Smartphone-integrated lateral flow assay for robust detection of Δ9-Tetrahydrocannabinol

Abstract

The increasing prevalence of cannabis use necessitates accurate and accessible monitoring methods to mitigate substance-associated risks and enhance public safety. We present iLine (integrated lateral-flow assay with intensity normalization and enhancement) for sensitive and on-site quantification of tetrahydrocannabinol (THC), the primary psychoactive component of cannabis. The iLine platform incorporates i) gold nanoclusters for enhanced optical signals and ii) a smartphone-based imaging system with intelligent algorithms to ensure consistent signal acquisition. These features enable a portable, hardware-minimal THC detection while maintaining laboratory-grade performance in real-world conditions. Specifically, iLine achieved a detection limit of 0.12 ng/mL for THC, representing an 85-fold improvement over conventional lateral-flow assays. The analytical performance was highly robust, with coefficient of variation values < 2 % under diverse imaging conditions, including daylight, direct lamp illumination, and shadow environments. When applied to human saliva samples, iLine successfully differentiated cannabis users (n = 10) from non-users (n = 10) within 10 min. The iLine approach facilitates the use of smartphones as ubiquitous detectors, broadening access to THC testing in field applications.

1. Introduction

Cannabis is the most widely consumed psychoactive substance globally; the World Health Organization estimates that over 147 million people (>2.5 % of the world’s population) use cannabis products (Ar-Sanork et al., 2025; WHO, 2025). The ongoing trend of cannabis legalization has led to a substantial increase in both recreational and medicinal cannabis consumption, creating a need for reliable monitoring technologies (Lee et al., 2016; Manthey et al., 2021; Yu et al., 2021; Churcher et al., 2023). The primary psychoactive component, Δ9-tetrahydrocannabinol (THC), is associated with significant cognitive impairment, including deficits in attention, coordination, and memory processing (Murray et al., 2007; Hall and Degenhardt, 2009; Stuyt, 2018; Couttas et al., 2023). These effects pose substantial risks in safety-critical environments, particularly in transportation and industrial settings (Moore et al., 2007; Hartman and Huestis, 2013). Indeed, epidemiological studies have demonstrated increased vehicular accident rates in regions following cannabis legalization, particularly among younger populations (Fergusson et al., 2008; Veldstra et al., 2015; Derne et al., 2024). Similarly, workplace safety data indicate that cannabis-positive employees experience 55 % more industrial accidents and 85 % more injuries compared to their cannabis-negative counterparts (O’Neill et al., 2023; Macdonald and Zhao, 2024). These statistics underscore the critical need for reliable THC detection methods, especially given the expanding global drug testing market that is expected to reach USD 73.6 billion by 2027 (Malabadi et al., 2024).

Current THC detection methods face practical challenges in field deployment. Traditional laboratory-based techniques, while highly accurate, require complex sample preparation, specialized equipment, and trained personnel, making them unsuitable for point-of-care (POC) applications (Breidi et al., 2012; Couttas et al., 2023). Conversely, existing rapid testing kits often suffer from poor sensitivity, a limited dynamic range, or require subjective visual interpretation, which can lead to inconsistent results (Cirimele et al., 2006; Bosker and Huestis, 2009). Lateral flow assay (LFA) technology is considered a promising alternative to enable POC detection of THC (Hudson et al., 2019; Liang et al., 2022; Deenin et al., 2023). Recent advances, such as thermographic-based systems (Thapa et al., 2020), achieved superior sensing performance, demonstrating 96 % accuracy. However, quantitative LFA approaches often require complex optical setups (Rong et al., 2019; Yu et al., 2021; Lin et al., 2023), controlled environmental conditions, or specialized instrumentation, limiting their practical utility in field settings where ambient conditions vary.

Here, we present an integrated LFA approach designed for highly sensitive and reliable THC detection, termed iLine (integrated LFA with intensity normalization and enhancement). The technique combined two key features: i) enhanced optical sensitivity enabled by gold nanoclusters (AuNCs) as detection labels, and ii) a smartphone-based imaging system that incorporated an intelligent algorithm to acquire consistent optical signals. Specifically, iLine utilized a reference card to compensate for variations in illumination, camera settings, and imaging orientation (Park et al., 2022). These capabilities enabled portable, hardware-minimal THC detection while maintaining laboratory-grade performance in real-world conditions. iLine demonstrated robust analytical performance, yielding coefficient of variation (CV) values < 2 % under diverse ambient conditions (e.g., daylight, direct lamp illumination, shadow). It achieved a detection limit of 0.12 ng/mL for THC, which is about 85-fold lower than that of conventional LFA methods. When applied to human saliva samples, iLine successfully stratified cannabis users (n = 10) and non-users (n = 10) within 10 min. The iLine approach could facilitate the use of smartphones as ubiquitous detectors, thereby broadening access to THC testing in field settings.

2. Materials and methods

2.1. Materials and reagents

Gold(III) chloride trihydrate (HAuCl₄, >99.99 %), polyvinylpyrrolidone (PVP, MW 10 kDa), bovine serum albumin (BSA, 99 %), Tween-20 (MW 1.228 kDa), anti-mouse IgG antibody (1 mg/mL), para-aminobenzenethiol (ABT, ≥97 %), acetaminophen (AAP), acetylsalicylic acid (ASA), and ibuprofen (IBF, ≥98 %) were obtained from Sigma-Aldrich (St. Louis, MO, USA). Borate buffer (20 ×, pH 8.5) and phosphate-buffered saline (1 × PBS, pH 7.4) were purchased from ThermoFisher Scientific (Waltham, MA, USA). Surfactant 10G (S-10G, 30–40 dyn/cm surface tension), anti-THC antibody (10-T43E), and THC-BSA conjugate (80-IT62) were acquired from Fitzgerald Industries (North Acton, MA, USA). Analytical standards, including Δ9-tetrahydrocannabinol (THC, 1 mg/mL), oxycodone (OCD, 1 mg/mL), and methadone (MTD, 1 mg/mL), were sourced from Cerilliant (Round Rock, TX, USA). Nitrocellulose membranes (Whatman Immunopore RP, 90–150 s/4 cm capillary flow rate) were obtained from Cytiva (Marlborough, MA, USA). All aqueous solutions were prepared using ultrapure water (18.2 MΩ cm) generated by a PURELAB Option-Q system (ELGA LabWater, UK).

2.2. Synthesis and characterization of gold nanoclusters

Gold nanoclusters (AuNCs) were synthesized via a rapid, room-temperature reduction method. In brief, the ABT solution (25 mM in 95 % ethanol) was combined with a HAuCl₄ solution (10 mM) containing 0.001 % PVP in a 1:9 vol ratio, resulting in the immediate formation of purple-brown AuNCs. The product was purified by triple centrifugation (6000 rpm, 30 min) with intermediate washing steps. Characterization was performed using UV–Vis spectroscopy and transmission electron microscopy.

2.3. Preparation of Antibody-AuNC conjugates

Anti-THC antibodies (10 μL, 1 mg/mL) were conjugated to purified AuNCs in 20 mM borate buffer (pH 8.5) through a 90-min incubation at room temperature. The conjugates were stabilized by the addition of BSA (50 μL, 5 % w/v) and a 45-min incubation. The resulting antibody-AuNC conjugates (Ab@AuNC) were purified by triple centrifugation (6000 rpm, 30 min).

2.4. Lateral flow strip fabrication

Nitrocellulose membranes were laminated onto adhesive-backed cards, followed by the attachment of absorption pads with a 2-mm overlap. The assembled materials were cut into 4-mm wide strips and stored in a desiccator at room temperature until use. Test and control lines were prepared by dispensing THC-BSA conjugate (0.5 mg/mL, 1.65 μL) and anti-mouse IgG antibody (1 mg/mL, 1.65 μL), respectively, followed by 2-h desiccation at room temperature.

2.5. LFA-based THC detection protocol

For THC quantification, test strips were prepared by dispensing THC-BSA conjugate (0.5 mg/mL, 1.65 μL) and anti-mouse IgG antibody (1 mg/mL, 1.65 μL) onto the test and control zones of the nitrocellulose membrane, respectively. Working buffer (0.025 % S-10G, 0.5 % BSA in 1 × PBS) was used as the assay medium. THC-containing samples were combined with the Ab@AuNC conjugate at a 9:1 volumetric ratio. The prepared LFA strips were immersed in this mixture solution, and the immunochromatographic reaction was allowed to proceed for 15 min at room temperature.

2.6. Smartphone-based signal acquisition

Following the immunoassay reaction, LFA strips were mounted onto the reference card for signal analysis. The reference card incorporates multiple calibration elements, including colorimetric standards (red, green, blue, white, and black), a quick-reference (QR) code for system registration, and a designated area for positioning the test strips. Image acquisition was performed using the custom-developed smartphone application (tested on Samsung Galaxy Note 20 5G and Apple iPhone 7 devices). The signal acquisition process follows a defined sequence: i) initial recognition and decoding of the QR code; ii) identification and analysis of colorimetric calibration elements; iii) test strip alignment and signal measurement; iv) automated compensation for ambient lighting conditions; v) signal normalization and quantification.

2.7. THC quantification

Calibration experiments were performed by measuring signal intensities at varying THC concentrations. The resulting data were fitted using a four-parameter logistic model, and the fitted curve was used to estimate THC concentrations from measured signal intensities through interpolation. The limit of detection (LOD) was defined as the analyte concentration corresponding to a signal equal to the mean of blank samples plus three times their standard deviation (mean + 3σ). The limit of quantification (LOQ) was similarly defined as the analyte concentration corresponding to the mean of blank samples plus 10 times their standard deviation (mean + 10σ), following standard analytical convention.

2.8. Sample collection

Oral fluid specimens were collected from adult, anonymous volunteers (age >21) following informed consent under protocols approved by the Massachusetts General Hospital Institutional Review Board (IRB #2019P003472). Information on the administered THC dose and sampling schedule post-ingestion was not available. No multiple samples were collected from the same donor, and baseline (pre-ingestion) samples were not obtained. Participants rinsed their mouths with water before placing an oral swab (SalivaBio Oral Swab, Salimetrics) under the tongue for 1 min. Collected samples were stored at 4 °C until use.

2.9. Sample processing

Samples were processed by centrifugation (10,000 rpm, 5 min) and diluted 9:1 with concentrated working buffer (0.25 % S-10G, 5 % BSA in 10 × PBS). For THC detection, 40 μL of the prepared sample mixture (containing diluted oral fluid, working buffer, and Ab@AuNC) was applied to each LFA strip. Following the 15-min reaction period, strips were analyzed using the smartphone-based detection system under various ambient lighting conditions. Signal intensities at both the test and control lines were quantified using an automated image processing algorithm.

3. Results

3.1. iLine development and optimization

We engineered iLine to achieve robust THC quantification under variable field conditions while maintaining the hardware simplicity. The system consisted of three main components: a precision-engineered reference card incorporating multiple calibration elements, a smartphone app for real-time image analysis, and an LFA strip with enhanced optical signal (Fig. 1A).

Fig. 1. Schematic overview of the smartphone-integrated THC detection platform.

(A) Workflow demonstrating analysis of human oral fluid samples using iLine. (B) Automatic compensation for ambient lighting variations, enabling consistent signal acquisition under both direct illumination and shadow conditions. Data are displayed as mean ± s.d. from triplicate measurements. a.u., arbitrary unit. (C) Representative smartphone interface images showing the sequential detection process. Software-generated visual guides facilitate accurate alignment and analysis of reference card components.

A critical innovation was the automated signal compensation that addressed the fundamental challenge of environmental variability in smartphone-based optical measurements. Unlike previous approaches that relied on hardware accessories or controlled environments, iLine achieved measurement consistency through intelligent software processing (Fig. 1B). The reference card (4 × 5 cm²) incorporated defined calibration features, including colorimetric standards (red, green, blue, white, and black), a QR code for device registration and calibration, and a designated area for LFA strip mounting (Fig. 1C).

Users were required to simply place a reacted LFA strip onto the reference card and open a smartphone app for image acquisition. The app automatically recognized calibration elements and took a photo of the LFA. The acquired image was then processed for alignment, color correction, and signal extraction. Through real-time normalization against printed color standards, this approach effectively eliminates variations introduced by ambient lighting conditions, camera specifications, and environmental factors.

3.2. Automated compensation of variable lighting conditions

The iLine utilized a software-driven approach to acquire robust optical signals. The algorithm sequentially identified printed elements in the reference card (Fig. 2A). It first located the center of the QR code (Ⓐ) and its three corner squares (Ⓑ), establishing geometric references for subsequent measurements. Recognition zones were then extended at 45° angles from the center point to define colorimetric calibration regions (Ⓒ, Ⓓ), followed by the establishment of the LFA signal measurement zone based on these geometric references (Ⓔ, Ⓕ).

Fig. 2. Automated ambient light compensation algorithm and validation.

(A) Software workflow demonstrating systematic recognition of reference card components. Red dots indicate scanning regions for signal acquisition. (B) Algorithm flowchart for signal normalization and analysis. (C) Raw images of LFA strips under various lighting conditions: daylight (4000K), soft white (3000K), warm white (2700K), and corresponding shadow conditions. (D) Quantitative analysis of test line intensities demonstrating iLine’s reproducibility across different lighting conditions. CV, coefficient of variation. Data are displayed as mean ± s.d. from triplicate measurements. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

The iLine signal processing pipeline consisted of multiple sequential steps (Fig. 2B). After the application was launched, a reference card was recognized to standardize imaging conditions (see Supporting Note for details) (Park et al., 2022). The captured image then underwent multi-stage error correction to mitigate variations from orientation, magnification, lighting, and imaging defects. Corrected signals were subsequently rescaled through device-specific algorithms that accounted for differences in imaging systems (e.g., iOS, Android). The converted signals were used for concentration calculation, and the results were displayed to the user in real time. Finally, all processed data were archived on a secure server, enabling subsequent analysis and cross-platform reliability assessment.

We validated the robustness of the iLine algorithm across diverse lighting environments. LFA strip images were acquired under daylight (4000K), soft white (3000K), warm light (2700K), and their corresponding shadow conditions (Fig. 2C). Processing these images with iLine demonstrated high measurement stability, with CV values between 1.09 % and 1.52 % (Fig. 2D). This consistency contrasts favorably with existing smartphone-based systems, which typically exhibit CV values exceeding 10 % under variable lighting conditions (Schneider et al., 2018; Riezk et al., 2025). This high stability effectively overcomes a key limitation for the field deployment of smartphone-based analytical systems.

3.3. Optimization of AuNC-based signal enhancement

We used gold nanoclusters (AuNCs) as detection labels, which produce stronger optical signals than conventional spherical gold nanoparticles (AuNPs) (Hu et al., 2013). The signal enhancement arises from the AuNCs’ multifaceted morphology, which generates abundant plasmonic hot spots and increases the light-scattering cross-sections (Liebig et al., 2020).

The colloidal AuNC preparation exhibited a characteristic purple-brown coloration (Fig. 3A, inset), reflecting a unique clustered nanostructure, as confirmed by electron microscopy analysis (Supplementary Figure S1). Furthermore, the developed synthesis protocol exhibited high batch-to-batch reproducibility, with the CV value of 0.94 % in absorbance peak measurements (Supplementary Figure S2).

Fig. 3. Characterization and optimization of AuNC-based signal enhancement.

(A) UV–Vis absorption spectra of AuNCs before and after antibody conjugation, with an inset showing representative TEM images. (B) Comparative analysis of LFA using spherical Au nanoparticles (AuNPs) and AuNCs conjugated with THC antibodies. THC concentrations were varied from 0 to 10⁴ ng/mL. Dotted arrows denote the THC concentration range where analytical signals were detectable on LFA strips. C, control-line signal; T, test-line signal. (C) Calibration curves comparing analytical performance between AuNC-iLine and AuNP-iLine systems. Using AuNC labels resulted in an approximately 85-fold reduction in LOD, limit of detection. Data are displayed as mean ± s.d. from technical duplicates and fitted to a four-parameter logistic regression model (R² = 0.98 for both AuNC and AuNP).

Conjugation with THC antibodies resulted in a characteristic bathochromic shift from 530 nm to 539 nm (Fig. 3A), indicating successful surface modification while preserving the clustered nanostructure. We subsequently evaluated a panel of ten anti-THC antibodies for their reactivity with three different THC-BSA complexes (Supplementary Fig. S3, Table S1); an antibody exhibiting strong test line signals and minimal non-specific binding was selected for further use.

Compared to conventional spherical gold nanoparticles (AuNPs, 60 nm diameter), AuNCs produced stronger optical signals, resulting in improved sensitivity for THC detection (Fig. 3B). The assay followed a competitive LFA format: increasing THC concentrations inhibited probe binding at the test line, leading to weaker test signals (T) and correspondingly stronger control signals (C). We applied a ratiometric readout, C/(C + T), in which higher THC levels produced higher ratios, enabling an intuitive and quantitative measure of analyte concentration.

Titration experiments revealed that the AuNC-iLine assay achieved a detection limit of 0.12 ng/mL for THC, which was approximately 85-fold lower than that of the AuNP-iLine assay (Fig. 3C). The limit of quantification (LOQ) showed a similar enhancement (66-fold), with LOQ values of 2.3 ng/mL for AuNC-iLine and 153 ng/mL for AuNP-iLine. The dynamic range spanned from 0 to 10⁴ ng/mL, encompassing physiologically relevant THC concentrations (Swortwood et al., 2017).

3.4. Robust quantification of THC

iLine’s signal compensation, combined with AuNC’s signal enhancement, enabled robust THC quantification regardless of the measurement environment. After performing LFAs with samples of varying THC concentrations, we imaged LFA strips under diverse illumination settings (Fig. 4A). Although variations in ambient lighting produced different background signals, iLine extracted corrected signal intensities (Fig. 4B), which yielded consistent dose-dependent signal patterns (Fig. 4C). For instance, the half-maximal effective concentration (EC₅₀) values were statistically identical (p = 0.99; extra sum-of-squares F-test) among the generated calibration curves. These curves were averaged to generate a master calibration curve, which was used to estimate THC concentrations from measured signal intensities through interpolation.

Fig. 4. Robust THC quantification.

(A) LFA strips used in THC quantification were imaged under diverse illumination conditions. (B) iLine-corrected signals at control (C) and test (T) lines. Signal intensities were similar at the given THC concentration regardless of illumination conditions. (C) THC calibration curves remained consistent across different illumination conditions. The half-maximal effective concentration (EC₅₀) values were statistically identical (p = 0.99; extra sum-of-squares F-test). Data are displayed as mean ± s.d. from duplicate measurements and fitted to a four-parameter logistic regression model (R² ≥ 0.99 for all different conditions). (D) iLine was tested on different phone models: Apple iPhone 7 (phone 1) and Samsung Galaxy 20 5G (phone 2). Six samples with varying THC concentrations (0, 1, 10, 10², 10³, and 10⁴ ng/mL) were assessed. The results from two phones showed an excellent match. The dotted line indicates the line of identity (slope = 1). All data are displayed as mean ± s.d. from duplicate measurements.

Further cross-platform validation was performed using smartphones from distinct manufacturers (Apple iPhone 7 and Samsung Galaxy 20 5G). Leveraging the reference card for automatic colorimetric calibration, iLine corrected device-specific differences in the red, green, and blue channels, achieving high concordance between platforms (Fig. 4D). These results demonstrate that iLine maintains consistent analytical performance across different hardware configurations and operating systems without requiring device-specific calibration.

We also assessed intra- and inter-day variability across multiple THC concentrations (1, 10, and 100 ng/mL). The intra-day coefficients of variation (CVs) ranged from 3.4 to 6.9 %, and the inter-day CVs obtained over three independent assay days ranged from 2.1 to 5.9 % (Supplementary Figure S4), confirming the precision and reproducibility of the iLine assay.

3.5. Assay validation and real sample analysis

We performed comprehensive assay validation in preparation for the iLine application with human specimens. The extraction buffer was formulated to include a nonionic surfactant having hydrophobic and hydrophilic groups (0.25 % S-10G) and BSA (5 %), which effectively solubilized hydrophobic THC while preventing the aggregation of AuNCs. Specificity analysis was performed to assess potential cross-reactivity with confounding substances, including common pharmaceuticals (oxycodone, methadone, acetaminophen, ibuprofen, acetylsalicylic acid) and endogenous salivary proteins (human salivary albumin, mucin, amylase). The iLine-THC tests demonstrated high specificity (Fig. 5A). The analytical signal was positive exclusively in the presence of THC, with signal levels remaining consistent irrespective of the presence of these potential interferents. Conversely, control samples without THC showed low values; from these results, we set the signal threshold (0.07) for THC negativity determination.

Fig. 5. iLine validation and analytical performance assessment.

(A) Specificity analysis demonstrating minimal cross-reactivity with potentially interfering substances. Control samples were without THC, and test samples were spiked with THC (10⁴ ng/mL). OCD, oxycodone; MTD, methadone; AAP, acetaminophen; IBF, ibuprofen; ASA, acetylsalicylic acid; HSA, human salivary albumin; AML, amylase. Data are displayed as mean ± s.d. from duplicate measurements. (B) iLine measurements demonstrated a strong correlation (Pearson’s r = 0.98) with ELISA results. (C) Samples prepared in buffer or oral fluid matrices were analyzed using iLine. Both conditions demonstrated similar response characteristics. Data are displayed as mean ± s.d. from duplicate measurements and fitted to a four-parameter logistic regression model (R² = 0.98 for the buffer and 0.99 for the saliva samples). (D) Human saliva samples were analyzed using iLine for the quantification of THC. Cannabis users had either consumed edibles (n = 6) or smoked (n = 4). The dotted line represents the signal threshold (0.07) used to determine THC positivity. The bar represents the signal from a single measurement. (E) The iLine saliva assay successfully distinguished cannabis users from non-users. THC concentrations from edible cannabis users were significantly higher (p = 0.034; two-sided unpaired t-test) than those from smokers. Samples were considered THC negative below the dotted line, which was calculated from the signal threshold.

We further compared iLine and standard ELISA results for THC quantification. The iLine measurements exhibited a strong correlation with ELISA (Pearson’s coefficient, 0.98; Fig. 5B), validating iLine’s analytical accuracy. The Bland–Altman analysis further confirmed a good agreement between iLine and ELISA (Supplementary Figure S5), with a mean bias of 0.024 (95 % limits of agreement ranging from − 0.32 to 0.36).

We next evaluated iLine’s performance in complex biological matrices using oral fluid samples. Notably, higher oral fluid concentrations improved analytical performance (Supplementary Figure S6), likely due to enhanced antibody stability in the presence of endogenous proteins. Comparison of THC detection in buffer versus spiked oral fluid samples showed similar response characteristics (Fig. 5C). Furthermore, THC recovery in saliva ranged from 98 % to 112 % at low and mid analytical concentrations (1, 10, 50, and 100 ng/mL; Supplementary Figure S7), confirming iLine’s analytical compatibility with human samples.

In a pilot testing of the iLine platform, we used oral fluid specimens from cannabis users (n = 10; 6 THC-infused food consumers, 4 smokers) and non-users (n = 10). Samples from cannabis users exhibited significantly elevated THC concentrations (mean: 5.61 μg/mL), while concentrations in samples from non-users remained below the detection threshold (Fig. 5D). THC concentrations in samples from edible cannabis users were higher than those from smokers (Fig. 5E), consistent with previous findings (Yu et al., 2021). Overall, these results confirmed the platform’s capability for point-of-use detection of THC in real-world scenarios.

4. Discussion

We developed and validated iLine, a smartphone-integrated platform for THC detection. The platform achieved exceptional sensitivity (0.12 ng/mL detection limit) by using AuNCs as detection labels and enabled reliable THC quantification across a physiologically relevant range (0–10,000 ng/mL). Importantly, iLine incorporated automatic compensation for ambient light and imaging conditions, which contributed to robust assay performance and high reproducibility (CV < 2 %). These features distinguish iLine from other smartphone-based or electrochemical THC detection systems, which typically rely on specialized optical setups or electrode interfaces (Yu et al., 2021; Churcher et al., 2023; Lin et al., 2023). The iLine prototype demonstrated its practical utility by accurately quantifying THC in human saliva samples. Its hardware-minimal design and automated operation make it suitable for non-expert users in field settings, thereby democratizing access to reliable THC testing.

Smartphones are increasingly utilized as a portable and accessible platform for LFA readouts. However, achieving reproducible quantification remains challenging due to inherent variations in illumination, optical alignment, and device-specific imaging algorithms (Park et al., 2022; Sena-Torralba et al., 2022). Prior smartphone-based systems have addressed these issues by using optical attachments, gamma-correction, or machine-learning-based color normalization (Cheng et al., 2017; Rong et al., 2019; Yu et al., 2021; Cui et al., 2024; Hagos et al., 2025), yet these methods are often device-specific and require specialized hardware, access to proprietary parameters, or extensive training datasets. In contrast, the iLine platform performs color-space correction in a deterministic, reference-guided manner. This is achieved by utilizing predefined color standards on a reference card to calibrate the red, green, and blue channels, as well as the overall image brightness. This direct and universally applicable methodology standardizes imaging across diverse devices and varying illumination conditions, enabling most smartphones to function as quantitative analytical signal readers.

Our study used oral fluid (saliva) as a practical and informative matrix for assessing recent cannabis exposure. It can be collected rapidly and under direct observation, minimizing the risk of sample tampering. Oral-fluid THC concentrations reflect short-term use, typically within the preceding several hours (Milman et al., 2012; Desrosiers and Huestis, 2019; Yu et al., 2021), making this matrix well-suited for evaluating recent cannabis use in safety and compliance settings. The Substance Abuse and Mental Health Services Administration (SAMHSA) recommends a 3 ng/mL cutoff for THC screening in oral fluid (Macdonald and Zhao, 2024), providing a benchmark for interpreting iLine test results. As an antibody-based immunoassay, iLine likely exhibits cross-reactivity with other structurally related cannabinoids, such as cannabidiol (CBD), 11-hydroxy-THC (11-OH-THC), and 11-nor-9-carboxy-THC (THC-COOH) (Schwope et al., 2010). However, prior oral-fluid studies showed that these cannabinoids occur at lower concentrations than parent THC in recent users, minimizing the likelihood of false-positive outcomes (Desrosiers et al., 2012).

iLine outperforms existing THC detection methods by combining simplicity, accuracy, and portability. It eliminates the need for specialized hardware while maintaining laboratory-grade analytical performance, marking a substantial advancement in point-of-care diagnostics. The hardware-minimal design lowers cost, enhances portability, and removes technical barriers that have limited the adoption of previous smartphone-based systems. Additionally, the automated environmental compensation algorithm ensures consistent measurements under diverse lighting conditions, obviating the need for controlled environments and enabling field-deployable optical testing. This capability is particularly valuable for law enforcement, workplace monitoring, and public health screening, where reliable on-site analysis is essential.

We envision future development to extend the current approach. The present pilot study (n = 20) demonstrated feasibility and reproducibility but may not fully capture population variability; larger and more diverse cohorts will be needed to refine diagnostic thresholds and confirm generalizability. While this work focused on THC as the primary analytical target, the detection panel could be expanded to incorporate additional cannabinoids and metabolites (Desrosiers et al., 2014). Such multi-analyte assays may differentiate recent cannabis use from residual compounds from prior consumption, thereby improving the temporal resolution of THC tests. Furthermore, expanding beyond the current oral fluid validation to other matrices, such as urine or sweat, would broaden the applicability of THC testing (Huestis and Smith, 2018). Additionally, developing simplified sample processing kits would enhance the usability of the iLine platform and facilitate the implementation of larger-scale studies to establish performance characteristics across diverse populations and use patterns.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.bios.2025.118349.

Data availability

Data will be made available on request.