Transferability limitations for Covid 3D Localization Using SARS-CoV-2 segmentation models in 4D CT images

TL;DR

This study examines the transferability challenges of deep learning models for Covid-19 lung segmentation in 4D CT images, highlighting that multiple retrainings on large datasets can reduce accuracy.

Contribution

It introduces a 4-channel input approach for segmentation and evaluates transferability limitations across multiple datasets, emphasizing careful use of transfer learning.

Findings

Transferability should be used carefully in Covid segmentation models.

Retraining multiple times on large datasets can decrease accuracy.

Proxy lung masks can be generated when not available.

Abstract

In this paper, we investigate the transferability limitations when using deep learning models, for semantic segmentation of pneumonia-infected areas in CT images. The proposed approach adopts a 4 channel input; 3 channels based on Hounsfield scale, plus one channel (binary) denoting the lung area. We used 3 different, publicly available, CT datasets. If the lung area mask was not available, a deep learning model generates a proxy image. Experimental results suggesting that transferability should be used carefully, when creating Covid segmentation models; retraining the model more than one times in large sets of data results in a decrease in segmentation accuracy.