Stepwise Transfer Learning for Expert-level Pediatric Brain Tumor MRI Segmentation in a Limited Data Scenario
Aidan Boyd1, Zezhong Ye1, Sanjay P Prabhu1
1From the Artificial Intelligence in Medicine (AIM) Program, Mass General Brigham, Harvard Medical School, Boston, Mass (A.B., Z.Y., Y.Z., A.Z., H.H., R.C., H.J.W.L.A., B.H.K.); Department of Radiation Oncology (A.B., Z.Y., M.C.T., Y.Z., A.Z., H.H., R.C., K.X.L., D.A.H.K., H.J.W.L.A., B.H.K.) and Department of Radiology (H.J.W.L.A.), Brigham and Women's Hospital, Dana-Farber Cancer Institute, Harvard Medical School, 75 Francis St, Boston, MA 02115; Department of Radiology, Boston Children's Hospital, Harvard Medical School, Boston, Mass (S.P.P., S.V., T.Y.P.); Department of Biostatistics and Computational Biology, Harvard T.H. Chan School of Public Health, Boston, Mass (P.J.C.); Center for Data-Driven Discovery in Biomedicine (D3b) (A.N., A.C.R.) and Department of Neurosurgery (A.C.R.), Children's Hospital of Philadelphia, Philadelphia, Pa; Department of Radiology, Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pa (A.N.); Departments of Neurology, Pediatrics, and Neurologic Surgery, University of California, San Francisco, San Francisco, Calif (S.M.); and Department of Radiology and Nuclear Medicine, CARIM & GROW, Maastricht University, Maastricht, the Netherlands (H.J.W.L.A.).
Abstract:
Purpose To develop, externally test, and evaluate clinical acceptability of a deep learning pediatric brain tumor segmentation model using stepwise transfer learning. Materials and Methods In this retrospective study, the authors leveraged two T2-weighted MRI datasets (May 2001 through December 2015) from a national brain tumor consortium (n = 184; median age, 7 years [range, 1-23 years]; 94 male patients) and a pediatric cancer center (n = 100; median age, 8 years [range, 1-19 years]; 47 male patients) to develop and evaluate deep learning neural networks for pediatric low-grade glioma segmentation using a stepwise transfer learning approach to maximize performance in a limited data scenario. The best model was externally tested on an independent test set and subjected to randomized blinded evaluation by three clinicians, wherein they assessed clinical acceptability of expert- and artificial intelligence (AI)-generated segmentations via 10-point Likert scales and Turing tests. Results The best AI model used in-domain stepwise transfer learning (median Dice score coefficient, 0.88 [IQR, 0.72-0.91] vs 0.812 [IQR, 0.56-0.89] for baseline model; P = .049). With external testing, the AI model yielded excellent accuracy using reference standards from three clinical experts (median Dice similarity coefficients: expert 1, 0.83 [IQR, 0.75-0.90]; expert 2, 0.81 [IQR, 0.70-0.89]; expert 3, 0.81 [IQR, 0.68-0.88]; mean accuracy, 0.82). For clinical benchmarking (n = 100 scans), experts rated AI-based segmentations higher on average compared with other experts (median Likert score, 9 [IQR, 7-9] vs 7 [IQR 7-9]) and rated more AI segmentations as clinically acceptable (80.2% vs 65.4%). Experts correctly predicted the origin of AI segmentations in an average of 26.0% of cases. Conclusion Stepwise transfer learning enabled expert-level automated pediatric brain tumor autosegmentation and volumetric measurement with a high level of clinical acceptability. Keywords: Stepwise Transfer Learning, Pediatric Brain Tumors, MRI Segmentation, Deep Learning Supplemental material is available for this article. © RSNA, 2024.
