Key result
RCS_UNet outperforms other models in carotid ultrasound plaque segmentation with a ~0.82 Dice score.
Why the study?
Carotid artery ultrasound plaque segmentation is important for disease diagnosis but challenging due to image noise, low contrast, and variable plaque shapes.
Does the RCS_UNet model accurately segment plaques in carotid artery ultrasound images compared to other advanced models?
Does the RCS_UNet model accurately segment plaques in carotid artery ultrasound images compared to other advanced models?
The proposed RCS_UNet model demonstrates high accuracy and generalization in segmenting carotid artery plaques from ultrasound images, offering potential clinical value for disease diagnosis.
May support automated plaque analysis in research; leaves open clinical validation before diagnostic use.
The plaque segmentation of ultrasound images is very important in the disease diagnosis of carotid artery, and it is hard to be realized due to the image noise, low contrast, and plaque shapes, etc. This paper proposed an improved network model called RCS_UNet based on the classical UNet architecture. The residual convolution, CBAM and the multi-scale supervision modules improve the gradient vanishing, the model attention on key features and the performance of decoder part, respectively. RCS_UNet was trained and tested using a dataset containing 443 longitudinal carotid artery ultrasound images. RCS_UNet can accurately segment plaques in carotid artery ultrasound images, with DSC of 0.819±0.091, IoU of 0.703±0.123, recall of 0.845±0.100, and precision of 0.822±0.150, which was better than that of other advanced models. The trained model was used to segment plaques in carotid artery ultrasound images obtained from another ultrasound device, DSC was 0.832±0.075, IoU was 0.719±0.105, recall was 0.792±0.136, precision was 0.900±0.065, which proved that the model had a certain degree of generalization. RCS_UNet possesses excellent encoding and decoding capabilities, it can effectively extract plaque features in carotid ultrasound images for boundary segmentation, which holds certain clinical application value.
No takes yet. Share an insight, caveat, or question.
Hao et al. (2024) studied Carotid artery plaque (n=443). RCS_UNet model vs. Other advanced models was evaluated on Plaque segmentation accuracy (Dice Similarity Coefficient). The RCS_UNet model accurately segmented plaques in carotid artery ultrasound images with a Dice Similarity Coefficient of 0.819±0.091, outperforming other advanced models.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: