TY - JOUR
T1 - Automated detection of cerebral microbleeds on MR images using knowledge distillation framework
AU - Sundaresan, Vaanathi
AU - Arthofer, Christoph
AU - Zamboni, Giovanna
AU - Murchison, Andrew G.
AU - Dineen, Robert A.
AU - Rothwell, Peter M.
AU - Auer, Dorothee P.
AU - Wang, Chaoyue
AU - Miller, Karla L.
AU - Tendler, Benjamin C.
AU - Alfaro-Almagro, Fidel
AU - Sotiropoulos, Stamatios N.
AU - Sprigg, Nikola
AU - Griffanti, Ludovica
AU - Jenkinson, Mark
N1 - Publisher Copyright:
Copyright © 2023 Sundaresan, Arthofer, Zamboni, Murchison, Dineen, Rothwell, Auer, Wang, Miller, Tendler, Alfaro-Almagro, Sotiropoulos, Sprigg, Griffanti and Jenkinson.
PY - 2023
Y1 - 2023
N2 - Introduction: Cerebral microbleeds (CMBs) are associated with white matter damage, and various neurodegenerative and cerebrovascular diseases. CMBs occur as small, circular hypointense lesions on T2*-weighted gradient recalled echo (GRE) and susceptibility-weighted imaging (SWI) images, and hyperintense on quantitative susceptibility mapping (QSM) images due to their paramagnetic nature. Accurate automated detection of CMBs would help to determine quantitative imaging biomarkers (e.g., CMB count) on large datasets. In this work, we propose a fully automated, deep learning-based, 3-step algorithm, using structural and anatomical properties of CMBs from any single input image modality (e.g., GRE/SWI/QSM) for their accurate detections. Methods: In our method, the first step consists of an initial candidate detection step that detects CMBs with high sensitivity. In the second step, candidate discrimination step is performed using a knowledge distillation framework, with a multi-tasking teacher network that guides the student network to classify CMB and non-CMB instances in an offline manner. Finally, a morphological clean-up step further reduces false positives using anatomical constraints. We used four datasets consisting of different modalities specified above, acquired using various protocols and with a variety of pathological and demographic characteristics. Results: On cross-validation within datasets, our method achieved a cluster-wise true positive rate (TPR) of over 90% with an average of <2 false positives per subject. The knowledge distillation framework improves the cluster-wise TPR of the student model by 15%. Our method is flexible in terms of the input modality and provides comparable cluster-wise TPR and better cluster-wise precision compared to existing state-of-the-art methods. When evaluating across different datasets, our method showed good generalizability with a cluster-wise TPR >80 % with different modalities. The python implementation of the proposed method is openly available.
AB - Introduction: Cerebral microbleeds (CMBs) are associated with white matter damage, and various neurodegenerative and cerebrovascular diseases. CMBs occur as small, circular hypointense lesions on T2*-weighted gradient recalled echo (GRE) and susceptibility-weighted imaging (SWI) images, and hyperintense on quantitative susceptibility mapping (QSM) images due to their paramagnetic nature. Accurate automated detection of CMBs would help to determine quantitative imaging biomarkers (e.g., CMB count) on large datasets. In this work, we propose a fully automated, deep learning-based, 3-step algorithm, using structural and anatomical properties of CMBs from any single input image modality (e.g., GRE/SWI/QSM) for their accurate detections. Methods: In our method, the first step consists of an initial candidate detection step that detects CMBs with high sensitivity. In the second step, candidate discrimination step is performed using a knowledge distillation framework, with a multi-tasking teacher network that guides the student network to classify CMB and non-CMB instances in an offline manner. Finally, a morphological clean-up step further reduces false positives using anatomical constraints. We used four datasets consisting of different modalities specified above, acquired using various protocols and with a variety of pathological and demographic characteristics. Results: On cross-validation within datasets, our method achieved a cluster-wise true positive rate (TPR) of over 90% with an average of <2 false positives per subject. The knowledge distillation framework improves the cluster-wise TPR of the student model by 15%. Our method is flexible in terms of the input modality and provides comparable cluster-wise TPR and better cluster-wise precision compared to existing state-of-the-art methods. When evaluating across different datasets, our method showed good generalizability with a cluster-wise TPR >80 % with different modalities. The python implementation of the proposed method is openly available.
KW - cerebral microbleed (CMB)
KW - deep learning
KW - detection
KW - knowledge distillation
KW - magnetic resonance imaging
KW - quantitative susceptibility mapping (QSM)
KW - susceptibility-weighted image (SWI)
UR - http://www.scopus.com/inward/record.url?scp=85165549539&partnerID=8YFLogxK
U2 - 10.3389/fninf.2023.1204186
DO - 10.3389/fninf.2023.1204186
M3 - Article
AN - SCOPUS:85165549539
SN - 1662-5196
VL - 17
JO - Frontiers in Neuroinformatics
JF - Frontiers in Neuroinformatics
M1 - 1204186
ER -