What if diagnosing cancer or other serious illnesses could be as quick and straightforward as taking a pregnancy test or monitoring blood sugar levels with a glucose meter? This transformative vision is taking shape at the Carl R. Woese Institute for Genomic Biology, where researchers have developed an innovative approach that brings point-of-care biosensing technologies closer to widespread, practical use. By harnessing the power of machine learning integrated directly into diagnostic devices, this new method, known as LOCA-PRAM, promises to eliminate the need for expert analysis and make early disease detection more accessible and efficient.
Conventional medical diagnostics often involve sending blood or tissue samples to centralized clinical laboratories, where specialized personnel perform intricate testing and data interpretation. This process can be time-consuming and costly, creating barriers for many patients, especially those who face logistical, financial, or geographical limitations in accessing healthcare facilities. Recognizing these challenges, the research team, led by graduate student Han Lee and Professor Brian Cunningham at the University of Illinois at Urbana-Champaign, set out to develop a solution that brings diagnostic power directly to the patient’s side.
Point-of-care testing refers to medical testing performed at or near the site of patient care, ranging from home settings to clinics or specialist appointments. By providing rapid, easy-to-use, and cost-effective diagnostic tools, these technologies enable clinicians and patients to make timely decisions that can dramatically improve health outcomes. Examples such as home pregnancy kits, at-home COVID-19 antigen tests, and blood glucose meters for diabetes management have already demonstrated how point-of-care devices can revolutionize healthcare delivery and patient autonomy.
The team’s breakthrough stems from advancing a cutting-edge biosensing technique originally reported in prior studies, called Photonic Resonator Absorption Microscopy—or PRAM. PRAM offers an unprecedented ability to detect individual biomarker molecules such as nucleic acids, antigens, and antibodies, which act as critical indicators of physiological or pathological states. Unlike many biosensors that measure the collective signal generated by thousands of molecules, PRAM achieves digital resolution by identifying single molecules, significantly enhancing detection sensitivity and diagnostic precision.
At its core, PRAM operates by shining red LED light onto a sophisticated photonic sensor where target molecules tagged with gold nanoparticles (AuNPs) bind to the surface. These AuNPs, minuscule particles approximately 1,000 times smaller than human hair, create detectable contrast spots against a red background when imaged. However, the raw images generated can be difficult to interpret because of the presence of artifacts such as dust, nanoparticle aggregates, or noise. Traditionally, accurately counting the true biomarker-related signals demands extensive expertise and manual adjustment of thresholding parameters, limiting the scalability and applicability of PRAM in everyday clinical use.
To overcome these challenges, Han Lee developed a novel integration of advanced machine learning algorithms with PRAM, pioneering a method termed Localization with Context Awareness (LOCA). This approach leverages deep learning techniques to automatically analyze PRAM images, accurately distinguishing genuine biomarker signals from artifacts, and enabling real-time, high-precision molecular detection. The incorporation of artificial intelligence dramatically reduces dependence on human expertise, facilitating point-of-care deployment by non-specialists and patients themselves.
Because machine learning models rely heavily on the quality of their training data, the researchers adopted an innovative validation strategy. Lee painstakingly imaged identical biomarker samples using both PRAM and scanning electron microscopy (SEM). SEM provides ultra-high-resolution images where individual AuNPs are clearly distinguishable, serving as a ground truth reference to annotate spots in the PRAM images precisely. This labor-intensive cross-validation process was akin to finding a needle in a haystack, requiring the creation of reference landmarks to reliably match image areas across the two different microscopy platforms.
The resulting dataset empowered the training of a physically grounded deep learning model capable of interpreting complex microscopic image features in PRAM while factoring in physical realities of nanoparticle behavior and sensor optics. When tested, LOCA-PRAM demonstrated remarkable improvements over conventional image analysis algorithms, exhibiting enhanced sensitivity in detecting lower biomarker concentrations and substantially reducing false-positive and false-negative rates. This leap in analytical performance opens the door to reliable and widespread clinical application of PRAM technology.
Professor Brian Cunningham emphasizes the clinical potential of rapid, point-of-care diagnostics powered by this technology. Physicians often encounter bacterial infections treated empirically with broad-spectrum antibiotics due to lack of rapid identification of the causative agent. LOCA-PRAM’s capability suggests a future where cancer patients could receive tailored therapeutic guidance during routine appointments, quickly determining the most effective anti-cancer drugs or monitoring treatment efficacy shortly after initiation. Such timely interventions could dramatically improve patient outcomes and reduce unnecessary side effects.
This project exemplifies how interdisciplinary collaboration—combining electrical and computer engineering, materials science, and biomedical research—can yield technologies that bridge fundamental science and clinical practice. The implementation of machine learning in biosensing not only exemplifies technical ingenuity but also reflects a commitment to addressing real-world healthcare disparities by enhancing diagnostic accessibility.
Han Lee’s journey highlights the transformative power of curiosity and cross-field learning. Inspired by a university course in machine learning, Lee independently explored how artificial intelligence could solve persistent image interpretation problems in biosensing. The result is not merely an academic advance but a potentially life-saving technology that contributes meaningfully to the evolution of personalized medicine and global health.
Published in the journal Biosensors and Bioelectronics, the study titled “Physically grounded deep learning-enabled gold nanoparticle localization and quantification in photonic resonator absorption microscopy for digital resolution molecular diagnostics” represents a significant milestone in biosensor development. Supported by prominent funding agencies including the National Institutes of Health, the USDA AFRI Nanotechnology grant, and the National Science Foundation, this research lays foundational work for next-generation diagnostic devices.
As the medical field moves towards more decentralized, patient-centered care, technologies like LOCA-PRAM could redefine how we detect, monitor, and manage diseases in real-time. This innovative blend of nanotechnology, photonics, and artificial intelligence heralds a new era of precision diagnostics—one where critical health information can be accessed rapidly and affordably, empowering both patients and practitioners alike. The implications for public health, especially in underserved communities, are profound and far-reaching.
Subject of Research: Biosensing technology, machine learning integration, and point-of-care molecular diagnostics
Article Title: Physically grounded deep learning-enabled gold nanoparticle localization and quantification in photonic resonator absorption microscopy for digital resolution molecular diagnostics
Web References: https://doi.org/10.1016/j.bios.2025.117455
References: Supported by National Institutes of Health, USDA AFRI Nanotechnology grant, and National Science Foundation
Image Credits: Julia Pollack
Keywords: Machine learning, Photonic crystals, Gold nanoparticles, Biomarkers, Medical diagnosis
