uf college of medicine office of research
Led by Pinaki Sarder, Ph.D., CMIL aims to make kidney tissue imaging data accessible for clinical use.
By Manny Rea
The Computational Microscopy Imaging Laboratory (CMIL) is pioneering the integration of artificial intelligence with histopathology, the study of disease at the tissue level, and spatial-omics data, information about active tissue genes and their location in the sample. Focusing on diabetic kidney disease, its latest Nature Communications paper, “FUSION: A web-based application for in-depth exploration of multi-omics data with brightfield histology,” makes tissue imaging data accessible for clinical use.
CMIL is led by Pinaki Sarder, Ph.D., an associate professor of AI in the Section of Quantitative Health, Department of Medicine, and the Associate Director for Imaging at the Intelligent Clinical Care Center (IC³). He and his lab analyze tissue micro-anatomy from digital image processes to uncover the cellular causes of disease.
Histopathology has long relied on the trained eye of pathologists. But many of their assessments are “internal, kind of semi-quantitative measurements,” Sarder said. The paper aimed to turn those insights into objective workflows helping to quantify patient health more continuously and accurately.
At the center of the paper is FUSION, Functional Unit State Identification in Whole Slide Images (WSIs), an AI-driven, web-based tool that overlays spatial-omics data onto traditional stained histology images. Using deep learning and cloud resources, FUSION segments “Functional Tissue Units” (FTUs) in gigapixel-scale histology slides and links these with gene-expression data to visualize healthy versus injured cell types across tissue regions.
Built with a user-friendly front end, FUSION allows researchers and clinicians to upload WSIs and spatial-omics files, explore segmented FTUs overlaid with heatmaps of gene expression, and interactively analyze their data with no advanced coding skills required.
The lab demonstrated the platform using kidney tissue samples, including healthy kidneys, those with chronic kidney disease, and acute kidney injury cases. The kidney served merely as prototype; the tool can be generalized to other organs and future spatial-omics technology.
Behind this innovation is the CMIL team, including the primary developer Samuel Border who successfully defended his Ph.D. thesis in May 2025. He, other lab members, and other collaborators worked on segmentation, system infrastructure, and usability.
CMIL doesn’t just build software, though. It also emphasizes training and accessibility. Through its recent HuBMAP FUSION Research Experience, undergraduate students as well as postdoctoral and faculty researchers used FUSION to annotate and analyze their own histological data, enriching their computational skills.
“Our mission is to unite computational cell biology, anatomy, and pathology under one cohesive framework — cultivating a new generation of renaissance scientists equipped to interpret and design complex biological systems from a truly holistic perspective,” Sarder said.
In line with its tool’s success, CMIL has also secured about $18 million in external grant funding to support its work. More recently, its innovations were spotlighted in major forums: lab members presented three oral abstracts at the European Renal Association Congress 2025 in Vienna and participated in the HuBMAP 2025 Annual Meeting.
Additionally, the editors of Nature Reviews Nephrology, the premier nephrology journal, invited lab member and Ph.D. student Nicholas Lucarelli to describe the development and application of FUSION in a journal article.
“Our lab members are taking the national and international stage to present our work across diverse domains — including kidney transplant, diabetes, lupus nephritis, and acute kidney injury — advancing novel cellular structural and functional analyses, system design, and multimodal AI approaches as powerful vehicles for discovery,” Sarder said.