DualDistill: a dual-guided self-distillation approach for carotid plaque analysis
Accurate classification of carotid plaques is critical to assessing the risk of cardiovascular disease. However, this task remains challenging due to several factors: temporal discontinuity caused by probe motion, the small size of plaques combined with interference from surrounding tissue, and the...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Frontiers Media S.A.
2025-05-01
|
| Series: | Frontiers in Medicine |
| Subjects: | |
| Online Access: | https://www.frontiersin.org/articles/10.3389/fmed.2025.1554578/full |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849727012957061120 |
|---|---|
| author | Xiaoman Zhang Jiang Xie Haibing Chen Haiya Wang |
| author_facet | Xiaoman Zhang Jiang Xie Haibing Chen Haiya Wang |
| author_sort | Xiaoman Zhang |
| collection | DOAJ |
| description | Accurate classification of carotid plaques is critical to assessing the risk of cardiovascular disease. However, this task remains challenging due to several factors: temporal discontinuity caused by probe motion, the small size of plaques combined with interference from surrounding tissue, and the limited availability of annotated data, which often leads to overfitting in deep learning models. To address these challenges, this study introduces a structured self-distillation framework, named DualDistill, designed to improve classification accuracy and generalization performance in analyzing ultrasound videos of carotid plaques. DualDistill incorporates two novel strategies to address the identified challenges. First, an intra-frame relationship-guided strategy is proposed to capture long-term temporal dependencies, effectively addressing temporal discontinuity. Second, a spatial-temporal attention-guided strategy is developed to reduce the impact of irrelevant features and noise by emphasizing relevant regions within both spatial and temporal dimensions. These strategies jointly act as supervisory signals within the self-distillation process, guiding the student layers to better align with the critical features identified by the teacher layers. Besides, the self-distillation process acts as an implicit regularization mechanism, which decreases overfitting in limited datasets. DualDistill is designed as a plug-and-play framework, enabling seamless integration with various existing models. Extensive experiments were conducted on 317 carotid plaque ultrasound videos collected from a collaborating hospital. The proposed framework demonstrated its versatility and effectiveness. It achieved consistent improvements in classification accuracy across 13 representative models. Specifically, the average accuracy improvement is 2.97%, with the maximum improvement reaching 4.74% on 3D ResNet50. These results highlight the robustness and generalizability of DualDistill. It shows strong potential for reliable cardiovascular risk assessment through automated carotid plaque classification. |
| format | Article |
| id | doaj-art-d297bb4f06ef4a9f96c19a481e4b937c |
| institution | DOAJ |
| issn | 2296-858X |
| language | English |
| publishDate | 2025-05-01 |
| publisher | Frontiers Media S.A. |
| record_format | Article |
| series | Frontiers in Medicine |
| spelling | doaj-art-d297bb4f06ef4a9f96c19a481e4b937c2025-08-20T03:09:59ZengFrontiers Media S.A.Frontiers in Medicine2296-858X2025-05-011210.3389/fmed.2025.15545781554578DualDistill: a dual-guided self-distillation approach for carotid plaque analysisXiaoman Zhang0Jiang Xie1Haibing Chen2Haiya Wang3School of Medicine, Shanghai University, Shanghai, ChinaSchool of Computer Engineering and Science, Shanghai University, Shanghai, ChinaUltrasonic Center, Luodian Hospital, Shanghai, ChinaDepartment of Geriatrics, Shanghai Ninth People's Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, ChinaAccurate classification of carotid plaques is critical to assessing the risk of cardiovascular disease. However, this task remains challenging due to several factors: temporal discontinuity caused by probe motion, the small size of plaques combined with interference from surrounding tissue, and the limited availability of annotated data, which often leads to overfitting in deep learning models. To address these challenges, this study introduces a structured self-distillation framework, named DualDistill, designed to improve classification accuracy and generalization performance in analyzing ultrasound videos of carotid plaques. DualDistill incorporates two novel strategies to address the identified challenges. First, an intra-frame relationship-guided strategy is proposed to capture long-term temporal dependencies, effectively addressing temporal discontinuity. Second, a spatial-temporal attention-guided strategy is developed to reduce the impact of irrelevant features and noise by emphasizing relevant regions within both spatial and temporal dimensions. These strategies jointly act as supervisory signals within the self-distillation process, guiding the student layers to better align with the critical features identified by the teacher layers. Besides, the self-distillation process acts as an implicit regularization mechanism, which decreases overfitting in limited datasets. DualDistill is designed as a plug-and-play framework, enabling seamless integration with various existing models. Extensive experiments were conducted on 317 carotid plaque ultrasound videos collected from a collaborating hospital. The proposed framework demonstrated its versatility and effectiveness. It achieved consistent improvements in classification accuracy across 13 representative models. Specifically, the average accuracy improvement is 2.97%, with the maximum improvement reaching 4.74% on 3D ResNet50. These results highlight the robustness and generalizability of DualDistill. It shows strong potential for reliable cardiovascular risk assessment through automated carotid plaque classification.https://www.frontiersin.org/articles/10.3389/fmed.2025.1554578/fullultrasound video classificationcarotid plaque recognitionself-distillationspatial-temporal attentionintra-frame relationshipdeep learning |
| spellingShingle | Xiaoman Zhang Jiang Xie Haibing Chen Haiya Wang DualDistill: a dual-guided self-distillation approach for carotid plaque analysis Frontiers in Medicine ultrasound video classification carotid plaque recognition self-distillation spatial-temporal attention intra-frame relationship deep learning |
| title | DualDistill: a dual-guided self-distillation approach for carotid plaque analysis |
| title_full | DualDistill: a dual-guided self-distillation approach for carotid plaque analysis |
| title_fullStr | DualDistill: a dual-guided self-distillation approach for carotid plaque analysis |
| title_full_unstemmed | DualDistill: a dual-guided self-distillation approach for carotid plaque analysis |
| title_short | DualDistill: a dual-guided self-distillation approach for carotid plaque analysis |
| title_sort | dualdistill a dual guided self distillation approach for carotid plaque analysis |
| topic | ultrasound video classification carotid plaque recognition self-distillation spatial-temporal attention intra-frame relationship deep learning |
| url | https://www.frontiersin.org/articles/10.3389/fmed.2025.1554578/full |
| work_keys_str_mv | AT xiaomanzhang dualdistilladualguidedselfdistillationapproachforcarotidplaqueanalysis AT jiangxie dualdistilladualguidedselfdistillationapproachforcarotidplaqueanalysis AT haibingchen dualdistilladualguidedselfdistillationapproachforcarotidplaqueanalysis AT haiyawang dualdistilladualguidedselfdistillationapproachforcarotidplaqueanalysis |