The cascaded deep learning model using ResNet-based classifier with CAM supervision and CAM-guided MedSAM segmentation improved plaque classification F1-score to 96.7% and segmentation Dice coefficient to 86.6%, outperforming existing methods by at least 3.2% and 3.6%, respectively.
Does a cascaded deep learning framework integrating Masked-ResNet-DS and MedSAM improve the classification and segmentation of carotid plaques from 2D ultrasound images compared to standard models?
A novel cascaded deep learning framework combining Masked-ResNet-DS and MedSAM significantly improves the accuracy of carotid plaque classification and segmentation from ultrasound images.
Effect estimate: Mean F1-score of classification 96.7% (Masked-ResNet-DS) vs 92.1% (best existing MSP-VGG); Dice similarity coefficient (DSC) 86.6% (CAM-MedSAM) vs 83.0% (best baseline nnU-Net)
Absolute Event Rate: 96.7% vs 92.1%
p-value: p=2.77e-83 to 2.58e-19 for DSC improvement statistical significance; p = 2.95e-10 for CAM-MedSAM vs Base-MedSAM
Carotid plaque classification based on ultrasound echogenicity and quantification of plaque burden are crucial in stroke risk assessment. In this work, we propose a framework that leverages the synergy between classification and segmentation by sharing plaque location information to enhance the performance of both tasks. Our cascaded framework integrates a ResNet-based classifier (Masked-ResNet-DS) with MedSAM, a medically adapted version of the Segment Anything Model for joint classification and segmentation of carotid plaques from 2D ultrasound images. Ground truth boundaries are used to guide region-specific feature pooling in the classifier, helping it focus on plaques during training. Since ground truth boundaries are unavailable at inference, we introduce a two-iteration strategy: the first generates a class activation map (CAM), which is then used for focused pooling in the second iteration to predict plaque type. The CAM is also used as a prompt to guide MedSAM for segmentation. To ensure accurate localization, the CAM is supervised during training using a Dice loss against the segmentation ground truth. Masked-ResNet-DS achieves a mean F1-score of 96.7% in plaque classification, at least 3.2% higher than competing methods. Ablation studies confirm that ground truth-based pooling and CAM supervision both improve classification. CAM-guided MedSAM achieves a Dice similarity coefficient (DSC) of 86.6%, outperforming U-Net and nnU-Net by 5.9% and 3.6%, respectively. In addition, CAM prompts improve MedSAM’s DSC by 2.2%. By sharing plaque location between classification and segmentation, the proposed method improves both tasks and provides a more accurate tool for stroke risk stratification.
Ren et al. (2026) studied Patients with carotid artery plaques undergoing 2D carotid ultrasound imaging for plaque echogenicity classification and plaque segmentation (n=925). Cascaded deep learning model combining ResNet-based classifier (Masked-ResNet-DS) and MedSAM segmentation using Class Activation Map guidance vs. Existing plaque classification and segmentation methods including VGG16, SPP-VGG, MSP-VGG, MB-DCNN, RCCM-Net, U-Net, nnU-Net, Base-MedSAM was evaluated on Plaque echogenicity classification accuracy and plaque segmentation quality (Dice similarity coefficient) on 2D carotid ultrasound images (Mean F1-score of classification 96.7% (Masked-ResNet-DS) vs 92.1% (best existing MSP-VGG); Dice similarity coefficient (DSC) 86.6% (CAM-MedSAM) vs 83.0% (best baseline nnU-Net), p=2.77e-83 to 2.58e-19 for DSC improvement statistical significance; p = 2.95e-10 for CAM-MedSAM vs Base-MedSAM). The cascaded deep learning model using ResNet-based classifier with CAM supervision and CAM-guided MedSAM segmentation improved plaque classification F1-score to 96.7% and segmentation Dice coefficient to 86.6%, outperforming existing methods by at least 3.2% and 3.6%, respectively.