Learning shape distributions from large databases of healthy organs: applications to zero-shot and few-shot abnormal pancreas detection

TL;DR

This paper introduces a scalable, data-driven method to learn healthy organ shape distributions using variational auto-encoders, enabling zero-shot and few-shot abnormal pancreas detection with promising results.

Contribution

It presents a novel approach to model shape distributions from large healthy organ databases for improved zero-shot and few-shot abnormality detection.

Findings

Achieved up to 65.41% AUC in zero-shot detection.

Reached 78.97% AUC with only 15 abnormal examples.

Outperformed volume-based baseline methods.

Abstract

We propose a scalable and data-driven approach to learn shape distributions from large databases of healthy organs. To do so, volumetric segmentation masks are embedded into a common probabilistic shape space that is learned with a variational auto-encoding network. The resulting latent shape representations are leveraged to derive zeroshot and few-shot methods for abnormal shape detection. The proposed distribution learning approach is illustrated on a large database of 1200 healthy pancreas shapes. Downstream qualitative and quantitative experiments are conducted on a separate test set of 224 pancreas from patients with mixed conditions. The abnormal pancreas detection AUC reached up to 65.41% in the zero-shot configuration, and 78.97% in the few-shot configuration with as few as 15 abnormal examples, outperforming a baseline approach based on the sole volume.