Deep learning-based hybrid clinical decision support system algorithm for COVID-19 diagnosis via PCR graphics and Thorax CT images, preliminary data

Abstract

Introduction: The limited sensitivity of microbiological testing, challenges in radiological differential diagnosis, and expectations of quick and accurate diagnosis required developing clinical decision support systems (CDSS). We propose a new deep learning-based hybrid CDSS that combines the advantageous aspects of thorax computed tomography(CT) and reverse transcriptase-polymerase chain reaction(PCR) to overcome the weakness of each one.

Methods: We retrospectively constructed a database that contains CT images of healthy subjects and patients with COVID-19 pneumonia(CP), bacterial/viral pneumonia(BVP), interstitial lung diseases(ILD), and PCR data of patients who were tested positive and negative for SARS-CoV-2. A new 3D-convolutional neural network (3D-CNN) and long short-term memory network(LSTM) based CDSS is developed to perform accurate and robust detection of COVID-19 using CT images and PCR data.

Results: Performance results of the proposed models (Fig1) provide highly reliable diagnosis of COVID-19 with 93.2% and 99.7% AUC for CT and PCR data, respectively.

Conclusion: Proposed CDSS with state-of-the-art deep learning methods provides similar performance compared to both radiologists in CT evaluation and microbiologists in PCR evaluation and can be safely used. We plan to develop a hybrid CDSS algorithm further, combining laboratory data with CT and PCR models.

• Download figure
• Open in new tab
• Download powerpoint