Smartphone-Assisted CRISPR Diagnostics for Rapid Pathogen Detection in Low-Resource Settings

Thesis Abstract

Rapid, accurate, and affordable pathogen detection remains a critical bottleneck for infectious disease control in low-resource settings, where laboratory infrastructure, trained personnel, and supply chains are often limited. This study investigates a smartphone-assisted CRISPR diagnostics platform that integrates portable microfluidics, lateral-flow readouts, and a mobile app-driven data pipeline to enable field-deployable, rapid pathogen detection. The central problem addressed is the gap between highly accurate CRISPR-based assays and their practical deployment in communities with constrained resources, which hampers timely clinical decision-making and outbreak response. The aim is to evaluate whether a fully integrated smartphone-enabled CRISPR diagnostic workflow can achieve comparable sensitivity and specificity to laboratory-based methods while reducing turnaround times and operational costs. Specific objectives are (i) to develop and optimize a CRISPR-Cas12/13-based assay targeting common household- and healthcare-associated pathogens (including SARS-CoV-2, Dengue virus, and Escherichia coli O157H7) using a single-field sample-processing cartridge; (ii) to implement a smartphone-imaged, real-time fluorescence readout and machine-learning–assisted result interpretation to determine presence/absence with predefined thresholds; (iii) to assess diagnostic performance (sensitivity, specificity, PPV, NPV) in a field trial across three diverse settings—rural, peri-urban, and urban clinics—enrolling 600 participants with suspected infectious disease symptoms; (iv) to analyze user acceptance, operational feasibility, and data timeliness through a mixed-methods approach; and (v) to model cost-effectiveness relative to conventional RT-PCR testing to inform scalability. The methodology adopts a convergent parallel mixed-methods design. A cross-sectional diagnostic accuracy study will be conducted with 600 participants recruited from three health districts, with parallel reference testing by gold-standard RT-PCR conducted in centralized laboratories. The diagnostic workflow comprises a lyophilized CRISPR-Cas enzymatic reaction integrated into a disposable microfluidic cartridge, a 12-minute isothermal amplification step, and a smartphone-based imaging module leveraging an open-source computer vision pipeline and a random forest classifier to translate fluorescence signals into binary diagnostic outcomes. Data collection instruments include a structured clinical form, device performance logs, the smartphone application’s metadata capture (time-to-result, ambient temperature, battery level), and semi-structured interviews with 40 healthcare workers and 30 community participants to gauge usability and acceptability. Validity and reliability are ensured through calibration against known positive and negative controls, inter-device reproducibility tests, and pilot-testing of the app’s readout algorithm with kappa statistics for agreement. Analytical approaches combine quantitative and qualitative techniques. Diagnostic performance will be evaluated using receiver operating characteristic (ROC) analysis, with area under the curve (AUC) estimates and 95% confidence intervals, along with McNemar’s test for paired proportions to compare against RT-PCR. Regression analyses will explore associations between diagnostic performance and variables such as sample type, ambient conditions, and user expertise. The cost-effectiveness model will employ a decision tree framework comparing per-test costs, turnaround times, and downstream clinical outcomes. The qualitative data will be subjected to thematic analysis, employing a priori codes informed by technology acceptance and workflow integration theories, and cross-validated with analyst triangulation. A theoretical lens drawing on Technology Acceptance Model (TAM) and Diffusion of Innovations (DOI) will frame determinants of adoption. Expected findings indicate that the smartphone-assisted CRISPR platform can achieve sensitivity and specificity in the range of 85–92% and 95–98%, respectively, with mean time-to-result under 25 minutes and per-test costs substantially lower than centralized RT-PCR when deployed at point-of-care. It is anticipated that the app’s ML component will maintain diagnostic accuracy across environmental conditions and user profiles, while qualitative insights will identify critical facilitators (rapid feedback, ease of use, low supply-chain burden) and barriers (data privacy concerns, device maintenance). The study contributes to knowledge by providing robust evidence for the feasibility, accuracy, and cost-effectiveness of mobile CRISPR diagnostics in real-world, resource-limited settings, and by offering a scalable blueprint for integrating ICT-enabled molecular testing with community health workflows. The main conclusion is that smartphone-assisted CRISPR diagnostics can bridge diagnostic gaps in low-resource contexts, enabling timely clinical decisions and improved surveillance, provided that rigorous quality assurance, user-centered design, and data governance are maintained. Recommendations include standardized training for frontline workers, deployment guidelines tailored to environmental conditions, integration with existing health information systems, iterative updates to the device firmware and ML models, and policy frameworks to address data privacy, regulatory approval, and sustainability.