1250-GRM Multi-Organ Tumor Classification and Zero-Shot Cross-Organ Transfer

Presenter Information

Location

https://campus.kennesaw.edu/colleges-departments/ccse/events/computing-showcase/fa26-cday-program.php

Document Type

Event

Start Date

20-11-2026 2:00 PM

Description

In this study, two methods of medical image analysis were performed using ResNet-50. Experiment 1 trains a shared convolutional backbone with separate four-class heads for brain MRI, kidney CT, and lung CT classification. Experiment 2 tests zero-shot cross-organ transfer: a binary (yes or no) lesion classifier trained only on brain and kidney images then tested on unseen lung CT scans. Zero-shot refers to the model performing a prediction on unseen data without training or fine-tuning. The supervised multi-organ experiment displayed accuracies of 92.95% for the brain, 99.96% for the kidney, and 90.28% for the lung. In contrast, the zero-shot evaluation only displayed an accuracy of 30%. These findings illustrate the challenge of generalizing across organs and imaging domains.

This document is currently not available here.

Share

COinS
 
Nov 20th, 2:00 PM

1250-GRM Multi-Organ Tumor Classification and Zero-Shot Cross-Organ Transfer

https://campus.kennesaw.edu/colleges-departments/ccse/events/computing-showcase/fa26-cday-program.php

In this study, two methods of medical image analysis were performed using ResNet-50. Experiment 1 trains a shared convolutional backbone with separate four-class heads for brain MRI, kidney CT, and lung CT classification. Experiment 2 tests zero-shot cross-organ transfer: a binary (yes or no) lesion classifier trained only on brain and kidney images then tested on unseen lung CT scans. Zero-shot refers to the model performing a prediction on unseen data without training or fine-tuning. The supervised multi-organ experiment displayed accuracies of 92.95% for the brain, 99.96% for the kidney, and 90.28% for the lung. In contrast, the zero-shot evaluation only displayed an accuracy of 30%. These findings illustrate the challenge of generalizing across organs and imaging domains.