An explainable AI approach for diagnosis of COVID-19 using MALDI-ToF mass spectrometry

Current artificial intelligence (AI) applications for the diagnosis of coronavirus disease 2019 (COVID-19) often lack a biological foundation in decision-making process. In this study, we have employed AI COVID-19 using mass spectrometry (MS) data and leveraged explainable (X-AI) to explain process on local (per-sample) global (all samples) basis. We first assessed eight machine learning models five feature engineering techniques five-fold stratified cross-validation. The best accuracy was achieved by Random Forest (RF) classifier ratio areas under curve (AUC) from MS as features. These features were chosen basis tentatively representing both human viral proteins gargle samples. evaluated RF 70%−30% train-test split strategy 152 samples, yielding an 94.12% test dataset. Employing X-AI, further interpreted model shapely additive explanations (SHAP) importance techniques, including permutation impurity-based importance. With these interpretation offering explanation decisions, devised straightforward, four-stage X-AI framework that can enable medical practitioners understand mechanisms black-box model. To practitioner, instills trust providing rationales it’s decisions.