I-Corps: Translation potential of optical phenotyping using label-free microscopy and deep learning

Award ID: 2627608

This I-Corps project is based on the development of a platform for identifying tumor types using optical imaging and artificial intelligence. Current precision medicine approaches depend on analyzing a patient's tissue or blood to identify specific genes, proteins, or other molecules unique to their disease, which are often expensive, time-intensive, and available only through specialized laboratories. This limits access for millions of patients worldwide and delays treatment decisions for many cancers. This technology addresses this challenge by combining advanced optical microscopy with machine learning to rapidly characterize the biological properties of tumors without the need for molecular labels or complex laboratory workflows. The technology is designed to integrate into existing pathology workflows or operate as a centralized diagnostic service, enabling broad adoption across hospitals, cancer centers, and diagnostic laboratories. Reducing the cost and turnaround time required for tumor typing has the potential to expand access to precision medicine, improve clinical decision-making, and reduce healthcare costs. In addition, this platform has the potential to support a wide range of diseases where tissue type informs diagnosis or treatment, improving healthcare accessibility and patient outcomes. This I-Corps project utilizes experiential learning coupled with first-hand investigation of the industry ecosystem to assess the translation potential of an optical phenotyping platform that combines label-free two-photon microscopy with deep learning to classify tumor phenotypes from intrinsic tissue signals. The technology exploits endogenous autofluorescence and second harmonic generation to quantify structural, biochemical, and microenvironmental characteristics of tissue without staining, molecular labeling, or genomic sequencing. Deep neural networks are used to extract high-dimensional image features that enable accurate classification of clinically relevant tumor phenotypes within minutes. Results using this approach demonstrate approximately 90% classification accuracy in pancreatic cancer tissue. This technology may be an improvement over conventional immunohistochemistry and genomic profiling methods by providing rapid, non-destructive, and low-cost molecular surrogate information using optical measurements alone. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF Program Director: Ruth Shuman