MedYOLO: A Medical Image Object Detection Framework

TL;DR

MedYOLO is a 3-D object detection framework tailored for medical imaging that reduces annotation effort and performs well on medium and large structures across multiple datasets.

Contribution

This paper introduces MedYOLO, a novel 3-D object detection framework based on YOLO for medical imaging, addressing the lack of general-purpose detection models in this domain.

Findings

High accuracy on medium and large structures

Effective across diverse datasets

Limited performance on small or rare structures

Abstract

Artificial intelligence-enhanced identification of organs, lesions, and other structures in medical imaging is typically done using convolutional neural networks (CNNs) designed to make voxel-accurate segmentations of the region of interest. However, the labels required to train these CNNs are time-consuming to generate and require attention from subject matter experts to ensure quality. For tasks where voxel-level precision is not required, object detection models offer a viable alternative that can reduce annotation effort. Despite this potential application, there are few options for general purpose object detection frameworks available for 3-D medical imaging. We report on MedYOLO, a 3-D object detection framework using the one-shot detection method of the YOLO family of models and designed for use with medical imaging. We tested this model on four different datasets: BRaTS, LIDC, an abdominal organ Computed Tomography (CT) dataset, and an ECG-gated heart CT dataset. We found our models achieve high performance on commonly present medium and large-sized structures such as the heart, liver, and pancreas even without hyperparameter tuning. However, the models struggle with very small or rarely present structures.