ABSTRACT
Introduction: Patients with MS are MRI scanned continuously throughout their disease course resulting in a large manual workload for radiologists which includes lesion detection and size estimation. Though many models for automatic lesion segmentation have been published, few are used broadly in clinic today, as there is a lack of testing on clinical datasets. By collecting a large, heterogeneous training dataset directly from our MS clinic we aim to present a model which is robust to different scanner protocols and artefacts and which only uses MRI modalities present in routine clinical examinations. Methods: We retrospectively included 746 patients from routine examinations at our MS clinic. The inclusion criteria included acquisition at one of seven different scanners and an MRI protocol including 2D or 3D T2-w FLAIR, T2-w and T1-w images. Reference lesion masks on the training (n = 571) and validation (n = 70) datasets were generated using a preliminary segmentation model and subsequent manual correction. The test dataset (n = 100) was manually delineated. Our segmentation model https://github.com/CAAI/AIMS/ was based on the popular nnU-Net, which has won several biomedical segmentation challenges. We tested our model against the published segmentation models HD-MS-Lesions, which is also based on nnU-Net, trained with a more homogenous patient cohort. We furthermore tested model robustness to data from unseen scanners by performing a leave-one-scanner-out experiment. Results: We found that our model was able to segment MS white matter lesions with a performance comparable to literature: DSC = 0.68, precision = 0.90, recall = 0.70, f1 = 0.78. Furthermore, the model outperformed HD-MS-Lesions in all metrics except precision = 0.96. In the leave-one-scanner-out experiment there was no significant change in performance (p < 0.05) between any of the models which were only trained on part of the dataset and the full segmentation model. Conclusion: In conclusion we have seen, that by including a large, heterogeneous dataset emulating clinical reality, we have trained a segmentation model which maintains a high segmentation performance while being robust to data from unseen scanners. This broadens the applicability of the model in clinic and paves the way for clinical implementation.
ABSTRACT
PURPOSE: To implement and validate an existing algorithm for automatic delineation of white matter lesions on magnetic resonance imaging (MRI) in patients with multiple sclerosis (MS) on a local single-center dataset. METHODS: We implemented a white matter hyperintensity segmentation model, based on a 2D convolutional neural network, using the conventional T2-weighted fluid attenuated inversion recovery (FLAIR) MRI sequence as input. The model was adapted for delineation of MS lesions by further training on a local dataset of 93 MS patients with a total of 3040 lesions. A quantitative evaluation was performed on ten test patients, in which model-generated masks were compared to manually delineated masks from two expert delineators. A subsequent qualitative evaluation of the implemented model was performed by two expert delineators, in which generated delineation masks on a clinical dataset of 53 patients were rated acceptable (<â¯10% errors) or unacceptable (>â¯10% errors) based on the total number of true lesions. RESULTS: The quantitative evaluation resulted in an average accuracy score (F1) of 0.71, recall of 0.77 and dice similarity coefficient of 0.62. Our implemented model obtained the highest scores in all three metrics, when compared to three out of the box lesion segmentation models. In the clinical evaluation an average of 94% of our 53 model-generated masks were rated acceptable. CONCLUSION: After adaptation to our local dataset, the implemented segmentation model was able to delineate MS lesions with a high clinical value as rated by delineation experts while outperforming popular out of the box applications. This serves as a promising step towards implementation of automatic lesion delineation in our MS clinic.
