AI-based multi-modal integration (ScanCov scores) of clinical characteristics, lab tests and chest CTs improves COVID-19 outcome prediction of hospitalized patients

Abstract

<jats:p>The SARS-COV-2 pandemic has put pressure on Intensive Care Units, and made the identification of early predictors of disease severity a priority. We collected clinical, biological, chest CT scan data, and radiology reports from 1,003 coronavirus-infected patients from two French hospitals. Among 58 variables measured at admission, 11 clinical and 3 radiological variables were associated with severity. Next, using 506,341 chest CT images, we trained and evaluated deep learning models to segment the scans and reproduce radiologists' annotations. We also built CT image-based deep learning models that predicted severity better than models based on the radiologists' reports. Finally, we showed that adding CT scan information-either through radiologist lesion quantification or through deep learning-to clinical and biological data, improves prediction of severity. These findings show that CT scans contain novel and unique prognostic information, which we included in a 6-variable ScanCov severity score.</jats:p>

Common pitfalls and recommendations for using machine learning to detect and prognosticate for COVID-19 using chest radiographs and CT scans

“…Nineteen papers developed models for the prognosis of patients with COVID-19 51,[63][64][65][66][67][68][69][70][71][72][73][74][75][76][77][78][79][80] , fifteen using CT and four using CXR. These models were developed for predicting severity of outcomes including: death or need for ventilation 72,78,79 , a need for intensive care unit (ICU) admission 63,73,[77][78][79] , progression to acute respiratory distress syndrome 80 , the length of hospital stay 51,74 , likelihood of conversion to severe disease 64,65,75 and the extent of lung infection 76 . Most papers used models based on a multivariable Cox proportional hazards model 51,72,78,79 , logistic regression 65,[73][74][75]80 , linear regression 75,76 , random forest 74,7...…”

“…These models were developed for predicting severity of outcomes including: death or need for ventilation 72,78,79 , a need for intensive care unit (ICU) admission 63,73,[77][78][79] , progression to acute respiratory distress syndrome 80 , the length of hospital stay 51,74 , likelihood of conversion to severe disease 64,65,75 and the extent of lung infection 76 . Most papers used models based on a multivariable Cox proportional hazards model 51,72,78,79 , logistic regression 65,[73][74][75]80 , linear regression 75,76 , random forest 74,77 or compare a huge variety of machine learning models such as tree-based methods, support vector machines, neural networks and nearestneighbour clustering 63,64 .…”

“…Predictors from radiological data were extracted using either handcrafted radiomic features 63,64,[68][69][70]72,74,75,[77][78][79][80] or deep learning 51,66,70,71,73,76 . Clinical data included basic observations, serology and comorbidities.…”

“…Clinical data included basic observations, serology and comorbidities. Only eight models integrated both radiological and clinical data 62,63,69,72,73,[77][78][79] .…”

“…Including both internal and external validation allows more insight to generalizability of the algorithm. We found that 48/62 papers consider internal validation only, with 13/62 using external validation 22,32,41,42,51,54,63,66,67,69,73,78,79 . Twelve used truly external test datasets and one tested on the same data the algorithm was trained on 32 .…”