Deep Active Lesion Segmentation

TL;DR

Deep Active Lesion Segmentation (DALS) combines CNNs and active contour models to improve lesion segmentation accuracy across multiple organs and imaging modalities, especially with limited training data.

Contribution

Introduces a novel fully automated segmentation framework that integrates CNNs with an improved level-set ACM for precise lesion boundary detection.

Findings

Outperforms competing methods on the MLS dataset.

Effective with small training datasets.

Works across various organs and imaging modalities.

Abstract

Lesion segmentation is an important problem in computer-assisted diagnosis that remains challenging due to the prevalence of low contrast, irregular boundaries that are unamenable to shape priors. We introduce Deep Active Lesion Segmentation (DALS), a fully automated segmentation framework for that leverages the powerful nonlinear feature extraction abilities of fully Convolutional Neural Networks (CNNs) and the precise boundary delineation abilities of Active Contour Models (ACMs). Our DALS framework benefits from an improved level-set ACM formulation with a per-pixel-parameterized energy functional and a novel multiscale encoder-decoder CNN that learns an initialization probability map along with parameter maps for the ACM. We evaluate our lesion segmentation model on a new Multiorgan Lesion Segmentation (MLS) dataset that contains images of various organs, including brain, liver, and lung, across different imaging modalities---MR and CT. Our results demonstrate favorable performance compared to competing methods, especially for small training datasets. Source code : $\text{https://github.com/ahatamiz/dals}$