Student Research Year Highlights: Lewis and Lodha

The Cleveland Clinic Lerner of College of Medicine of Case Western Reserve University (CCLCM) is a five-year program dedicated to developing the next generation of physician investigators. Students are assigned dedicated physician and research advisors to help ensure they reach their goals. In their fourth year, students work with a mentor to develop a master’s-level thesis in basic science, translational medicine, clinical medicine or health systems. When the students graduate, they each receive an MD with Special Qualification in Biomedical Research from Case Western Reserve University.

The research in which students are involved is critical, timely and often results in presentations at national conferences and publication in scholarly journals. The examples below briefly describe two students’ research projects, one that involves predicting postoperative kidney function in high-risk patients with upper tract urothelial carcinoma and another that involves using a deep learning model to personalize clinical care for patients with breast cancer:

“During my research year at Cleveland Clinic’s Glickman Urological & Kidney Institute, I focused on improving how we predict postoperative kidney function in patients with upper tract urothelial carcinoma (UTUC) undergoing radical nephroureterectomy (RNU). Accurate prediction of new baseline glomerular filtration rate (NBGFR) after RNU is critical, as many patients may become ineligible for cisplatin-based chemotherapy depending on their kidney function. Existing prediction tools often lack precision and do not account for split renal function (SRF), which is especially important in UTUC due to frequent anatomic challenges such as hydronephrosis and infiltrative tumors. Our team developed a novel SRF-based model called PVA+, which incorporates both parenchymal kidney volume and degree of enhancement on contrast-enhanced CT scans.

We conducted imaging analyses using 3D segmentation software on a large patient population undergoing RNU at Cleveland Clinic and performed statistical evaluations of model performance. Compared to existing methods, PVA+ achieved the highest accuracy for predicting NBGFR, especially in patients with moderate hydronephrosis, a group where other models often fail. These findings suggest that PVA+ can serve as a reliable, noninvasive tool to help clinicians decide whether to proceed directly to surgery or offer neoadjuvant chemotherapy first. By addressing a key gap in perioperative decision making, our work offers a practical and scalable solution to better guide treatment in high-risk UTUC patients using only standard preoperative imaging.”

-Kieran Lewis (’26)

“My research focused on developing and validating Path2Space, a deep learning model that predicts spatial transcriptomics directly from routine hematoxylin and eosin (H&E) histopathology images in breast cancer. Spatial transcriptomics enables us to understand gene expression across tissue landscapes, but it remains prohibitively expensive and technically challenging for large-scale clinical use. Working in the Cancer Data Science Laboratory at the National Institutes of Health under Eytan Ruppin, MD, PhD, we sought to overcome this by training machine learning models to predict spatial gene expression directly from routinely collected H&E slides. We validated the model across multiple breast cancer subtypes and external datasets, showing it could generalize well to hormone receptor–positive, HER2-positive and triple-negative cases in both fresh frozen and formalin-fixed samples.

We also developed a novel spatial metric to quantify spatial co-variation between gene pairs, which outperformed traditional methods in predicting patient survival. Ultimately, our results demonstrate the potential to democratize personalized clinical care by enabling large-scale, low-cost spatial transcriptomics directly from existing pathology slides. This could revolutionize biomarker discovery, treatment prediction and clinical decision making in oncology.”

-Roshan Lodha (’26)