Related Experiment Video
Updated: May 8, 2026
Combined Near-infrared Fluorescent Imaging and Micro-computed Tomography for Directly Visualizing Cerebral Thromboemboli
Published on: September 25, 2016
DCCAT: Dual-Coordinate Cross-Attention Transformer for thrombus segmentation on coronary OCT
Miao Chu1, Giovanni Luigi De Maria2, Ruobing Dai3
1Biomedical Instrument Institute, School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, China; Oxford Heart Centre, Oxford University Hospitals NHS Trust, UK; Division of Cardiovascular Medicine, Radcliffe Department of Medicine, University of Oxford, UK.
Insights
A new dual-coordinate cross-attention transformer network (DCCAT) accurately segments thrombus in optical coherence tomography (OCT) images. This automated method improves accuracy for acute coronary syndromes (ACS) diagnosis and treatment planning.
Area of Science:
- Cardiovascular Imaging
- Artificial Intelligence in Medicine
- Medical Image Analysis
Background:
- Acute coronary syndromes (ACS) are a major global health concern, with thrombus formation being central to their pathology.
- Accurate thrombus burden evaluation is crucial for effective treatment and prognosis in ACS patients.
- Coronary optical coherence tomography (OCT) offers detailed in-vivo visualization of thrombus, but automated quantification remains a challenge due to thrombus variability and limited data.
Purpose of the Study:
- To develop and validate a novel deep learning model for automatic thrombus segmentation in coronary OCT images.
- To address the challenges of thrombus variability and small datasets through an innovative network architecture.
Main Methods:
- A novel dual-coordinate cross-attention transformer network (DCCAT) was proposed, integrating features from both Cartesian and polar coordinates.
- The DCCAT model utilizes a multi-head cross-attention mechanism for feature fusion and hierarchical stacking with convolutional layers.
- The model was trained on 5,649 OCT frames from 339 patients and validated on an independent dataset of 548 frames from 52 patients.
Main Results:
- DCCAT achieved a Dice similarity score (DSC) of 0.706 for thrombus segmentation, outperforming existing CNN-based (0.656) and Transformer-based (0.584) models.
- The inclusion of polar image features enhanced model robustness against geometrical transformations.
- The model demonstrated high data efficiency, achieving competitive performance with only 10% of the training data.
Conclusions:
- The proposed DCCAT model represents a significant advancement in automated thrombus segmentation for coronary OCT.
- The dual-coordinate cross-attention approach effectively handles thrombus variability and improves segmentation accuracy.
- This method holds promise for enhancing clinical decision-making in the management of acute coronary syndromes.
Abstract:
Acute coronary syndromes (ACS) are one of the leading causes of mortality worldwide, with atherosclerotic plaque rupture and subsequent thrombus formation as the main underlying substrate. Thrombus burden evaluation is important for tailoring treatment therapy and predicting prognosis. Coronary optical coherence tomography (OCT) enables in-vivo visualization of thrombus that cannot otherwise be achieved by other image modalities. However, automatic quantification of thrombus on OCT has not been implemented. The main challenges are due to the variation in location, size and irregularities of thrombus in addition to the small data set. In this paper, we propose a novel dual-coordinate cross-attention transformer network, termed DCCAT, to overcome the above challenges and achieve the first automatic segmentation of thrombus on OCT. Imaging features from both Cartesian and polar coordinates are encoded and fused based on long-range correspondence via multi-head cross-attention mechanism. The dual-coordinate cross-attention block is hierarchically stacked amid convolutional layers at multiple levels, allowing comprehensive feature enhancement. The model was developed based on 5,649 OCT frames from 339 patients and tested using independent external OCT data from 548 frames of 52 patients. DCCAT achieved Dice similarity score (DSC) of 0.706 in segmenting thrombus, which is significantly higher than the CNN-based (0.656) and Transformer-based (0.584) models. We prove that the additional input of polar image not only leverages discriminative features from another coordinate but also improves model robustness for geometrical transformation.Experiment results show that DCCAT achieves competitive performance with only 10% of the total data, highlighting its data efficiency. The proposed dual-coordinate cross-attention design can be easily integrated into other developed Transformer models to boost performance.
More Related Videos
12:54Simultaneous Brightfield, Fluorescence, and Optical Coherence Tomographic Imaging of Contracting Cardiac Trabeculae Ex Vivo
Published on: October 2, 2021
13:07Optical Coherence Tomography Based Biomechanical Fluid-Structure Interaction Analysis of Coronary Atherosclerosis Progression
Published on: January 15, 2022
