Recent advances in fluorescence-colorimetric sensor arrays and their applications in biomedical fields

Abstract

Fluorescence colorimetric dual-mode sensor arrays have drawn significant attention in biomedical analysis. This is attributed to their remarkable advantages, such as high sensitivity, multi-parameter detection capability, and visible signal output. By integrating fluorescence sensing with colorimetric detection, these arrays enable multi-target detection in biological samples through changes in fluorescence intensity, ratiometric fluorescence, or color information. The continuous optimization of fluorescent nanomaterial-based probes has demonstrated their potential application in biological fluid detection, cancer biomarker analysis, metabolic/disease monitoring, and infectious disease screening. Furthermore, the integration of microfluidic technology, flexible electronics, and image analysis into the fluorescence-colorimetric sensor arrays (FCSAs) has enhanced the portability, sensitivity, and data-processing capabilities of the sensor arrays. This endows them with considerable potential for point-of-care testing (POCT) and telemedicine applications. This review summarizes the representative achievements of FCSAs over the past three years. First, the principles of array design are outlined, encompassing the sensing mechanism, methods for array construction, and data processing strategies. Then, the innovations and applications of FCSAs are highlighted for biomarker analysis and disease diagnosis. Finally, the challenges and future advances of this technology in biomedical fields are discussed.

---

1. Introduction

With the rapid development of personalized medicine, there is higher demand for the sensitivity and reliability of biomedical detection technologies. Traditional detection methods for biomarkers, such as enzyme-linked immunosorbent assay and mass spectrometry, generally suffer from limitations such as long detection cycles and strong dependence on specialized equipment. These limitations make it difficult to meet the requirements for clinical point-of-care testing (POCT) and high-throughput screening of complex samples.^1,2 In particular, in scenarios such as early disease diagnosis and routine health monitoring, there is an urgent need to develop novel analytical techniques that combine rapid response with multiplexing capability.

Fluorimetry is regarded as a highly promising technology due to its significant advantages, including rapid response and high sensitivity. The interaction between the target analyte and the fluorescent probe generates specific changes in fluorescence signals, which can be amplified as “fingerprint” features and read through spectral analysis or imaging systems.^3–5 Although fluorescence sensors have made significant progress in sensitivity and selectivity compared with traditional detection methods, they still have certain limitations. For instance, fluorescence signals are susceptible to interference from the autofluorescence of the sample matrix, photobleaching, variations in media pH, and ambient light, thereby compromising their detection stability and reliability. Meanwhile, a single fluorescence signal makes it difficult to distinguish different targets in complex biological systems, which limits its applicability in high-throughput detection and POCT testing (Fig. 1).^6,7

Fig. 1 A review of the research progress in FCSAs (all images are copyrighted and noted in each section of this article).

Fluorescence-colorimetric sensor arrays (FCSAs) were first developed by Rakow and Suslick in 2000 and have since been successfully applied to the identification of organic vapors.⁸ FCSAs are usually constructed by immobilizing multiple chemo/bio-responsive fluorescent probes on a substrate in specific geometric patterns, creating a sensing matrix capable of simultaneous multiplex multi-target detection. When exposed to biological samples, distinct probes in the sensor arrays undergo specific reactions with designated targets (e.g., glucose, inflammatory factors), generating unique changes in fluorescence intensity/wavelength and discernible colorimetric signals. By employing pattern recognition algorithms, these complex fluorescence-colorimetric responses are converted into classifiable feature profiles, enabling high-throughput analysis of targets.^9–11 This approach overcomes the inherent limitations of “one-to-one” specific recognition in traditional fluorescence sensors.

FCSAs integrate the high sensitivity of fluorescence detection with the intuitiveness of colorimetric visualization, achieving synchronous recognition and quantitative or semi-quantitative analysis of multiple targets in complex samples.¹² Furthermore, their capability of multi-channel signal output not only effectively circumvents false positives/negatives arising from cross-interference in a single-sensor system but also facilitates the dimensionality reduction of multi-dimensional data through pattern recognition algorithms. The detection throughput and result reliability can be enhanced by optimizing signal interpretation.¹³ In comparison with traditional single sensors, such as surface-enhanced Raman scattering (SERS) sensors, although FCSAs are somewhat less proficient in the specific recognition of single molecules, they exhibit more pronounced advantages in the simultaneous visual screening of multiple analytes. This makes them particularly well suited for the holistic and pattern-based rapid discrimination of complex systems. Furthermore, SERS typically relies on noble metal substrates and sophisticated optical equipment.¹⁴ In contrast, FCSAs can be more readily integrated with flexible sensing materials and low-power detection modules, enabling the miniaturization and lightweight design of devices.^15,16 When combined with artificial intelligence (AI) technology, the compatibility of FCSAs in complex biological matrices like serum and urine is further improved. This propels their crucial role in telemedicine, personalized health monitoring, and on-site analysis in resource-constrained regions.^17–19 By virtue of their multi-analyte simultaneous detection capabilities, low hardware requirements, and promising portability potential, FCSAs are better adapted to high-throughput and rapid screening scenarios at the primary care level and in field settings.

This review summarizes the recent advances in FCSAs in biomedical fields over the past three years, focusing on their applications in disease diagnostics and biomarker analysis. The key technological breakthroughs in data processing and optimization of sensing materials have been analyzed. Finally, this review discusses the future prospects of FCSAs in precision medicine and personalized health monitoring.

2. Design principles of FCSAs

Under certain excitation, electrons of fluorophores experience transition from the ground state to the excited state, and then return to the ground state through radiation, releasing fluorescence. Due to the distinct emission wavelengths and intensities of different fluorophores, the probes can interact with the target analytes through specific interactions, causing changes in the fluorescence signals. This forms the basis of the signal response of FCSAs. The colorimetric signal is typically analysed according to the Lambert–Beer Law, which is particularly suitable for portable devices and enables semi-quantitative detection with the naked eye.

The core design concept of FCSAs aims to achieve multi-target detection by integrating multiple sensing units and utilizing signal amplification and cross-validation strategies for fluorescence and colorimetric signals. Compared to the traditional single-mode sensors, FCSAs achieve enhanced specificity and visualization ability in detection by using different types of probe materials in their design, regulating their spatial arrangement, and combining data processing algorithms. The design principle of the FCSAs will be elaborated from three aspects: the generation of array signals, array construction, data acquisition and signal processing.

2.1 Signal generation mechanisms

2.1.1 Fluorescent materials. Diverse types of fluorescent materials exhibit distinct photophysical characteristics. Materials such as metal nanoclusters (NCs), quantum dots (QDs), carbon dots (CDs), and metal–organic frameworks (MOFs) have become the core components of FCSAs.²⁰ Metal-based nanomaterials possess characteristics such as surface plasmon resonance and photoluminescence, which enable their fluorescence behaviour to be tuned by adjusting parameters like nanoparticle size, surface functionalization, or metal–ligand bonding interactions. As shown in Fig. 2A, the Cyc/NAC-AuNC fluorescent probe developed by Noreldeen's group can generate different fluorescence responses based on π–π stacking interactions between thiol groups and vitamin B6 derivatives, coupled with machine learning-assisted recognition to realize quantitative analysis of different derivatives.²¹

Fig. 2 (A) Schematic illustration of the deep learning-based fluorescence sensor array for qualitative and quantitative analyses of VB6Ds. Reprinted with permission from ref. 21. Copyright 2022, American Chemical Society. (B) Schematic diagram of the synthesis steps of B-QD/G-QD/R-QD sensing systems. Reprinted with permission from ref. 22. Copyright 2022, Elsevier. (C) Schematic illustration of the preparation of Eu-BDC@FM. Reprinted with permission from ref. 25. Copyright 2023, Elsevier. (D) Schematic illustration of the fabrication of BSA-AuNCs@Eu-MOF-based ATP nanosensors. Reprinted with permission from ref. 26. Copyright 2022, Elsevier.

The quantum confinement effect confers upon QDs the characteristics of narrow-band emission and remarkable stability. Moreover, their fluorescence properties can be precisely modulated through surface functionalization strategies. Lu et al.²² synthesized QDs with tricolour emissions (B-QDs, G-QDs, and R-QDs) driven by electrostatic interactions, as shown in Fig. 2B. These QDs were employed for the quantitative analysis of Cu²⁺. The fluorescence intensity ratio demonstrated an excellent linear correlation with the concentration of Cu²⁺, and the detection limit was found to be less than 0.05 μM. Meanwhile, the recognition capabilities of CDs can be expanded through the design of surface functional groups. Li et al.²³ utilized citric acid and Congo Red as carbon sources to fabricate a sensor array composed of two types of CDs. In this system, the CDs bind to target proteins through electrostatic and hydrophobic interactions, resulting in fluorescence quenching. This approach enabled the accurate discrimination of four amyloid proteins and two serum proteins, thereby facilitating the preliminary cancer screening for liver cancer and breast cancer.

Metal–organic frameworks (MOFs) are crystalline porous materials formed through the self-assembly of metal ions and organic ligand templates. Their fluorescence mechanism predominantly relies on coordination chemistry and energy transfer processes.²⁴ Enzyme-free fluorescence sensors (Eu-BDC@FM) were prepared through electrospinning based on lanthanide metal–organic frameworks (Eu-BDC), as shown in Fig. 2C, for the detection of uric acid in human urine.²⁵ MOFs can couple with other fluorescent nanomaterials, such as AuNCs, to form composite probes. As shown in Fig. 2D, isophthalic acid absorbs the excitation light and transfers energy to Eu³⁺ in the AuNCs@Eu-MOF nanocomposite via the “antenna effect” to emit red fluorescence.²⁶ This probe was used for the detection of ATP in serum, achieving a detection limit as low as 25 μM.

By leveraging multi-channel fluorescence signals, colorimetric responses, and nanocomposite effects, the limitations of single-component materials can be overcome. Bhatt synthesized a carbon dots-copper fluorescent nanocomposite (CDs-Cu) using sucrose as the carbon source. This material was used for the quantitative analysis of creatinine in the urine of patients with chronic kidney disease.²⁷ A fluorescent probe was developed based on carbon dots/gold nanoclusters (CDs-AuNCs) for detecting iodide ions in urine, achieving a detection limit of 0.7 μM.²⁸ Similarly, Norouzi and colleagues anchored red-emitting carbon nanostructures (r-QCNSs) onto a molecularly imprinted Er-BTC metal–organic framework (MOF) to fabricate a paper-based fluorescent biosensor. This sensor exhibited a detection limit of 1.28 μM for dipicolinic acid (DPA), a biomarker of Bacillus anthracis, in urine samples. This finding further validates the potential of nanocomposites for use in the detection of clinical samples.²⁹

2.1.2 Sensing mechanism. When a target analyte interacts with a fluorescent probe, the fluorescence signal can be modulated via a variety of mechanisms, encompassing fluorescence quenching, fluorescence enhancement, and ratiometric fluorescence. Table 1 summarizes the commonly used sensing materials and sensing mechanisms in FCSAs.

Table 1 Fundamental sensing materials and sensing mechanisms of FCSAs

Biomarkers	Sensing materials	Sensing mechanism	Applications	Ref.
Amyloid	CDs	Fluorescence quenching	Serum	23
DA	ZGC/ZIF-8-NH2	Fluorescence quenching	Sweat, urine	24
Uric acid	Eu-BDC@FM	Fluorescence quenching	Urine	25
DPA	Er-BTC MOF/MIP-r-QCNSs	Fluorescence quenching	Water, urine	29
PPs	CDs	Fluorescence quenching	Serum	30
Cu^2+, Thiram	B-QDs, G-QDs, R-QDs	Ratiometric fluorescence	Water, urine, serum	22
HSA	r HSAp	Ratiometric fluorescence, 490 nm/580 nm	Urine	33
PH	TTA-UC/ET	Ratiometric fluorescence, 431 nm/491 nm	Blood	34
2,6-DPA	r-CDs/EY-ZIF-67	Ratiometric fluorescence, 573 nm/673 nm	Water, urine	35
ATP	AuNCs@Eu-MOF	Ratiometric fluorescence, 446 nm/616 nm	Serum	26
Creatinine	CDs-Cu^2+	Fluorescence enhancement	Urine	27
I^−	CDs-Au NCs	Fluorescence enhancement	Urine	28
Metal ions	PS-AuNCs	Multi-mechanism synergy	Water	32
Flavonoids	F1/F2/CTAB	Multi-mechanism synergy	Serum, urine	36
PPs	G-AuNCs, R-AuNCs	Multi-mechanism synergy	Serum, cell lysate	37
VB6 derivatives	Cys/NAC-AuNCs	Multi-mechanism synergy	Blood, urine	21

---

Fluorescence quenching is mainly caused by the energy or electron transfer between the quencher and the fluorophore. Taking Förster resonance energy transfer (FRET) as an example, Pushina et al. developed a fluorescence sensor with dual-fluorophores for detecting sugar levels in urine.¹⁷ This sensor achieved an “ON/OFF–OFF/ON” signal switch via FRET from the donor (tryptophan) to the acceptor (diol). In the absence of the target molecules, the donor (tryptophan) exhibited weak fluorescence while the acceptor (diol) emitted strong fluorescence. When sugars entered the system, the FRET process was suppressed by the separation of the acceptor from the phenylboronic acid group, to recover the donor fluorescence and to quench the acceptor fluorescence. Huang et al. constructed a three-signal sensing system. It utilized the significantly different binding capabilities between phosphate ions and three metal ions (Nd³⁺, Ag⁺, and Fe³⁺) to selectively quench the fluorescence of blue, green, and red-emitting carbon dots respectively, thereby achieving the discrimination of phosphate species in serum samples.³⁰

Fluorescence enhancement effects typically originate from intramolecular electronic reorganization and environmental perturbations. Zhou et al.³¹ achieved specific cysteine (Cys) detection by incorporating an electron-withdrawing group (3,5-bis(trifluoromethyl) benzenethiol). Upon Cys exposure, the probe exhibited significant near-infrared fluorescence enhancement, which was attributed to the reaction between Cys and the probe leading to the substitution of electron-withdrawing groups, resulting in intramolecular rearrangement to form high-fluorescent N-substituted products. Further mechanistic validation revealed that Cys addition significantly modulated the HOMO–LUMO energy gap, promoting intramolecular charge transfer (ICT) and amplifying the fluorescence signals. Dong's group³² developed an AuNC based fluorescent probe, where Cu²⁺ induced cluster aggregation, triggering a restriction of the intramolecular rotation (RIR) effect for signal amplification.

Ratiometric fluorescent probes typically exhibit distinct fluorescence dual-emission and quantify analytes by measuring intensity ratios at different emission wavelengths. For instance, Sarkar et al.³³ developed an intramolecular hydrogen bond-driven ratiometric probe (rHSAp). Upon binding with human serum albumin (HSA), the probe exhibited a 3.1-fold increase in the fluorescence intensity ratio at 490/580 nm, effectively resisting environmental interference. An up-conversion (TTA-UC) ratiometric probe based on triplet–triplet annihilation was created by utilizing PdOEP/DPA (palladium(II) octaethylporphyrin/9,10-diphenylanthracene) as the energy donor and pH-responsive p-nitrophenol (PNP) as the acceptor (Fig. 3A).³⁴ Variations in media pH value modulated the energy transfer efficiency between DPA and PNP, inducing ratiometric fluorescence changes for a precise detection of blood pH. The r-CDs/EY-ZIF-67 sensor developed by Norouzi et al. also operated on this mechanism. Taking the stable fluorescence of red-emitting carbon dots (r-CDs, with an emission wavelength at 673 nm) as a reference, it monitored the fluorescence changes of eosin Y (EY, with an emission wavelength at 573 nm). The detection of the anthrax biomarker 2,6-pyridinedicarboxylic acid (2,6-DPA) was achieved through the F₅₇₃/F₆₇₃ ratio. Even in complex matrices such as urine and tap water, it could still maintain high accuracy.³⁵

Fig. 3 (A) Diagram of the TTA-UC/ET probe for pH detection. Reprinted with permission from ref. 34. Copyright 2022 American Chemical Society. (B) Ratiometric probe working principle diagram. Reprinted with permission from ref. 36. Copyright 2024, Elsevier. (C) Schematic illustration of the construction of the AuNC-metal ion sensor array and discrimination of multiple PPs. Reprinted with permission from ref. 37. Copyright 2024, American Chemical Society.

Integrating multiple signal modulation mechanisms into the same sensor array can significantly expand detection dimensions. Fan et al.³⁶ developed a dual-probe system (F1/F2) that discriminates flavonoids in serum and urine via differential response patterns. Flavones and flavonols quenched fluorescence of the F1 probe, while isoflavones induced ratiometric changes in the F2 probe (Fig. 3B). Zhou's group³⁷ advanced this strategy and developed an array of gold nanoclusters regulated by metal ions for the simultaneous detection of multiple physiological phosphates (PPs). As shown in Fig. 3C, Zn²⁺ and Eu³⁺ complexed with AuNCs to quench the fluorescence of green-emitting AuNCs (G-AuNCs) while enhancing the emission of red-emitting AuNCs (R-AuNCs). When PPs were present, they triggered the disassembly of metal aggregates, restoring the fluorescence of the AuNCs. This array can distinguish eight PP species and quantify ATP in complex matrices with a detection limit of 5 μM.

2.2 Array fabrication strategies

When designing the spatial arrangement of a sensor array, each sensor unit must be able to interact strongly with the analyte. At the same time, its colorimetry or fluorescence changes should be closely correlated with the chemical interactions, and it is essential to ensure that there is no interference among the individual units. For example, a cross-sensing array was designed based on six chelates.³⁸ As depicted in Fig. 4A, the results can be detected either through visual inspection or by using a microplate reader in combination with pattern analysis, which enabled the accurate identification of four metal ions.

Fig. 4 (A) Photographs of the four developed ion-ligand arrays, as well as the CCSSA discriminant LDA plots. Reprinted with permission from ref. 38. Copyright 2021, Elsevier. (B) Process diagram for fluorescence compound array analysis of glucose in tear fluid using the random forest technique with a mobile device supporting machine learning. Reprinted with permission from ref. 44. Copyright 2023, Elsevier. (C) Schematic diagram of hydrogel microparticle colorimetric array chip design. Reprinted with permission from ref. 46. Copyright 2023, The Royal Society of Chemistry.

Standard 96-well plates, with their uniform microplate architecture, served as the primary platform for early array designs. Mitchell's group used a standard 96-well plate as the detection platform and developed a six-fluorescence-sensor array that combined fluorescence responses with multivariate analysis to discriminate platinum-based drugs at varying concentrations effectively.³⁹ However, such arrays are not suitable for application scenarios such as on-site or POCT detection. Inspired by the conventional urinalysis colorimetric strips,^40,41 a smart eye patch using qualitative filter paper as the sensing layer and textile fibre as the support substrate was engineered to enable simultaneous detection of pH, proteins, ascorbic acid, and glucose in tear.⁴² Microfluidic technology was further introduced into this kind of smart eye patch. Whatman filter paper was used as the chromogenic sensing layer, and a smartphone analytical system was combined to develop a microfluidic colorimetric patch for the simultaneous detection of multiple biomarkers in human tears.^43,44 Microfluidic technology can guide the analyte to flow into the detection zone along the predefined channels, effectively overcoming the cross-interference between paper-based sensor array units.

Compared to paper-based materials, hydrogels can effectively suppress the coffee-ring effect inherent to cellulose arrays and reduce the uneven distribution of analytes in the detection zones to enhance chromogenic stability and uniformity. Hydrogels have the advantages of moisturizing ability, good biocompatibility, and controllable physicochemical properties. Wang et al.⁴⁵ reported a novel fluorescence-colorimetric array using sodium alginate (SA) hydrogel membranes immobilized with multi-enzymes to realize highly sensitive visual detection of organophosphates. Microfluidics was integrated with hydrogel arrays to further improve the sensitivity and selectivity. De Masi et al.⁴⁶ fabricated a hydrogel microparticle colorimetric chip for detecting human immunoglobulin G (IgG) in biofluids (Fig. 4C). This system employed spatial compartmentalization to direct fluid flow, and the signal interference caused by the diffusion of analytes was effectively diminished, making the colorimetric response more stable and controllable. Furthermore, the swelling properties and directional channels of hydrogels could be precisely modulated to integrate with multi-channel microfluidic chips, which effectively circumvents cross-interference.⁴⁷ Alternatively, a double-layer fibrous cotton pad could be incorporated to reinforce the mechanical properties of the hydrogel.⁴⁸ While hydrogel-based sensor arrays offer significant advantages, challenges remain in long-term storage stability, mechanical robustness, and electronic integration. Future development may focus on novel composite materials to advance these applications.

Furthermore, within the realm of sensor array optimization, improving binding affinity constitutes one of the central focuses. A multitude of strategies have been substantiated as effective. For example, polymers featuring specific recognition sites are constructed via molecular imprinting technology to obtain molecularly imprinted polymers (MIPs). The porous architecture of these polymers gives rise to multiple non-covalent interactions with target molecules, thereby significantly enhancing the binding ability between the sensor and the target analyte.⁴⁹ Biomolecular functional modification also serves as a crucial approach. For instance, biological receptors such as aptamers and antibodies are immobilized onto the sensor surface. Aptamers are capable of specifically binding to target molecules through base-complementary pairing. Antibodies, on the other hand, capitalize on the high specificity of the “antigen–antibody” immune response to achieve strong-affinity recognition of low-concentration analytes within complex matrices.⁵⁰ Simultaneously, the manipulation of the material's microstructural features (such as the fabrication of porous or core–shell structures) can augment the sensor's specific surface area, expose a greater number of active binding sites, and further intensify the interaction with the analyte.⁵¹ Although MIPs showcase exceptional binding performance under ideal conditions, their recognition capabilities deteriorate significantly when applied to untreated biological samples. This is attributable to the fact that proteins and other components within biological samples can non-specifically occupy active sites. Moreover, traditional MIPs are predominantly synthesized in organic solvents or single-phase aqueous media. The hydrophobicity/hydrophilicity and charge characteristics of their imprinted cavities are ill-matched to the complex biological sample environment, rendering them prone to structural collapse or swelling, which in turn disrupts the spatial complementarity of the recognition sites.⁵² In addition to the aforementioned receptors, bioenzymes are also commonly employed in sensor arrays. Leveraging their catalytic specificity, the concentration variations of target analytes can be converted into quantifiable optical signals.

2.3 Data acquisition and processing

2.3.1 Data acquisition. Within FCSA platforms, the specific interaction between the probes and analytes elicits alterations in fluorescence intensity, spectral shifts, or observable colorimetry changes, thereby generating a distinct signal pattern. These signals make use of the optical diffraction and spectroscopy techniques of spectrometers. By means of gratings or interference filters, the fluorescence signals are decomposed into different wavelengths. The spectral information of the light intensity varying with wavelength is accurately recorded with the assistance of a charge-coupled device(CCD) or a photomultiplier tube (PMT).⁵³ Through the analysis of the peak wavelength, intensity, and emission bandwidth of the fluorescence spectrum, highly sensitive and high-resolution detection of the target object can be realized. Simultaneously, the visual colorimetric information of the sensor array is typically captured with a camera or a smartphone. Subsequently, the color changes are extracted and analyzed using image-processing software (such as Adobe Photoshop, Image J, etc.).⁵⁴ Nevertheless, the detection environment and the excitation light often interfere with camera acquisition. Moreover, smartphone cameras of different models exhibit differences in color reproduction, exposure control, and sensor sensitivity. This may result in deviations between the collected data and the actual optical signals.

To faithfully reproduce the colorimetric signals, optical optimization approaches can be implemented, including the use of filter wheels or adjustable filters and the employment of light-tight detection chambers. This can effectively minimize the interference from excitation light and ambient light.^55,56 For instance, Binabaji et al.⁵⁷ installed a 400 nm optical filter between the sensor and the smartphone. In combination with a light-shielding device for a 375 nm UV lamp and a multi-channel microfluidic device, the interference of excitation light and ambient light was mitigated. Furthermore, the characteristic information of fluorescence arrays can be converted into optoelectronic signals.⁵⁸ By leveraging integrated chips to amplify and process these electrical signals, there is strong potential for enhancing the accuracy of portable detection devices and realizing a diverse array of functions.

2.3.2 Data processing. The multimodal data generated by FCSAs encompass fluorescence intensity, emission wavelength shifts, time-resolved decay curves, and colorimetric RGB values. Among these, the colorimetric RGB values, serving as the core signal of digital images, despite having a lower dimensionality compared to spectral data, are still influenced by factors such as variations in sensor fabrication and environmental light interference. Additionally, the non-linear relationship between color intensity and analyte concentration needs to be alleviated through pre-processing. For example, the normalized R value can be utilized.⁵⁹ Alternatively, difference spectra can be employed, which involves subtracting the RGB values of the blank sample from those of the test sample, and then establishing a linear relationship via piecewise fitting.⁵⁷ Moreover, the RGB values can be transformed into other color space values. For instance, by converting RGB to HSV (H: Hue, S: Saturation, V: Value). Subsequently, linear relationships are established using the respective change amounts, namely ΔH, ΔS, and ΔV.⁴⁴ The values are calculated according to eqn (1)–(3):

ΔH = H_(after) − H_(before) (1)

ΔS = S_(after) − S_(before) (2)

ΔV = V_(after) − V_(before) (3)

Regarding high-dimensional fluorescence data, advanced algorithms are employed to extract features and establish distinctive response libraries, typically involving techniques such as principal component analysis (PCA), linear discriminant analysis (LDA), and hierarchical cluster analysis (HCA).^60,61

PCA is an unsupervised dimensionality reduction technique. It maps high-dimensional data onto a low-dimensional orthogonal space via orthogonal transformation. In the computational process, the data are first centered. Then, the covariance matrix Σ of the centered data is computed. As depicted in eqn (4), X′ represents the centered data matrix, and n represents the number of samples.

Eigenvalue decomposition on Σ is performed to determine the directions of the principal components. The essence of PCA lies in maximizing the data variance and projecting the data onto a low-dimensional space. Specifically, the first principal component preserves the maximum variance, and the variances of subsequent principal components decrease in turn, aiming to retain as much information as possible. LDA is a supervised classification algorithm, the goal of which is to maximize the inter-class differences and minimize the intra-class differences. Specifically, LDA constructs the between-class scatter matrix (S_B) and the within-class scatter matrix (S_W). By solving for the projection direction that maximizes the value of J(W), it achieves optimal classification. Its expression is given in eqn (5) (where W is the projection matrix):

HCA is an unsupervised clustering approach. The core principle is to quantify the degree of association among samples by leveraging distances (e.g. Euclidean distance, Manhattan distance) or similarities (e.g. correlation coefficients). Grounded in the similarity among samples, HCA iteratively calculates these distances to successively merge or split samples, thereby constructing a hierarchical clustering outcome.⁶²

Eqn (6) shows the formula for calculating the Euclidean distance, and eqn (7) shows the formula for calculating the Manhattan distance. Here, d denotes the dimension of the sample features, and the dependent variable d(x_i, x_j) represents the distance between samples x_i and x_j. The smaller the value of d(x_i, x_j), the greater the similarity between the samples. In the practical processing of fluorescence data, PCA can be initially employed to project the multi-channel high-dimensional data onto a low-dimensional space. This step serves to extract the principal components, thereby reducing the complexity of the data. Subsequently, HCA can be utilized to classify similar samples. This approach is adequate for screening and differentiating various samples.^63,64 Meanwhile, visualization analysis should be conducted in conjunction with relevant charts.⁶⁵

Applying deep learning models (e.g., convolutional neural networks, graph neural networks) to extract latent features can effectively mitigate errors and achieve the fitting of high-dimensional data.⁶⁶ During the data processing stage, these approaches are capable of automatically learning the spatial and spectral characteristics of images, thereby enhancing the accuracy of classification and recognition. By integrating with technologies such as software development or WeChat mini-programs,⁶⁷ the detection platform can be incorporated into smartphones, facilitating convenient data analysis and remote monitoring.

3. Biomedical applications of FCSAs

The emergence and development of pathological states typically do not occur as isolated physiological anomalies. Instead, they are concomitant with biomolecular metabolic dysregulations across multiple systems and dimensions in the whole body. These dysregulations can manifest as specific alterations in carriers, such as respiratory gases, cutaneous secretions, and body fluids. FCSAs can enable early detection of conditions such as cancer, infectious diseases, and metabolic disorders by analyzing these alterations.^68,69 Furthermore, applications of FCSAs in environmental and food safety monitoring are increasingly prevalent, including the detection of heavy metal ions and organic pollutants in aqueous systems, as well as antibiotic residues in food products.^70–72 This section focuses on elucidating the technological applications and innovations in the detection of biomarkers and disease diagnostics.

3.1 Detection of disease biomarkers

3.1.1 Detection of urine. Serving as a critical excretory route for metabolic waste and electrolytes, the composition of urine reflects renal function, endocrine system status, and metabolic conditions.⁷³ Traditional colorimetric test strips employ immobilized enzymes or chromogenic agents (e.g., sodium nitroprusside for ketone detection) to achieve semi-quantitative analysis of urine. However, they suffer from limited sensitivity and crosstalk interference between adjacent reaction zones.

Iodide ions (I⁻) play a critical role in the regulation of thyroid function, and the abnormal concentration of I⁻ implies hypothyroidism or thyroid cancer.⁷⁴ Although traditional methods such as atomic absorption spectroscopy and ion chromatography possess high precision, they suffer from complex operation, high cost, and time-consuming procedures. A fluorescent-colorimetric probe based on cyan-emitting carbon dots and red-emitting gold nanoclusters was developed to achieve dual-mode detection of I⁻ through fluorescence quenching and colorimetric response assisted by mercury ions (Hg²⁺).²⁸ The method demonstrates a detection limit as low as 0.70 µM and enables portable on-site testing via test strips, achieving recovery rates of 88.9% to 112.7% in urine samples. Given the complex anions in urine, Yang et al. refined an anion-sensing strategy by designing a cucurbit[8]uril (Q[8])-based host–guest fluorescent probe array.⁷⁵ This system integrated anion-induced fluorescence quenching with an LDA algorithm, achieving the detection of five distinct anions with 100% classification accuracy.

Urinary tract infection (UTI) is a common disease caused by diverse pathogens, with an incidence rate of 2% to 15% among pregnant women. Delayed diagnosis and treatment may elevate risks of adverse pregnancy outcomes including pyelonephritis, preterm birth, and low birth weight.⁷⁶ Conventional UTI detection methods, such as urine culture, fluorescence quantification, and polymerase chain reaction (PCR), have disadvantages of complex operation, time-consuming procedures, and reliance on specialized equipment. Li's group engineered a paper-based fluorescence-colorimetric sensing platform for the rapid detection of PCR amplicons by preloading fluorescein, 3,3′,5,5′-tetramethylbenzidine (TMB) and buffer solution onto a paper tray.⁷⁷ This design attained 99.4% concordance with standard PCR, surpassing that of the traditional urine culture (96.7%). Furthermore, the method demonstrated high accuracy and reliability when applied to 200 clinical urine samples.

3.1.2 Detection of sweat. Sweat, a filtrate of blood plasma, primarily consists of water, electrolytes, metabolites, and trace proteins. A fluorescent probe using a CsPbBr₃ perovskite nanocrystal modified by amphiphilic polymer ligands showed a significant blue shift in emission wavelengths when responding to various concentrations of Cl⁻, achieving visual detection of Cl⁻ in sweat samples.⁷⁸ Jia and Yan⁷⁹ constructed a ratiometric fluorescence sensor based on europium metal–organic frameworks (Eu-NDC) which emitted Eu³⁺ characteristic emission and ligand fluorescence. This system can real-time monitor L-lactate in sweat, providing novel strategies for hypoxia warning in sports medicine. These studies not only deepen the understanding of analysis technologies for sweat components, but also lay the groundwork for developing wearable health monitoring devices.

Sweat secretion is significantly influenced by physiological status, temperature, and physical activity, posing dual challenges of complex matrix interference and dynamic secretion for in situ monitoring. Wang et al. integrated fluorescent gold-copper bimetallic nanoclusters (AuCuNCs) with green-emitting fluorescein isothiocyanate-modified cellulose nanofibers (F-CNF), to construct a hydrogel sensor with ratiometric fluorescence response to urea (Fig. 5A).⁸⁰ Under urease catalysis, the hydrolysis of urea increases the media pH, causing attenuation of AuCuNCs’ red emission and enhancement of F-CNF green fluorescence. Additionally, the hydrogel encapsulates pH-responsive bromothymol blue (BTB), which transitions from yellow to blue upon the presence of urea. The BAF-CPu hydrogel adheres directly to the skin for sweat absorption, avoiding complex sampling procedures. This sensor enables in situ sweat detection while mitigating interference from the complex sweat matrix, achieving a detection limit as low as 0.19 μM for urea. Building on this foundation, Quan et al. developed a pH-responsive composite gel film sensor (Fig. 5B).⁸¹ Under H⁺ variation, the film exhibited fluorescence quenching and room-temperature phosphorescence (RTP) enhancement, facilitating the dual-mode monitoring of sweat pH value. Under weakly acidic sweat conditions, the composite gel film exhibits characteristic fluorescence quenching and RTP intensification. Beyond sensing applications, this composite material shows potential for information encryption and anti-counterfeiting. By engineering stimulus-responsive multiplexed patterns, it enables sweat-based anti-counterfeiting implementations.

Fig. 5 (A) Relationship between film fluorescence intensity and sweat pH. Reprinted with permission from ref. 80. Copyright 2023, John Wiley and Sons. (B) Wearable hydrogel patch for in situ urea detection in sweat. Reprinted with permission from ref. 81. Copyright 2024, Elsevier. (C) Production process diagram of a degradable sweat glucose sensor based on paper-based and hydrophilic cotton threads. Reprinted with permission from ref. 82. Copyright 2024, Elsevier. (D) Structure illustration of the smartphone-based fluorescence microfluidic-chip for glucose detection. Reprinted with permission from ref. 83. Copyright 2025, Elsevier.

There is a certain correlation between the concentration of glucose in sweat and blood sugar levels. As blood sugar increases, the concentration of glucose in sweat will also increase accordingly. The Kansay group synthesized boric acid-functional CQDs from fruit extracts through a one-step hydrothermal method and developed a flexible, wearable, biocompatible and degradable sweat glucose sensor by combining a paper-based analysis device with hydrophilic cotton thread microfluidic channels.⁸² Li et al.⁸³ designed a microfluidic chip containing six microchambers (Fig. 5D). Each microchamber was equipped with a glucose sensing membrane, which utilized glucose oxidase (GOD) to catalyse the consumption of O₂ and the generation of H₂O₂ during the glucose oxidation process. The chip generated double fluorescence signals of green (H₂O₂-responsive) and red (O₂-sensitive) emission under single-wavelength excitation. By simultaneously calibrating glucose with R and G channel signals, this dual monitoring mode enhanced the accuracy and reliability of the detection. The chip incorporates miniaturization, high-resolution signal acquisition, wireless transmission, low sample consumption, and smartphone integration, providing novel tools for personalized diabetes health management.

3.1.3 Detection of tears. Tears not only lubricate the eyeball and remove external pollutants, but also contain various biomarkers, such as glucose, electrolytes, inflammatory factors, enzymes, and antibodies. Changes in their composition can be used to evaluate ocular and systemic diseases. Diabetes patients exhibit significantly higher tear glucose concentrations than healthy individuals, showing a positive correlation with blood glucose levels.⁸⁴ Deng et al.⁸⁵ immobilized a glucose fluorescent probe and a reference fluorescent dye within the hydrogel network of contact lenses. As shown in Fig. 6A, the fluorescence colorimetry of the lenses shifts from pink to blue with increasing glucose concentration. Smartphones can quantitatively determine glucose levels through RGB signals, achieving a detection limit of 9.3 μM. Diabetic retinopathy (DR), a common microvascular complication of diabetes, can be effectively screened using the separable immunofluorescence sensor tear analysis chip developed by Das et al.⁸⁶ This innovation overcomes the limitations of traditional detection methods with an accuracy 99%.

Fig. 6 (A) The schematic illustration of smart contact lenses for monitoring glucose. Reprinted with permission from ref. 85. Copyright 2022, Elsevier. (B) Portable optical platforms (lateral flow detection, capillary tube detection and contact lens detection) for lactoferrin fluorescence sensing in tears. Reprinted with permission from ref. 87. Copyright 2023, Elsevier. (C) Development of the wearable eye sensor patch and portable imaging platform. Reprinted with permission from ref. 55. Copyright 2022, American Chemical Society.

Lactoferrin is a protein with antibacterial, anti-inflammatory, and immunomodulatory functions, and is associated with various diseases such as dry eye syndrome, Sjögren's syndrome, and Alzheimer's disease. Shi et al.⁸⁷ utilized terbium(III) chloride (TbCl₃) as a fluorescent probe, integrating the probe across multiple platforms including nitrocellulose membranes, capillaries, and contact lenses. As shown in Fig. 6B, this system enables on-site rapid detection of lactoferrin. The platform presents excellent linearity within the lactoferrin concentration range of 0–5 mg mL⁻¹. Compared to traditional electrochemical and spectroscopic detection methods, the system is easy to operate and economical.

Beyond detecting ocular-related diseases, some eye patches can also detect residual pharmaceutical levels in tears. Yin et al.⁵⁵ designed a wearable eye patch integrated with a quadruple nanosensor chip (Fig. 6C), achieving continuous monitoring of fluoroquinolone antibiotic residues in tears via a smartphone platform. Tear analysis sampling is non-invasive and painless, which can avoid the discomfort and infection risks of traditional invasive testing, facilitating early disease screening and drug tracking. However, the composition of tears is easily affected by the environment and sampling methods. The detection sensitivity of low-concentration markers is insufficient, and complex matrices may also interfere with the results. Therefore, it is necessary to optimize the technology to break through the bottleneck.

3.1.4 Detection of breath. Some diseases can be detected by assaying breath biomarkers. For example, in patients infected with Helicobacter pylori, the increased urease activity in the stomach causes an excessive amount of urea to be decomposed into ammonia, which is then excreted through respiration. In a study, Yang et al.⁸⁸ developed a glass cellulose membrane modified with phospholipid-coated perovskite quantum dots (PM-CsPbBr₃@GCM). By utilizing the substitution reaction between ammonia and the quantum dots, which leads to fluorescence quenching and a color change from yellow to colorless, dual-channel fluorescence-visual detection of exhaled ammonia was achieved. Likewise, patients with abnormal liver and kidney functions also experience elevated levels of exhaled ammonia due to abnormal urea metabolism.^89,90 In the exhaled breath of lung cancer patients, the concentration of aniline is abnormally high, which can be used as an important biomarker for lung cancer diagnosis. Gao et al.⁹¹ designed a fluorescence sensing film of o-carborane-modified pyromellitic diimide derivative (CB-PMI). Through the electron-transfer process between aniline and CB-PMI, fluorescence quenching was triggered. This sensing film can achieve an ultra-low detection limit of 0.1 ppt even under an environment of 100% relative humidity and shows high selectivity towards other interfering respiratory gases (such as ethanol, dichloromethane, etc.). The above-mentioned studies have verified the feasibility of fluorescence detection of respiratory biomarkers. But current research on colorimetric detection of respiratory biomarkers still has obvious limitations, such as interference from complex gas mechanisms, and a universal technical framework has not yet been established. Particularly in the clinical scenario, the convenience and low-cost advantages of colorimetric detection are more in line with the requirements of primary healthcare. In the future, further breakthroughs in material design and detection mechanisms are required to promote the practical application and popularization of colorimetric detection of respiratory biomarkers.

3.1.5 Cancer diagnostics. Cancer is a disease that involves multiple internal and external factors, including genetics, environment, lifestyle and infection, and is caused by gene mutations and abnormal cell proliferation. Patients typically exhibit a spectrum of physiological, biochemical, and metabolic characteristics, such as specific compounds in exhaled breath gas or abnormal molecular levels in body fluids.

Detection of ovarian cancer faces challenges due to its inconspicuous early symptoms and the lack of effective screening methods. Current clinical diagnosis for ovarian cancer primarily relies on traditional approaches, including imaging examinations, tests of tumour biomarkers in serum, and histopathological analysis. Kim et al. proposed an innovative solution for ovarian cancer detection.⁹² They developed quantum-defect-modified single-walled carbon nanotube sensors (OCC-DNA), where ssDNA encapsulation could improve biocompatibility. By integrating fluorescence spectral characteristics with machine learning algorithms, this platform achieved highly sensitive and specific detection of high-grade serous ovarian carcinoma (HGSOC), outperforming conventional serum biomarkers methods. A fluorescence sensor array based on gold nanoclusters functionalized with histidine (AuNCs@His) was designed, and it showed response to reactive oxygen species (ROS).⁵³ As shown in Fig. 7A, leveraging cancer cell-specific endocytosis and differential fluorescence quenching effects toward six ROS types, this system enabled not only discrimination of cancer cell subtypes but also precise identification of cell proliferation states through principal component analysis. Its uniform surface properties significantly enhanced diagnostic consistency. Compared to previously reported methods, this sensor array demonstrated three key advantages of similar surface characteristics, consistent endocytosis rates, and superior diagnostic efficiency.

Fig. 7 (A) Schematic diagram of precise diagnosis of cancer via an ROS-responsive fluorescence sensor array based on pH controlled multicolour histidine-templated gold nanoclusters. Reprinted with permission from ref. 53. Copyright 2023, American Chemical Society. (B) The linear relationship for varying quantities of SKBR3 cells. Reprinted with permission from ref. 68. Copyright 2025, Springer Nature. (C) Illustration of the signal capture device that integrates a smartphone with a capillary. Reprinted with permission from ref. 93. Copyright 2025, American Chemical Society.

Similarly, significant progress has also been made in nanozyme-based multimodal detection for other cancers. Xue et al.⁶⁸ constructed a Pt/DMSN nanozyme platform that was functionalized with HER2 monoclonal antibodies and aptamers to realize specific recognition of HER2-positive breast cancer cells (Fig. 7B). By integrating TMB chromogenic reactions with fluorescence imaging, this system could detect as few as 50 cancer cells within 30 min and effectively distinguished breast cancer subtypes (e.g., luminal A and triple-negative type). The core innovation lay in the synergistic effect between platinum nanoclusters and dendritic mesoporous silica nanoparticles (DMSNs), which not only enhanced peroxidase-like activity but also enabled target recognition through dual-fluorescent probes. This work provided a highly sensitive tool for early subtyping diagnosis of breast cancer.

Another study focusing on prostate-specific antigen (PSA) detection was developed by Liu's team.⁹³ A ratiometric fluorescence/colorimetric dual-mode sensing platform using Fe–N–C nanozymes integrated with red-emissive carbon quantum dots (R-CQDs@Fe-NC) was prepared, and aptamers were adsorbed on the probe to regulate the competitive catalytic activity and fluorescence signaling. As shown in Fig. 7C, coupled with smartphone imaging for portable analysis, the platform achieved detection limits of 0.054 ng mL⁻¹ (fluorescence mode) and 0.16 ng mL⁻¹ (colorimetric mode).

These studies have jointly demonstrated the universal advantages of nanomaterials in the detection of multiple cancer biomarkers. By integrating catalytic, targeting, and signal transduction modules within multifunctional nanocomposite systems, they overcome the sensitivity limitations of traditional single-mode detection, establishing a scalable technological paradigm for precision cancer diagnostics and treatment.

3.1.6 Neurodegenerative disease detection. Neurodegenerative diseases are a type of chronic conditions characterized by progressive degeneration, denaturation, and death of neurons. These disorders are typically irreversible with no current curative treatments.⁹⁴ Catecholamines (CAs) are a class of bioactive compounds containing catechol and amine groups such as dopamine (DA), norepinephrine (NE), and epinephrine (E). Their abnormal concentrations are closely related to neurodegenerative diseases and cardiovascular lesions. Due to the molecular structural homology of CAs, conventional detection methods suffer from poor selectivity. Recent advances in sensing strategies for CAs have overcome this limitation through innovative signal design and material engineering. Jia et al.⁹⁵ pioneered a fluorescence turn-on approach to discriminate distinct CAs using the reactions of ethylenediamine (EDA) with CAs to generate differentially coloured fluorescent nanoparticles. By integrating internal reference calibration with gold nanoclusters and carbon dots, they constructed a ratiometric fluorescence sensor array with high precision (Fig. 8A), which enables the rapid discrimination of CAs in complex biological samples.

Fig. 8 (A) Scheme for the discrimination of CAs and single/dual-emission fluorescence arrays. Reprinted with permission from ref. 95. Copyright 2021, Elsevier. (B) CuNC-based catechol oxidase nanozymes to build tri-probe colorimetric sensor arrays for the determination of multiple catecholamine neurotransmitters. Reprinted with permission from ref. 96. Copyright 2023, Elsevier.

Copper ions exhibit unique catalytic potency in catecholamine (CA) detection. Yin's group regulated the surface activity of copper nanoclusters (CuNCs@TA, CuNCs@AA, and CuNCs@PMAA) through ligand engineering to simulate the characteristics of catechol oxidase.⁹⁶ These nanoclusters catalysed the conversion of CAs to quinones, triggering the distinct chromogenic reaction. The sensor array was constructed based on the catalytic reaction to accurately identify target substances such as adrenaline and dopamine and their complex mixtures, with a detection limit as low as 10⁻⁸ M. Successful discrimination of four catecholamines was demonstrated in spiked biological samples. Copper-based nanozymes (Cu-MNs) were developed by Li et al.⁹⁷via one-pot synthesis to enable the colorimetric-fluorometric dual-mode detection of 2,4-dichlorophenol and epinephrine. Leveraging synergistic signal amplification, detection limits reached 0.91 μM (colorimetric mode) and 1.67 μM (fluorometric mode).

Conventional methods for DA detection rely on invasive blood analysis, which involves complex procedures and high costs. Pan et al.²⁴ developed a composite sensor comprising fluorescent nanoparticles and modified mesoporous metal–organic frameworks (MOFs). Under UV excitation, the composite exhibited purple emission, with fluorescence intensity progressively quenching as DA concentration increased (Fig. 8B). The sensor established a novel pathway for non-invasive clinical monitoring. It possessed excellent selectivity and sensitivity of DA detection in urine and sweat matrices, with a detection limit of 0.075 μM. In contrast to urine and sweat detection, saliva-based detection offers more prominent application advantages in the field of non-invasive diagnosis, primarily owing to its more convenient sampling procedure. In a recent study,⁹⁸ the research team led by Giuseppe developed artichoke-extract-synthesized fluorescent carbon nanoparticles (CNPs-ART). The catechol groups on the surface of these nanoparticles can bind specifically to DA. When using a smartphone as a detector in simulated saliva, the detection limit could reach as low as 100 pM. Moreover, CNPs-ART exhibited high selectivity towards common interfering substances in saliva and could be reused up to five times. Nevertheless, the issue of interference from the complex saliva matrix, including mucoproteins and enzymes, on the chromogenic signal has not been fully resolved. And the scarcity of multi-center clinical data restricts its widespread implementation in primary healthcare settings. Furthermore, the team engineered a functionalized PAMAM dendrimer-based fluorescence sensor array to suggest a potential solution for the early diagnosis of Alzheimer's disease (AD). Integrated with machine learning algorithms, this platform discriminated the pathological aggregation states of Aβ40 and Aβ42 while maintaining high specificity in complex biological matrices including serum and cerebrospinal fluid.⁹⁹ The sensor array was capable of accurately differentiating 11 proteins with robust anti-interference capacity, demonstrating the significant clinical utility.

3.2 Personalized medicine and POCT

The emergence of flexible sensors has facilitated the advancement of wearable devices, transitioning health monitoring from conventional single-use intermittent testing to continuous real-time monitoring.^100–102 This paradigm shift is driven by the advances in flexible electronics and fluorescent nanomaterials, which enables sensors to adapt to dynamic physiological environments.^103,104 Integrating fluorescence sensors with flexible substrates, such as nanofibers, hydrogels, and biocompatible polymers, can significantly enhance the mechanical flexibility, wearable comfort, and long-term stability.^105,106 The polypeptide composite hydrogel sensing platform developed by Qiao et al. for detecting biomarkers in sweat had superhydrophilicity, which enables it to effectively repel lipid contaminants in sweat (Fig. 9A). Moreover, its robust mechanical properties and self-healing capability ensure long-term wearing stability.¹⁰⁷

Fig. 9 (A) Operation of the user-friendly home diagnosis system. Reprinted with permission from ref. 108. Copyright 2023, Elsevier. (B) Schematic drawing and images of the wearable plasmonic microneedle sensor. Reprinted with permission from ref. 109. Copyright 2024, Elsevier. (C) Schematic illustration of the fabrication process of fluorescent MN patches. Reprinted with permission from ref. 110. Copyright 2024, American Chemical Society. (D) Diagram illustrating operation of the microfluidic chip. Reprinted with permission from ref. 112. Copyright 2022, Elsevier. (E) Miniaturized multicolour fluorescence imaging system integrated with a PDMS LGP. Reprinted with permission from ref. 113. Copyright 2023, Springer Nature. (F) Schematic diagram of a microfluidic platform based on droplets. Reprinted with permission from ref. 114. Copyright 2024, American Chemical Society.

The minimal dimensions of microneedles allow penetration through the stratum corneum to direct contact with dermal interstitial fluid (ISF), offering virtually painless operation that significantly enhances user comfort. Sang et al. developed biodegradable microneedle arrays integrated with fluorescent sensors, where glucose concentration changes in the ISF were continuously monitored via fluorescence intensity measurements (Fig. 9B).¹⁰⁸ The microneedles need not be removed after use as the biodegradable material can be absorbed in the body. Similarly, Xiao et al. created a microfluidic-based plasmonic microneedle biosensing platform (Fig. 9C). ISF was extracted under negative pressure using a hollow microneedle array, and ultrasensitive detection of uric acid was achieved by surface-enhanced Raman scattering (SERS) of a three-dimensional gold nanoarray, with a detection limit as low as 0.51 µM. Portable real-time analysis was achieved by coupling the hollow microneedle array with a handheld Raman spectrometer.¹⁰⁹ Tabatabaee et al. designed skin tattoo-like epidermal sensors based on microneedle arrays.¹¹⁰ Integrated with wearable optoelectronic readers, these devices can be utilized for long-term painless monitoring of bilirubin, for smart electronic diagnosis at home, or therapeutic monitoring for neonatal jaundice management.

Microfluidic technology can precisely control the flow of liquids, facilitating the ultralow-volume sample processing and the synchronous detection of multiple components. Its modular design allows seamless adaptation to diverse requirements of biological analyses, such as large and small molecules and cells.¹¹¹ For instance, a microfluidic chip platform was developed for the detection of serum procalcitonin and optimized the interaction between analytes and fluorescent probes according to magnetic, thermal, or photonic mechanisms to achieve automated analysis (Fig. 9D).¹¹² Shin et al. presented a miniaturized multicolour fluorescence imaging system integrated with PDMS light-guiding plates (Fig. 9E). Coupled with CMOS image sensors and multi-wavelength LEDs, this platform achieved dynamic monitoring of green/red fluorescently labelled cells within incubators, demonstrating the potential of microfluidic-optical integration for real-time bioimaging.¹¹³ Furthermore, the automated droplet platform was designed with microfluidic technology that could simulate pan-proteomic alterations under pathological conditions, enabling protein identification in physiological buffers and human serum (Fig. 9F).¹¹⁴

However, it is essential to note that despite the remarkable performance demonstrated by the aforementioned FCSAs technologies in laboratory research, the majority of them have not yet been incorporated into routine hospital clinical applications. Upon closer examination, several key challenges have emerged as the primary obstacles. Firstly, the issue of interference from complex biological matrices remains incompletely resolved. Substances such as proteins and lipids in real samples tend to induce non-specific adsorption, thereby compromising the accuracy of detection. Secondly, there is a pressing need to enhance the long-term stability and batch-to-batch reproducibility of sensors. In the context of medical practice, equipment is expected to maintain consistent performance over several months or even years. Thirdly, the absence of standardized validation procedures and substantial clinical data support poses a significant hurdle, making it arduous to meet the requirements for regulatory approval. Furthermore, the levels of system integration and automation necessitate optimization to align with the efficient and standardized workflows within hospitals. Although these technologies hold great promise for future medical monitoring, translating them from the laboratory bench to clinical settings demands significant breakthroughs in areas such as material stability, standardization development, and clinical validation.

4. Conclusions

FCSAs have evolved from single-mode detectors to integrated analytical platforms, which can simultaneously achieve fluorescence emission tracking, chromaticity quantitative analysis, and pattern recognition based on machine learning. This evolution, achieved by optimized probe spatial organization and nanoscale interface engineering, has enabled multi-channel sensing with sub-picomolar sensitivity in complex biofluids. Current results in FCSA research demonstrate robust performance in the detection of serum and tissue homogenates, notably achieving multiplexed target discrimination based on differential quenching/ratiometric responses across different probe elements. Although FCSAs have demonstrated excellent performance in laboratory research, they still encounter numerous crucial challenges in practical applications. These challenges primarily encompass limitations in material stability, selectivity, sensitivity, reproducibility, and long-term performance. Some fluorescent probes are prone to photobleaching or degradation under long-term illumination or in complex biochemical environments, which can compromise the reliability of the signal. In complex sample matrices, probes may exhibit cross-responses to non-target analytes, thereby interfering with specific recognition. Despite having achieved sub-picomolar detection limits, the detection of extremely low-concentration targets is still vulnerable to background noise. The preparation variations among sensor batches and the fluctuations in environmental conditions make it arduous to standardize the signal output. Moreover, during continuous or repeated use, ensuring the stability of the sensing interface and the consistency of the signal remains a formidable task. In light of these issues, several novel solutions have come to the fore. These encompass the employment of anti-fouling probe coatings featuring orthogonal response mechanisms, the integration of microfluidic sample pre-processing modules, and the training of convolutional neural networks using spectral-colorimetric fusion datasets. The integration of flexible photonics, edge-computing chipsets, and clinical validation frameworks is propelling the practical development of FCSAs. Moreover, FCSAs hold significant promise in diverse fields such as environmental monitoring, food safety inspection, and security applications. To expedite the clinical translation and commercialization of FCSAs, the following aspects merit particular attention: (1) the development of self-calibrating arrays equipped with embedded reference channels, (2) the establishment of standardized signal acquisition protocols tailored to different environmental conditions, and (3) the realization of a co-design approach for low-power electronic devices and disposable sensor cartridges. By intensifying interdisciplinary collaborations with cutting-edge technologies, including organ-on-a-chip platforms and 5G-enabled telemedicine, the practical implementation of FCSAs within decentralized healthcare systems can be effectively promoted. This will ultimately facilitate the transition from laboratory research to clinical practice.

Author contributions

X-WX designed and supervised this work. S-LW collected the data, drafted the manuscript, and created the figures. L-JN and C-XX edited and revised this review. All authors have read and approved the final manuscript.

Conflicts of interest

There are no conflicts to declare.

Data availability

The authors confirm that all the data corroborating the outcomes of this deliberation are available within this review, and no new code, software, or research outputs have been analysed or included as part of this review.

Acknowledgements

This research was financially supported by the Innovation Project of Graduate Education of Guilin University of Electronic Technology (No. 2025YCXS210) and the Innovation Project of Guangxi Graduate Education (No. YCSW2025344), the National Natural Science Foundation of China (No. 61761013), the Guangxi Human Physiological Information Non-Invasive Detection Engineering Technology Research Centre, the Guangxi Colleges and Universities Key Laboratory of Biomedical Sensors and Intelligent Instruments, and the Guangxi Key Laboratory of Metabolic Reprogramming and Intelligent Medical Engineering for Chronic Diseases.

References

1. J. Cabo and J. Favresse, Clin. Chim. Acta, 2023, 550, 117534.
2. L. Marques, B. Costa, M. Pereira, A. Silva, J. Santos, L. Saldanha, I. Silva, P. Magalhães, S. Schmidt and N. Vale, Pharmaceutics, 2024, 16, 332.
3. Z. Yan, Y. Cai, J. Zhang and Y. Zhao, Measurement, 2022, 187, 110355.
4. Y. Hang, A. Wang and N. Wu, Chem. Soc. Rev., 2024, 53, 2932–2971.
5. D. Zhu, A. Ren and L. Xue, Org. Biomol. Chem., 2024, 22, 9113–9120.
6. L. Zhang, H. Wang, H. Zhang, N. Zhang, X. Zheng, W. Li, X. Qiu and D. Yu, Sens. Actuators, B, 2022, 369, 132272.
7. H. Kim, S. Lee, J. S. Min, E. Kim, J. Choi, J. Ko and E. Kim, Dyes Pigm., 2021, 193, 109492.
8. K. S. Suslick and N. A. Rakow, Nature, 2000, 406, 710–713.
9. Z. Xu, Y. Zhan, S. Zhang, Z. Xun, L. Wang, X. Chen, B. Liu and X. Peng, Chem. Commun., 2025, 61, 564–567.
10. H. A. A. Noreldeen, S. B. He, G. W. Wu, H. P. Peng, H. H. Deng and W. Chen, Talanta, 2024, 280, 126679.
11. H. Xi, N. Li, Z. Shi, P. Wu, N. Pan, D. Wang, T. You, X. Zhang, G. Xu, Y. Gao, X. Liang and P. Yin, Anal. Chim. Acta, 2022, 1220, 340023.
12. H. Han, L. Zhu, S. Deng, Y. Wan, K. Ren, Z. Liu, J. Gao, B. Zhu, F. An, J. Luo and H. Qian, Anal. Chem., 2024, 96, 13862–13869.
13. P. Behera, K. K. Singh, S. Pandit, D. Saha, D. K. Saini and M. De, ACS Appl. Nano Mater., 2021, 4, 3843–3851.
14. M. Arabi, A. Ostovan, Z. Zhang, Y. Wang, R. Mei, L. Fu, X. Wang, J. Ma and L. Chen, Biosens. Bioelectron., 2021, 174, 112825.
15. L. Cai, J. Li, H. Lv, W. Liu, H. Niu and Z. Wang, J. Biomed. Inf., 2023, 143, 104418.
16. A. Childs, B. Mayol, J. A. Lasalde-Ramírez, Y. Song, J. R. Sempionatto and W. Gao, ACS Nano, 2024, 18, 24605–24616.
17. M. Pushina, A. Penavic, S. Farshbaf and P. Anzenbacher, ACS Sens., 2021, 6, 4001–4008.
18. F.-F. Li, Y. Liu, W.-L. Jia, R.-N. Ma, W. Zhang, L. Shang, X.-J. Li, H.-S. Wang and L.-P. Jia, Sens. Actuators, B, 2025, 429, 137294.
19. J. Guo, F. Zhao, Z. Yue and Z. Lei, Food Chem., 2025, 465, 142075.
20. J. Yang, W. Ni, B. Ruan, L.-C. Tsai, N. Ma, D. Shi, T. Jiang and F.-C. Tsai, ECS J. Solid State Sci. Technol., 2021, 10, 056003.
21. H. A. A. Noreldeen, K. Y. Huang, G. W. Wu, H. P. Peng, H. H. Deng and W. Chen, Anal. Chem., 2022, 94, 9287–9296.
22. Z. Lu, M. Chen, M. Li, T. Liu, M. Sun, C. Wu, G. Su, J. Yin, M. Wu, P. Zou, L. Lin, X. Wang, Q. Huang, H. Yin, H. Rao, X. Zhou, J. Ye and Y. Wang, Chem. Eng. J., 2022, 439, 135686.
23. Q.-Y. Li, L. Ma, L. Li, S. Wang, X. Li, C. Zhang, Y. Zhang, M. Jiang, H. Wang, K. Huang, X. Yu and L. Xu, Chem. Eng. J., 2022, 430, 132696.
24. L. Pan, F. Yang, S. Xu, D. Lin and C. Jiang, Talanta, 2024, 278, 126356.
25. Q. Zheng, L. Xiong, L. Yu, D. Wu, C. Yang and Y. Xiao, J. Lumin., 2021, 236, 118076.
26. X. Zhou, X. Wang and L. Shang, Chin. Chem. Lett., 2023, 34, 108093.
27. P. Bhatt, D. Kukkar, A. K. Yadav and K. H. Kim, Spectrochim. Acta, Part A, 2024, 307, 123666.
28. R. Jiang, Y. Zhang, Q. Zhang, L. Li and L. Yang, ACS Appl. Nano Mater., 2021, 4, 9760–9767.
29. S. Norouzi, K. Dashtian, F. Amourizi and R. Zare-Dorabei, Analyst, 2023, 148, 3379–3391.
30. L. Huang, T. Huang, J. Liao, M. Liu, Y. Zhang, X. Zhu and S. Yao, Microchem. J., 2024, 207, 111992.
31. M. Zhou, C. Liu, Y. Lin, T. Bai, T. Ye, Z. Qian, L. Li, L. Guo, H. Liu and J. Wang, Food Chem., 2025, 464, 141755.
32. X. Dong, Y. Tian, F. Ai, D. Wei, H. Yin, N. Zhu, Z. Zhang and H. Zhao, Inorg. Chem., 2024, 63, 16264–16273.
33. S. Sarkar, A. Shil, Y. L. Jung, S. Singha and K. H. Ahn, ACS Sens., 2022, 7, 3790–3799.
34. C. Zhang, L. Li, L. Xu, C. Ye, P. Han, M. Wang, R. Liu, S. Chen, X. Wang and Y. Song, J. Phys. Chem. Lett., 2022, 13, 5758–5765.
35. S. Norouzi, A. Esmaeili, K. Dashtian, R. Zare-Dorabei, M. Mahdavi, M. Noroozifar and K. Kerman, Microchim. Acta, 2025, 192, 121.
36. J. Fan, R. Zhu, W. Han, H. Han and L. Ding, Spectrochim. Acta, Part A, 2024, 310, 123893.
37. X. Zhou, S. Huang, W. Liu and L. Shang, Anal. Chem., 2024, 96, 4224–4231.
38. L. Yu, Y. Pang, Z. Mo, Y. Huang and X. Shen, Talanta, 2021, 231, 122357.
39. L. Mitchell, C. Shen, H. C. Timmins, S. B. Park and E. J. New, ACS Sens., 2021, 6, 1261–1269.
40. M. Flaucher, M. Nissen, K. M. Jaeger, A. Titzmann, C. Pontones, H. Huebner, P. A. Fasching, M. W. Beckmann, S. Gradl and B. M. Eskofier, IEEE J. Transl. Eng. Health Med., 2022, 10, 1–9.
41. F. Lu, S. Yang, Y. Ning, F. Wang, X. Ji and Z. He, Analyst, 2021, 146, 5074–5080.
42. J. Xu, X. Tao, X. Liu and L. Yang, Anal. Chem., 2022, 94, 8659–8667.
43. Z. Wang, Y. Dong, X. Sui, X. Shao, K. Li, H. Zhang, Z. Xu and D. Zhang, npj Flexible Electron., 2024, 8, 35.
44. H. Kim, S. Lee, K. W. Lee, E. S. Kim, H.-M. Kim, H. Im, H. C. Yoon, J. Ko and E. Kim, Dyes Pigm., 2023, 215, 111287.
45. J. Wang, Y. Huang, Z. Kuai, Y. Zhang, Q. Shen, P. Tian, W. Nong, W. Jiang, Y. He, N. Ran, Y. Yin, T. Li and Q. Luo, Int. J. Biol. Macromol., 2024, 283, 137778.
46. A. De Masi, P. L. Scognamiglio, E. Battista, P. A. Netti and F. Causa, Lab Chip, 2023, 23, 2458–2468.
47. K. Dashtian, F. Binabaji and R. Zare-Dorabei, Anal. Chem., 2023, 95, 16315–16326.
48. H. Hashemian, M. Ghaedi, K. Dashtian, S. Khan, S. Mosleh, S. Hajati and D. Razmjoue, Sens. Actuators, B, 2024, 406, 135354.
49. A. Ostovan, M. Arabi, Y. Wang, J. Li, B. Li, X. Wang and L. Chen, Adv. Mater., 2022, 34, 2203154.
50. M. Arabi, A. Ostovan, J. Li, X. Wang, Z. Zhang, J. Choo and L. Chen, Adv. Mater., 2021, 33, 2100543.
51. M. Arabi, A. Ostovan, Y. Wang, R. Mei, L. Fu, J. Li, X. Wang and L. Chen, Nat. Commun., 2022, 13, 5757.
52. X. Chen, A. Ostovan, M. Arabi, Y. Wang, L. Chen and J. Li, Anal. Chem., 2024, 96, 6417–6425.
53. H. Lu, Q. Lu, H. Sun, Z. Wang, X. Shi, Y. Ding, X. Ran, J. Pei, Y. Pan and Q. Zhang, ACS Appl. Mater. Interfaces, 2023, 15, 38381–38390.
54. B. Dhanasekaran, M. Chandran, G. Chellasamy, M. Veerapandian, S. Govindaraju and K. Yun, ACS Appl. Bio Mater., 2024, 7, 6065–6077.
55. S. Yin, X. Chen, R. Li, L. Sun, C. Yao and Z. Li, ACS Nano, 2024, 18, 18522–18533.
56. Y. Guo, L. Zhong, L. Min, J. Wang, Y. Wu, K. Chen, K. Wei and C. Rao, Opto-Electron. Adv., 2022, 5, 200082.
57. F. Binabaji, K. Dashtian, R. Zare-Dorabei, N. Naseri, M. Noroozifar and K. Kerman, Anal. Chem., 2024, 96, 13522–13532.
58. H. Im, J. Choi, H. Lee, Z. Y. Al Balushi, D.-H. Park and S. Kim, ACS Sens., 2023, 8, 3370–3379.
59. M. Karimian, K. Dashtian and R. Zare-Dorabei, Talanta, 2024, 266, 125138.
60. J. Wang, Z. Zhou, Y. Luo, T. Xu, L. Xu and X. Zhang, Anal. Chem., 2023, 96, 381–387.
61. K. Sun, B. Wang, J. Lin, L. Han, M. Li, P. Wang, X. Yu and J. Tian, Sensors, 2023, 23, 6114.
62. M. Li, L. Jia, X. Chen, Y. Li, D. Zhao, L. Zhang, T. Zhao and J. Xu, Sens. Actuators, B, 2024, 407, 135491.
63. J.-Q. Du, W.-C. Luo, J.-T. Zhang, Q.-Y. Li, L.-N. Bao, M. Jiang, X. Yu and L. Xu, Sens. Actuators, B, 2024, 417, 136173.
64. H. A. A. Noreldeen, S. B. He, K. Y. Huang, C. T. Zhu, Q. L. Zhou, H. P. Peng, H. H. Deng and W. Chen, Anal. Bioanal. Chem., 2022, 414, 8365–8378.
65. X. Li, P. Xue, H. Wang, H. Li, R. Du, J. Gao, K.-Y. Wong and Y. Qin, J. Mater. Chem. C, 2024, 12, 16523–16532.
66. Y. Zhang, M. Wang, C. Shao, T. Liu, M. Sun, C. Wu, G. Su, Y. Wang, J. Ye, H. Hu, Y. Li, H. Rao and Z. Lu, Anal. Chim. Acta, 2024, 1297, 342373.
67. L. Nie, L. Jiang, S. Li, D. Song, G. Dong, L. Bu, C. Chen and Q. Zhou, Talanta, 2024, 275, 126119.
68. X. Xue, F. Zheng, Y. Luo, W. Chen, Y. Gao and K. Wei, Microchim. Acta, 2025, 192, 228.
69. J. Yu, D. Wang, V. V. Tipparaju, W. Jung and X. Xian, Biosens. Bioelectron., 2022, 195, 113650.
70. Z. Lu, J. Li, K. Ruan, M. Sun, S. Zhang, T. Liu, J. Yin, X. Wang, H. Chen, Y. Wang, P. Zou, Q. Huang, J. Ye and H. Rao, Chem. Eng. J., 2022, 435, 134979.
71. Z. Lu, M. Chen, T. Liu, C. Wu, M. Sun, G. Su, X. Wang, Y. Wang, H. Yin, X. Zhou, J. Ye, Y. Shen and H. Rao, ACS Appl. Mater. Interfaces, 2023, 15, 9800–9812.
72. Y. Wang, W. Liu, Z. Wang, J. Chi, C. Lyu, Q. Du, P. Ju, Y. Ding and Z. Chi, J. Hazard. Mater., 2024, 480, 136292.
73. Y.-D. Deng, Q. Liu, D. Wang, Z.-W. Pan, T.-T. Du, Z.-X. Yuan and W.-J. Yi, Bioorg. Chem., 2024, 152, 107742.
74. J.-N. Jiang, Q. Wan, N. Sun, Y.-L. Zhang, B. Wang, J.-F. Zheng, Y. Shao, Y.-H. Wang and X.-S. Zhou, ACS Sens., 2024, 9, 5578–5586.
75. R. P. Yang, Q. Tang, H. L. Yi, Z. Tao, X. Xiao, A. T. Zhao and Y. Huang, Mikrochim. Acta, 2024, 191, 401.
76. J. A. Stocki, R. C. Fleck, I. B. Nguyen, R. Walde, H. L. T. Mobley and A. E. Shea, Appl. Environ. Microbiol., 2024, 90, e0156724.
77. X. Li, T. Zheng, Y. Xiao, Y. Zhao and P. Wu, Anal. Chem., 2024, 96, 14679–14687.
78. Y. Shu, Y. Wang, J. Guan, Z. Ji, Q. Xu and X. Hu, Anal. Chem., 2022, 94, 5415–5424.
79. Y. Jia and B. Yan, Spectrochim. Acta, Part A, 2023, 297, 122764.
80. X. Wang, H. Zhang, J. Li and G. Zou, Chin. J. Chem., 2023, 42, 491–498.
81. Z. Quan, Z. Chen, H. Li, S. Sun and Y. Xu, Int. J. Biol. Macromol., 2024, 276, 133760.
82. V. Kansay, V. D. Sharma, V. Srivastava, N. Batra, S. Chakrabarti and M. K. Bera, Microchem. J., 2024, 201, 110624.
83. Z. Li, T. Li, X. Wang, J. Ping and H. Peng, Talanta, 2025, 281, 126883.
84. N. Fotovat-Ahmadi, O. Siddiqui, J. Ong, C. Thanitcul, C. Reinhardt, S. M. Cologna and V. K. Aakalu, Ocul. Surf., 2025, 37, 283–300.
85. M. Deng, G. Song, K. Zhong, Z. Wang, X. Xia and Y. Tian, Sens. Actuators, B, 2022, 352, 131067.
86. D. Das, H.-A. Chen, C.-L. Weng, Y.-C. Lee, S.-M. Hsu, J.-S. Kwon and H.-S. Chuang, Anal. Chim. Acta, 2024, 1325, 343117.
87. Y. Shi, Y. Zhang, Y. Hu, R. Moreddu, Z. Fan, N. Jiang and A. K. Yetisen, Sens. Actuators, B, 2023, 378, 133128.
88. Z. Yang, X. Yu, B. Huang, J. Li, X. Yan, K. Feng, R. Cai, Y. Yuan, T. Yue and Q. Sheng, Microchem. J., 2025, 212, 113263.
89. Y. Wang, J. Yao, S. Wu, C. Zhi, L. Yin, Z. Song, J. Wang, L. Ling, Y. Ma, D. Zhang, J. Li, L. Li and B. Chen, Chem, 2025, 11, 102457.
90. G. Song, D. Jiang, J. Wu, X. Sun, M. Deng, L. Wang, C. Hao, J. Shi, H. Liu, Y. Tian and M. Chen, Chem. Eng. J., 2022, 440, 135979.
91. X. Gao, R. Huang, W. Fang, W. Huang, Z. Yin, Y. Liu, X. Huang, L. Ding, H. Peng and Y. Fang, Adv. Mater. Interfaces, 2022, 9, 2201275.
92. M. Kim, C. Chen, P. Wang, J. J. Mulvey, Y. Yang, C. Wun, M. Antman-Passig, H.-B. Luo, S. Cho, K. Long-Roche, L. V. Ramanathan, A. Jagota, M. Zheng, Y. Wang and D. A. Heller, Nat. Biomed. Eng., 2022, 6, 267–275.
93. P. Liu, L. Du, F. Luan, C. Shi, Y. Liu, Z. Gai, F. Yang and Y. Yang, ACS Appl. Mater. Interfaces, 2025, 17, 7553–7567.
94. I. Grabska-Kobyłecka, P. Szpakowski, A. Król, D. Książek-Winiarek, A. Kobyłecki, A. Głąbiński and D. Nowak, Nutrients, 2023, 15, 3454.
95. J. An, Y. Shi, J. Fang, Y. Hu and Y. Liu, Chem. Eng. J., 2021, 425, 130595.
96. J.-H. Yin, M. Liu, C. Lan, B. Chu, L. Meng and N. Xu, Anal. Chim. Acta, 2023, 1279, 341823.
97. J.-X. Li, T.-Q. Chai, G.-Y. Chen, M.-L. Luo, J.-B. Wan and F.-Q. Yang, Anal. Chim. Acta, 2024, 1330, 343298.
98. R. Puglisi, L. M. Mancuso, R. Santonocito, A. Gulino, V. Oliveri, R. Ruffino, G. L. Destri, V. Muccilli, N. Cardullo, N. Tuccitto, A. Pappalardo, G. Sfuncia, G. Nicotra, M. Petroselli, F. Pappalardo, V. Zaccaria and G. T. Sfrazzetto, J. Mater. Chem. B, 2024, 12, 7826–7836.
99. L. Xu, H. Wang, Y. Xu, W. Cui, W. Ni, M. Chen, H. Huang, C. Stewart, L. Li, F. Li and J. Han, ACS Sens., 2022, 7, 1315–1322.
100. X. Yue, F. Xu, L. Zhang, G. Ren, H. Sheng, J. Wang, K. Wang, L. Yu, J. Wang, G. Li, G. Lu and H.-D. Yu, ACS Sens., 2022, 7, 2198–2208.
101. Y. Zhang, X. Cao, J. Zhang, G. Zhang, M. Zhu, H. Yan and Y. Li, Int. J. Biol. Macromol., 2024, 258, 129000.
102. R. S. Mehrbani, M. Mohhamadian, K. Dashtian, N. Naseri, R. Zare-Dorabei, M. Noroozifar, K. Kerman and M. Mahdavi, Anal. Chem., 2025, 97, 11424–11435.
103. Q. Chen, S. Li, X. Tu and X. Zhang, Biosens. Bioelectron., 2024, 263, 116606.
104. Y. Shi, Y. Hu, N. Jiang and A. K. Yetisen, ACS Sens., 2022, 7, 1615–1633.
105. B. Shen, W. Peng, B. Su, L. Wu, Z. Liu, H. Xu, J. Zhao, P. Feng and F. Li, Anal. Chem., 2022, 94, 1910–1917.
106. Y. Liu, Y. Li, S. Yu, K. Niu, H. Liu, L. B. Xing and S. Ma, Macromol. Rapid Commun., 2024, 45, 2400297.
107. X. Qiao, Y. Cai, Z. Kong, Z. Xu and X. Luo, ACS Sens., 2023, 8, 2834–2842.
108. M. Sang, M. Cho, S. Lim, I. S. Min and Y. Han, Sci. Adv., 2023, 9, 22.
109. J. Xiao, S. Zhang, Q. Liu, T. Xu and X. Zhang, Sens. Actuators, B, 2024, 398, 134685.
110. R. S. Tabatabaee, T. Naghdi, M. Peyravian, M. A. Kiani and H. Golmohammadi, ACS Nano, 2024, 18, 28012–28025.
111. K. Nagy, R. McBride, S. R. Head, P. Ordoukhanian and M. Law, in Peptide Microarrays, 2023, ch. 6, pp. 63–81.
112. E. Huang, D. Huang, Y. Wang, D. Cai, Y. Luo, Z. Zhong and D. Liu, Biosens. Bioelectron., 2022, 195, 113684.
113. H. Shin, G.-W. Yoon, W. Choi, D. Lee, H. Choi, D. S. Jo, N. Choi, J.-B. Yoon and I.-J. Cho, npj Flexible Electron., 2023, 7, 7.
114. M. S. Bosco, D. Naud-Martin, C. Gonzalez-Galindo, M. Auvray, M. Araya-Farias, G. Gropplero, Y. Rozenholc, Z. Topcu, J.-F. Gaucher, V. Tsatsaris, S. Descroix, F. Mahuteau-Betzer and N. Gagey-Eilstein, ACS Appl. Bio Mater., 2024, 7, 8236–8247.

---