Task

Abstract

Recent advancements in visual model pre-training have increased the complexity of models, making the clinical deployment of computer-aided diagnosis (CAD) systems cumbersome and resource-intensive. Additionally, the medical imaging community lacks an open-source CAD architecture that could facilitate the rapid construction of fair and accessible platforms. To address these challenges, we introduce UniCAD, a novel approach that leverages the robust capabilities of large pre-trained vision models to enhance 2D/3D medical image analysis.

Specific Expert Preview

AdrenalMNIST3D

DermaMNIST

FractureMNIST3D

INBreast

NIH Chest X-ray

NoduleMNIST3D

OAI

OCTMNIST

PathMNIST

PneumoniaMNIST

RetinaMNIST

VesselMNIST3D