Automated Detection of Pancreatic Cystic Lesions on CT Using Deep Learning.

摘要: Pancreatic cystic lesions (PCL) are a frequent and underreported incidental finding on CT scans can transform into neoplasms with devastating consequences. We developed evaluated an algorithm based two-step nnU-Net architecture for automated detection of PCL CTs. A total 543 cysts 221 abdominal CTs were manually segmented in 3D by radiology resident consensus board-certified radiologist specialized radiology. This information was used to train the performance assessed depending lesions' volume location comparison three human readers varying experience. Mean sensitivity 78.8 ± 0.1%. The highest large 87.8% ≥220 mm3 distal pancreas up 96.2%. number false-positive detections 0.1 per case. algorithm's comparable readers. To conclude, is feasible. proposed model could serve radiologists as second reading tool. All imaging data code this study freely available online.