Detecting Eye Cancer Earlier with Ultrasound and AI

What biomedical issues did you address in your research, and what did your studies find?

My research focused on improving the early detection of choroidal melanoma, the most common primary eye cancer in adults. One major challenge is that current imaging methods can make it difficult to distinguish malignant choroidal melanomas from benign choroidal nevi, which appear as freckle-like spots in the back of the eye. Detecting these tumors earlier and more accurately is especially important because choroidal melanoma can spread to other parts of the body, most often the liver, where outcomes are poor.

To address this problem, I optimized a noninvasive ultrasound imaging technique called quantitative high-definition microvasculature imaging (qHDMI) to visualize tiny blood vessels within choroidal tumors. Since cancer growth is often linked to abnormal blood vessel formation, I wanted to see whether the patterns of these vessels could help distinguish malignant tumors from benign ones.

We found that malignant choroidal melanomas had more complex, twisted and disorganized vessel patterns than benign choroidal nevi. Importantly, these differences were measurable even when the tumors were similar in size, which is one of the primary factors used in current clinical evaluation. I also developed an artificial intelligence (AI) model called "DopNet" that significantly reduced the time required to reconstruct microvascular images. Together, these findings suggest that qHDMI and AI-based tools could help physicians detect choroidal melanoma earlier and more accurately, with the potential for real-time use in the clinic.

What aspects of your training at Mayo helped you grow as a scientist and as a thinker?

The collaborative and patient-centered environment at Mayo Clinic helped me grow beyond the technical aspects of my research. Under the mentorship and close supervision of Mostafa Fatemi, Ph.D., and Azra Alizad, M.D., along with the guidance of Lauren Dalvin, M.D., I learned how important it is to remain calm, focused and thorough when navigating challenges in research. Their guidance provided invaluable experience in conducting translational research and inspired me to think beyond engineering solutions and toward addressing real clinical needs. Since my project relied heavily on clinical data, I learned to approach research questions with patient care in mind and to think critically about how new technologies can move toward clinical use.

Communication across disciplines was another important part of my graduate training. At Mayo, I had opportunities to present my work, collaborate with researchers from different departments, mentor students and learn from experts in imaging, ophthalmology and AI. These experiences helped me become a more thoughtful, adaptable and well-rounded researcher.

What's next?

I hope to continue working at the intersection of biomedical engineering, AI and clinical translation. My goal is to develop ultrasound- and AI-based technologies that support earlier disease detection and make diagnosis faster, safer and more accessible. My training at Mayo strengthened my interest in translational research and showed me how scientific advances can influence clinical decision-making and, ultimately, improve patients' lives.

This article was written by Meredith Lilley, a Ph.D. candidate in Neuroscience at Mayo Clinic Graduate School of Biomedical Sciences.