Checklist for responsible deep learning modeling of medical images based on COVID-19 detection studies

Abstract

The sudden outbreak and uncontrolled spread of COVID-19 disease is one of the most important global problems today. In a short period of time, it has led to the development of many deep neural network models for COVID-19 detection with modules for explainability. In this work, we carry out a systematic analysis of various aspects of proposed models. Our analysis revealed numerous mistakes made at different stages of data acquisition, model development, and explanation construction. In this work, we overview the approaches proposed in the surveyed Machine Learning articles and indicate typical errors emerging from the lack of deep understanding of the radiography domain. We present the perspective of both: experts in the field - radiologists and deep learning engineers dealing with model explanations. The final result is a proposed checklist with the minimum conditions to be met by a reliable COVID-19 diagnostic model.

Keywords: COVID-19; Computed tomography; Deep learning; Explainable AI; Lungs; X-ray.

Fig. 1

PRISMA Flow Diagram shows the flow of information through the different phases of a systematic review including inclusions and exclusions.

Fig. 2

Taxonomy of AI applications in 25 reviewed studies.

Fig. 3

Differences between AP and PA chest projections.

Fig. 4

Examples of explanations for COVID-related models from studies: , , , , , . The following explanations are used: a) Grad-CAM, b) CAM, c) saliency, d) guided backpropagation, e) integrated gradients, f) LIME. Such explanations can be divided into 4 types: heat maps (image a) - c)), contour lines (d)), points (e)), and image pieces (f)).

Fig. 5

Examples of biased model explanations: a) , b) , c) , d) . Red arrows in the image b) are marked by a radiologist to help locate the lesions. They were not present in the training set.