| Sumario: | Purpose: To develop an end-to-end DL model for automated classification of affected territory in DWI of stroke patients. Materials and methods: In this retrospective multicenter study, brain DWI studies from January 2017 to April 2020 from Center 1, from June 2020 to December 2020 from Center 2, and from November 2019 to April 2020 from Center 3 were included. Four radiologists labeled images into five classes: anterior cerebral artery (ACA), middle cerebral artery (MCA), posterior circulation (PC), and watershed (WS) regions, as well as normal images. Additionally, for Center 1, clinical information was encoded as a domain knowledge vector to incorporate into image embeddings. 3D convolutional neural network (CNN) and attention gate integrated versions for direct 3D encoding, long short-term memory (LSTM-CNN), and time-distributed layer for slice-based encoding were employed. Balanced classification accuracy, macro averaged f1 score, AUC, and interrater Cohen's kappa were calculated. Results: Overall, 624 DWI MRIs from 3 centers were utilized (mean age, interval: 66.89 years, 29–95 years; 345 male) with 439 patients in the training, 103 in the validation, and 82 in the test sets. The best model was a slice-based parallel encoding model with 0.88 balanced accuracy, 0.80 macro-f1 score, and an AUC of 0.98. Clinical domain knowledge integration improved the performance with 0.93 best overall accuracy with parallel stream model embeddings and support vector machine classifiers. The mean kappa value for interrater agreement was 0.87. Conclusion: Developed end-to-end deep learning models performed well in classifying affected regions from stroke in DWI. Clinical relevance statement: The end-to-end deep learning model with a parallel stream encoding strategy for classifying stroke regions in DWI has performed comparably with radiologists. Key results: • Parallel encoding of slices followed by application of convolutional block attention module was the best performing image-based deep learning approach with 0.88 balanced accuracy, 0.80 macro-f1 score, and 0.98 AUC. • Integration of clinical domain knowledge to the image embeddings and support vector machine-based training increased the classification accuracy to 0.93. • Ablation studies with reduced class number and pocket algorithm with image embeddings from different encoding strategies yielded different winner strategies, which was interpreted as evidence of the strong requirement for dataset and training methodology matching.
|
|---|