A Novel Implementation of Machine Learning for the Efficient, Explainable Diagnosis of COVID-19 from Chest CT

TL;DR

This paper presents a machine learning approach using transfer learning and explainability techniques to accurately diagnose COVID-19 from chest CT scans, aiming to improve speed, reliability, and interpretability of diagnostics.

Contribution

The study introduces a novel deep learning framework with explainability for COVID-19 detection from CT scans, utilizing a large dataset and transfer learning with EfficientNetB7.

Findings

Achieved 92.7% accuracy and 95.8% sensitivity in COVID-19 detection.

Utilized explainability to localize infected regions and validate model decisions.

Demonstrated potential for AI to assist radiologists in COVID-19 diagnosis.

Abstract

In a worldwide health crisis as exigent as COVID-19, there has become a pressing need for rapid, reliable diagnostics. Currently, popular testing methods such as reverse transcription polymerase chain reaction (RT-PCR) can have high false negative rates. Consequently, COVID-19 patients are not accurately identified nor treated quickly enough to prevent transmission of the virus. However, the recent rise of medical CT data has presented promising avenues, since CT manifestations contain key characteristics indicative of COVID-19. This study aimed to take a novel approach in the machine learning-based detection of COVID-19 from chest CT scans. First, the dataset utilized in this study was derived from three major sources, comprising a total of 17,698 chest CT slices across 923 patient cases. Image preprocessing algorithms were then developed to reduce noise by excluding irrelevant features. Transfer learning was also implemented with the EfficientNetB7 pre-trained model to provide a backbone architecture and save computational resources. Lastly, several explainability techniques were leveraged to qualitatively validate model performance by localizing infected regions and highlighting fine-grained pixel details. The proposed model attained an overall accuracy of 0.927 and a sensitivity of 0.958. Explainability measures showed that the model correctly distinguished between relevant, critical features pertaining to COVID-19 chest CT images and normal controls. Deep learning frameworks provide efficient, human-interpretable COVID-19 diagnostics that could complement radiologist decisions or serve as an alternative screening tool. Future endeavors may provide insight into infection severity, patient risk stratification, and prognosis.