Kidney Cancer Detection Using 3D-Based Latent Diffusion Models

TL;DR

This paper introduces a novel 3D latent diffusion pipeline for kidney anomaly detection in CT scans, leveraging weak supervision and generative models, showing promising results for annotation-efficient medical imaging analysis.

Contribution

The work presents a new 3D latent diffusion approach combining DDPMs, DDIMs, and VQ-GANs for weakly supervised kidney anomaly detection directly on image volumes.

Findings

Operates directly on 3D image volumes, unlike slice-wise methods.

Demonstrates feasibility of weakly supervised anomaly detection with generative models.

Current results suggest directions for improving reconstruction and localization.

Abstract

In this work, we present a novel latent diffusion-based pipeline for 3D kidney anomaly detection on contrast-enhanced abdominal CT. The method combines Denoising Diffusion Probabilistic Models (DDPMs), Denoising Diffusion Implicit Models (DDIMs), and Vector-Quantized Generative Adversarial Networks (VQ-GANs). Unlike prior slice-wise approaches, our method operates directly on an image volume and leverages weak supervision with only case-level pseudo-labels. We benchmark our approach against state-of-the-art supervised segmentation and detection models. This study demonstrates the feasibility and promise of 3D latent diffusion for weakly supervised anomaly detection. While the current results do not yet match supervised baselines, they reveal key directions for improving reconstruction fidelity and lesion localization. Our findings provide an important step toward annotation-efficient, generative modeling of complex abdominal anatomy.