Mary Gates Endowment for Students

Sri Varshitha Pinnaka

Project

Virtual Hematoxylin and Eosin Colorization for Point-of-Care Breast Cancer Detection

Early identification of breast cancer is imperative for improving patient outcomes, but is limited in rural settings. Currently, CoreView is a lab-on-a-chip technology developed by the Human Photonics Lab at the University of Washington that offers rapid, automated point-of-care diagnostics for breast cancer tissue samples. However, the colorization of raw digital images of breast cancer tissue samples within CoreView is inadequate. These images do not resemble commonly used hematoxylin and eosin (H&E) stains, which is the current gold standard for definitive cancer diagnoses. CoreView’s images are often too dark and lack proper definition of cellular features when converted to H&E standard colorization. To mitigate this bottleneck, I am testing different computational H&E conversion methods. Methods tested thus far include manual image processing using GIMP/Sedeen viewer, color curve-based corrections, and hue/saturation/brightness (HSV) based corrections. I am also testing machine learning-based methods, including multiple instance learning/attention-based and StainGAN-based models. These methods will be evaluated against each other using feedback from practicing pathologists specializing in breast pathology, as they comprise the current standard for pathology. We expect that machine learning based methods will have the highest performance. In the coming months, these methods will be integrated into CoreView’s main functionality, and we expect that this technology will improve accessibility to reliable diagnostics in low-resource environments.

What I learned

While working on this project, I tried new things outside my comfort zone, and not all of them worked. I learned that designing useful and effective point-of-care diagnostic tools requires a lot of creativity and collaboration. Especially in a high-stress environment, not being afraid to ask questions and express your opinion is important for building confidence in your abilities and developing your skill set.

Advice for future applicants

Don't be afraid to ask for feedback! Your first draft will likely be far from perfect, but the only way to improve your application is to ask for help.

More scholar profiles