Incorporating functional soft tissue deformations in AI model training for spatially accurate prostate cancer detection.
Authors
Affiliations (16)
Affiliations (16)
- German Cancer Research Center (DKFZ) Heidelberg, Division of Medical Image Computing, Heidelberg, Germany; DKFZ Heidelberg, Division of Radiology, Heidelberg, Germany; Medical Faculty Heidelberg, Heidelberg University, Heidelberg, Germany; HIDSS4Health - Helmholtz Information and Data Science School for Health, Karlsruhe/Heidelberg, Germany.
- DKFZ Heidelberg, Division of Radiology, Heidelberg, Germany.
- German Cancer Research Center (DKFZ) Heidelberg, Division of Medical Image Computing, Heidelberg, Germany; Faculty of Mathematics and Computer Science, Heidelberg University, Heidelberg, Germany; DKFZ Heidelberg, Helmholtz Imaging, Heidelberg, Germany.
- DKFZ Heidelberg, Division of Statistics, Heidelberg, Germany.
- German Cancer Research Center (DKFZ) Heidelberg, Division of Medical Image Computing, Heidelberg, Germany; Medical Faculty Heidelberg, Heidelberg University, Heidelberg, Germany.
- DKFZ Heidelberg, Division of Radiology, Heidelberg, Germany; Medical Faculty Heidelberg, Heidelberg University, Heidelberg, Germany.
- German Cancer Research Center (DKFZ) Heidelberg, Division of Medical Image Computing, Heidelberg, Germany.
- DKFZ Heidelberg, Helmholtz Imaging, Heidelberg, Germany; DKFZ Heidelberg, Interactive Machine Learning Group, Heidelberg, Germany.
- German Cancer Research Center (DKFZ) Heidelberg, Division of Medical Image Computing, Heidelberg, Germany; DKFZ Heidelberg, Helmholtz Imaging, Heidelberg, Germany.
- DKFZ Heidelberg, Junior Clinical Cooperation Unit 'Multiparametric methods for early detection of prostate cancer', Heidelberg, Germany; Department of Urology, University of Heidelberg Medical Center, Germany.
- Department of Urology, University of Heidelberg Medical Center, Germany.
- Institute of Pathology, University of Heidelberg Medical Center, Germany.
- DKFZ Heidelberg, Division of Radiology, Heidelberg, Germany; German Cancer Consortium (DKTK), DKFZ, core center Heidelberg, Germany.
- Mannheim University of Applied Sciences, Mannheim, Germany.
- German Cancer Research Center (DKFZ) Heidelberg, Division of Medical Image Computing, Heidelberg, Germany; HIDSS4Health - Helmholtz Information and Data Science School for Health, Karlsruhe/Heidelberg, Germany; DKFZ Heidelberg, Helmholtz Imaging, Heidelberg, Germany; German Cancer Consortium (DKTK), DKFZ, core center Heidelberg, Germany; Pattern Analysis and Learning Group, Department of Radiation Oncology, Heidelberg University Hospital, Heidelberg, Germany.
- DKFZ Heidelberg, Division of Radiology, Heidelberg, Germany; Medical Faculty Heidelberg, Heidelberg University, Heidelberg, Germany; German Cancer Consortium (DKTK), DKFZ, core center Heidelberg, Germany; National Center for Tumor Diseases (NCT) Heidelberg, Heidelberg, Germany. Electronic address: [email protected].
Abstract
To increase performance and generalization ability of artificial intelligence prostate cancer detection systems by simulating physiological size changes of the bladder and rectum and, thereby, associated deformations of the prostate and its lesions. This retrospective study included 1028 bi-parametric MRI examinations of men (age range: 40-90 years) performed between 2014 and 2019, divided into training/test sets (771/257). We integrated an 'anatomy-informed' transformation into the training of nnU-Net, by simulating soft-tissue deformations of the prostate resulting from size changes of the rectum and bladder. The effects of these strategies were evaluated using free-response receiver operating characteristic (FROC) to assess lesion-level performance, along with a variant: weighted alternative FROC (wAFROC), which prioritizes patient-level effects with localization criteria. Change in sensitivity was tested using a clustered McNemar test. Patient-level performance was assessed with standard and localized receiver operating characteristics (ROC/LROC) analysis. On the independent test set, the anatomy-informed model simulating changes of both rectum and bladder, significantly increased lesion-level detection of true positive lesions by 18.8% (from 48 to 57, p = 0.01) and demonstrated significantly higher performance in the wAFROC analysis (from 0.597 to 0.639, p < 0.01). Patient-level ROC increased slightly (from 0.779 to 0.782, p = 0.89), while LROC analysis demonstrated increased performance (from 0.471 to 0.546). Simulation of rectum and bladder size variations during model training led to significant improvement in lesion detection performance, which may be crucial for diagnostics and therapeutic measures depending on correct lesion localization, e.g. MRI-guided biopsies or focal therapy regimes.