Deep Bayesian networks for uncertainty estimation and adversarial resistance of white matter hyperintensity segmentation. Issue 7 (28th January 2022)

• Record Type: Journal Article
• Authors: Mojiri Forooshani, Parisa; Biparva, Mahdi; Ntiri, Emmanuel E.; Ramirez, Joel; Boone, Lyndon; Holmes, Melissa F.; Adamo, Sabrina; Gao, Fuqiang; Ozzoude, Miracle; Scott, Christopher J. M.; Dowlatshahi, Dar; Lawrence‐Dewar, Jane M.; Kwan, Donna; Lang, Anthony E.; Marcotte, Karine; Leonard, Carol; Rochon, Elizabeth; Heyn, Chris; Bartha, Robert; Strother, Stephen; Tardif, Jean‐Claude; Symons, Sean; Masellis, Mario; Swartz, Richard H.; Moody, Alan; Black, Sandra E.; Goubran, Maged
• Is Part Of: Human brain mapping. Volume 43:Issue 7(2022)
• Page Start: 2089
• Page End: 2108
• Publication Date: 2022-01-28
• Subjects: adversarial attacks -- Bayesian neural networks -- deep learning -- image segmentation -- uncertainty estimation -- vascular lesions -- white matter hyperintensity; Brain mapping -- Periodicals; 611.81
• DOI: 10.1002/hbm.25784
• Languages: English
• ISSNs: 1065-9471

Abstract

Abstract: White matter hyperintensities (WMHs) are frequently observed on structural neuroimaging of elderly populations and are associated with cognitive decline and increased risk of dementia. Many existing WMH segmentation algorithms produce suboptimal results in populations with vascular lesions or brain atrophy, or require parameter tuning and are computationally expensive. Additionally, most algorithms do not generate a confidence estimate of segmentation quality, limiting their interpretation. MRI‐based segmentation methods are often sensitive to acquisition protocols, scanners, noise‐level, and image contrast, failing to generalize to other populations and out‐of‐distribution datasets. Given these concerns, we propose a novel Bayesian 3D convolutional neural network with a U‐Net architecture that automatically segments WMH, provides uncertainty estimates of the segmentation output for quality control, and is robust to changes in acquisition protocols. We also provide a second model to differentiate deep and periventricular WMH. Four hundred thirty‐two subjects were recruited to train the CNNs from four multisite imaging studies. A separate test set of 158 subjects was used for evaluation, including an unseen multisite study. We compared our model to two established state‐of‐the‐art techniques (BIANCA and DeepMedic), highlighting its accuracy and efficiency. Our Bayesian 3D U‐Net achieved the highest Dice similarity coefficient of 0.89 ± 0.08 and the lowest modified Hausdorff distance of 2.98 ± 4.40 mm. We further validated our models highlighting their robustness on "clinical adversarial cases" simulating data with low signal‐to‐noise ratio, low resolution, and different contrast (stemming from MRI sequences with different parameters). Our pipeline and models are available at: https://hypermapp3r.readthedocs.io . Abstract : We present a robust and efficient WMH segmentation model, which also generates an uncertainty map for quality control. In addition, we present a second model to classify dWMH and pvWMH using the initial total WMH segmentation. Our segmentation models achieved high accuracy compared to SOTA algorithms on a wide spectrum of WMH burdens, especially mild WMH. Additionally, we used an augmentation scheme to make our model robust to simulated images with SNR, low resolution, and different contrasts.