Automated detection of cerebral microbleeds on MR images using knowledge distillation framework
<p><b>Introduction:</b> 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 imagin...
| Main Authors: | , , , , , , , , , , , , , , |
|---|---|
| Format: | Journal article |
| Language: | English |
| Published: |
Frontiers Media
2023
|
| _version_ | 1824458757512364032 |
|---|---|
| author | Sundaresan, V Arthofer, C Zamboni, G Murchison, AG Dineen, RA Rothwell, PM Auer, DP Wang, C Miller, KL Tendler, BC Alfaro-Almagro, F Sotiropoulos, SN Sprigg, N Griffanti, L Jenkinson, M |
| author_facet | Sundaresan, V Arthofer, C Zamboni, G Murchison, AG Dineen, RA Rothwell, PM Auer, DP Wang, C Miller, KL Tendler, BC Alfaro-Almagro, F Sotiropoulos, SN Sprigg, N Griffanti, L Jenkinson, M |
| author_sort | Sundaresan, V |
| collection | OXFORD |
| description | <p><b>Introduction:</b> 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.</p>
<p><b>Methods:</b> 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.</p>
<p><b>Results:</b> 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.</p> |
| first_indexed | 2025-02-19T04:30:58Z |
| format | Journal article |
| id | oxford-uuid:07053d31-5d90-4aed-9198-657e186b3463 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2025-02-19T04:30:58Z |
| publishDate | 2023 |
| publisher | Frontiers Media |
| record_format | dspace |
| spelling | oxford-uuid:07053d31-5d90-4aed-9198-657e186b34632025-01-03T11:43:48ZAutomated detection of cerebral microbleeds on MR images using knowledge distillation frameworkJournal articlehttp://purl.org/coar/resource_type/c_dcae04bcuuid:07053d31-5d90-4aed-9198-657e186b3463EnglishSymplectic ElementsFrontiers Media2023Sundaresan, VArthofer, CZamboni, GMurchison, AGDineen, RARothwell, PMAuer, DPWang, CMiller, KLTendler, BCAlfaro-Almagro, FSotiropoulos, SNSprigg, NGriffanti, LJenkinson, M<p><b>Introduction:</b> 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.</p> <p><b>Methods:</b> 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.</p> <p><b>Results:</b> 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.</p> |
| spellingShingle | Sundaresan, V Arthofer, C Zamboni, G Murchison, AG Dineen, RA Rothwell, PM Auer, DP Wang, C Miller, KL Tendler, BC Alfaro-Almagro, F Sotiropoulos, SN Sprigg, N Griffanti, L Jenkinson, M Automated detection of cerebral microbleeds on MR images using knowledge distillation framework |
| title | Automated detection of cerebral microbleeds on MR images using knowledge distillation framework |
| title_full | Automated detection of cerebral microbleeds on MR images using knowledge distillation framework |
| title_fullStr | Automated detection of cerebral microbleeds on MR images using knowledge distillation framework |
| title_full_unstemmed | Automated detection of cerebral microbleeds on MR images using knowledge distillation framework |
| title_short | Automated detection of cerebral microbleeds on MR images using knowledge distillation framework |
| title_sort | automated detection of cerebral microbleeds on mr images using knowledge distillation framework |
| work_keys_str_mv | AT sundaresanv automateddetectionofcerebralmicrobleedsonmrimagesusingknowledgedistillationframework AT arthoferc automateddetectionofcerebralmicrobleedsonmrimagesusingknowledgedistillationframework AT zambonig automateddetectionofcerebralmicrobleedsonmrimagesusingknowledgedistillationframework AT murchisonag automateddetectionofcerebralmicrobleedsonmrimagesusingknowledgedistillationframework AT dineenra automateddetectionofcerebralmicrobleedsonmrimagesusingknowledgedistillationframework AT rothwellpm automateddetectionofcerebralmicrobleedsonmrimagesusingknowledgedistillationframework AT auerdp automateddetectionofcerebralmicrobleedsonmrimagesusingknowledgedistillationframework AT wangc automateddetectionofcerebralmicrobleedsonmrimagesusingknowledgedistillationframework AT millerkl automateddetectionofcerebralmicrobleedsonmrimagesusingknowledgedistillationframework AT tendlerbc automateddetectionofcerebralmicrobleedsonmrimagesusingknowledgedistillationframework AT alfaroalmagrof automateddetectionofcerebralmicrobleedsonmrimagesusingknowledgedistillationframework AT sotiropoulossn automateddetectionofcerebralmicrobleedsonmrimagesusingknowledgedistillationframework AT spriggn automateddetectionofcerebralmicrobleedsonmrimagesusingknowledgedistillationframework AT griffantil automateddetectionofcerebralmicrobleedsonmrimagesusingknowledgedistillationframework AT jenkinsonm automateddetectionofcerebralmicrobleedsonmrimagesusingknowledgedistillationframework |