A hybrid machine learning/deep learning COVID-19 severity predictive model from CT images and clinical data.

COVID-19 clinical presentation and prognosis are highly variable, ranging from asymptomatic and paucisymptomatic cases to acute respiratory distress syndrome and multi-organ involvement. We developed a hybrid machine learning/deep learning model to classify patients in two outcome categories, non-ICU and ICU (intensive care admission or death), using 558 patients admitted in a northern Italy hospital in February/May of 2020. A fully 3D patient-level CNN classifier on baseline CT images is used as feature extractor. Features extracted, alongside with laboratory and clinical data, are fed for selection in a Boruta algorithm with SHAP game theoretical values. A classifier is built on the reduced feature space using CatBoost gradient boosting algorithm and reaching a probabilistic AUC of 0.949 on holdout test set. The model aims to provide clinical decision support to medical doctors, with the probability score of belonging to an outcome class and with case-based SHAP interpretation of features importance.

SARS-CoV-2-related encephalitis with prominent parkinsonism: clinical and FDG-PET correlates in two patients.

Considering the similarities with other pandemics due to respiratory virus infections and subsequent development of neurological disorders (e.g. encephalitis lethargica after the 1918 influenza), there is growing concern about a possible new wave of neurological complications following the worldwide spread of SARS-CoV-2. However, data on COVID-19-related encephalitis and movement disorders are still limited. Herein, we describe the clinical and neuroimaging (FDG-PET/CT, MRI and DaT-SPECT) findings of two patients with COVID-19-related encephalopathy who developed prominent parkinsonism. None of the patients had previous history of parkinsonian signs/symptoms, and none had prodromal features of Parkinson's disease (hyposmia or RBD). Both developed a rapidly progressive form of atypical parkinsonism along with distinctive features suggestive of encephalitis. A possible immune-mediated etiology was suggested in Patient 2 by the presence of CSF-restricted oligoclonal bands, but none of the patients responded favorably to immunotherapy. Interestingly, FDG-PET/CT findings were similar in both cases and reminiscent of those observed in post-encephalitic parkinsonism, with cortical hypo-metabolism associated with hyper-metabolism in the brainstem, mesial temporal lobes, and basal ganglia. Patient's FDG-PET/CT findings were validated by performing a Statistical Parametric Mapping analysis and comparing the results with a cohort of healthy controls (n = 48). Cerebrum cortical thickness map was obtained in Patient 1 from MRI examinations to evaluate the structural correlates of the metabolic alterations detected with FDG-PET/CT. Hypermetabolic areas correlated with brain regions showing increased cortical thickness, suggesting their involvement during the inflammatory process. Overall, these observations suggest that SARS-CoV-2 infection may trigger an encephalitis with prominent parkinsonism and distinctive brain metabolic alterations.