Percival, developed by Penn Medicine, can identify signs of conditions such as atrial fibrillation, heart failure, and chronic kidney disease from routine CT scans and predict patients’ subsequent diagnoses. The model was trained on approximately 400,000 scans and radiology reports from about 50,000 patients and tested on more than 20,000 patients. It scored 0.84 for atrial fibrillation and 0.91 for one type of heart failure, with average scores for predicting subsequent diagnoses of 0.69 and 0.70 for chest and abdominal scans, respectively.
Source:Earth.com · Read Original