Weakly Supervised Context Encoder using DICOM metadata in Ultrasound Imaging

TL;DR

This paper introduces a weakly supervised learning method that utilizes DICOM metadata to improve ultrasound image representation learning, addressing data scarcity issues in clinical AI applications.

Contribution

It presents a novel approach that leverages DICOM metadata for ultrasound image encoding, outperforming traditional non-metadata methods in various tasks.

Findings

Outperforms non-metadata approaches in downstream tasks

Utilizes DICOM metadata to enhance ultrasound image representations

Addresses data scarcity in clinical AI applications

Abstract

Modern deep learning algorithms geared towards clinical adaption rely on a significant amount of high fidelity labeled data. Low-resource settings pose challenges like acquiring high fidelity data and becomes the bottleneck for developing artificial intelligence applications. Ultrasound images, stored in Digital Imaging and Communication in Medicine (DICOM) format, have additional metadata data corresponding to ultrasound image parameters and medical exams. In this work, we leverage DICOM metadata from ultrasound images to help learn representations of the ultrasound image. We demonstrate that the proposed method outperforms the non-metadata based approaches across different downstream tasks.