The National Engineering Laboratory of Visual Information Processing and Applications and the Institute of Artificial Intelligence and Robotics, Xi’an Jiaotong University, Xi’an 710049, China
Peiwen Shi, Jingmin Xin, and Nanning Zheng, "A-line-based thin-cap fibroatheroma detection with multi-view IVOCT images using multi-task learning and contrastive learning," J. Opt. Soc. Am. A 39, 2298-2306 (2022)
Automatic detection of thin-cap fibroatheroma (TCFA) is essential to prevent acute coronary syndrome. Hence, in this paper, a method is proposed to detect TCFAs by directly classifying each A-line using multi-view intravascular optical coherence tomography (IVOCT) images. To solve the problem of false positives, a multi-input–output network was developed to implement image-level classification and A-line-based classification at the same time, and a contrastive consistency term was designed to ensure consistency between two tasks. In addition, to learn spatial and global information and obtain the complete extent of TCFAs, an architecture and a regional connectivity constraint term are proposed to classify each A-line of IVOCT images. Experimental results obtained on the 2017 China Computer Vision Conference IVOCT dataset show that the proposed method achieved state-of-art performance with a total score of $88.7 \pm 0.88\%$, overlap rate of $88.64 \pm 0.26\%$, precision rate of $84.34 \pm 0.86\%$, and recall rate of $93.67 \pm 2.29\%$.
Data underlying the results presented in this paper were publicly available for the 2017 China Computer Vision Conference Vulnerable Plaque (CCCV-VP) detection challenge, and are not publicly available at this time but may be obtained from the sponsor upon reasonable request.
Cited By
You do not have subscription access to this journal. Cited by links are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
You do not have subscription access to this journal. Figure files are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
You do not have subscription access to this journal. Article tables are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
You do not have subscription access to this journal. Equations are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
Comparison of Image-Level Classification Results Using Different Views of IVOCT Images with and without the Preprocessing Step, Evaluated by Precison, Recall, and -Score
View
Preprocessing
Precision
Recall
-Score
Cartesian
Without
0.8564
0.8434
0.8498
With
0.909
0.87
0.889
Polar
Without
0.85
0.8585
0.8542
With
0.8651
0.9393
0.9007
Multi-views
Without
0.8808
0.8585
0.8695
With
0.9195
0.9242
0.9219
Table 3.
Comparison of A-Line-Based Classification Results Using Polar-IVOCT Images with and without the Preprocessing Step, Evaluated by Quality Score (), Overlap Rate (), Precision (), and Recall ()
Preprocessing
Without
With
Table 4.
Comparison of A-Line-Based Classification Results Using Different Encoder–Decoder Structure with and without A-Line Classifier, Evaluated by Quality Score (), Overlap Rate (), Precision (), and Recall ()
Encoder–Decoder
A-Line Classifier
Asymmetric
Without
With
Symmetric
Without
With
Table 5.
Comparison of TCFA Detection Results Using Different Ensemble Methods, Evaluated by Quality Score (), Overlap Rate (), Precision (), and Recall ()
Ensemble Methods
Precision
Recall
Baseline
Polar domain fusion
Result-level fusion
MM-net
Table 6.
Comparison of TCFA Detection Using MM-Net with Different Constraints, Evaluated by Quality Score (), Overlap Rate (), Precision (), and Recall ()
Baseline
RCC
CC
Precision
Recall
✓
✗
✗
✓
✓
✗
✓
✗
✓
✓
✓
✓
Tables (6)
Table 1.
Performance Comparisons of Different TCFA Detection Methods, Evaluated by Quality Score (), Overlap Rate (), Precision (), and Recall ()a
Comparison of Image-Level Classification Results Using Different Views of IVOCT Images with and without the Preprocessing Step, Evaluated by Precison, Recall, and -Score
View
Preprocessing
Precision
Recall
-Score
Cartesian
Without
0.8564
0.8434
0.8498
With
0.909
0.87
0.889
Polar
Without
0.85
0.8585
0.8542
With
0.8651
0.9393
0.9007
Multi-views
Without
0.8808
0.8585
0.8695
With
0.9195
0.9242
0.9219
Table 3.
Comparison of A-Line-Based Classification Results Using Polar-IVOCT Images with and without the Preprocessing Step, Evaluated by Quality Score (), Overlap Rate (), Precision (), and Recall ()
Preprocessing
Without
With
Table 4.
Comparison of A-Line-Based Classification Results Using Different Encoder–Decoder Structure with and without A-Line Classifier, Evaluated by Quality Score (), Overlap Rate (), Precision (), and Recall ()
Encoder–Decoder
A-Line Classifier
Asymmetric
Without
With
Symmetric
Without
With
Table 5.
Comparison of TCFA Detection Results Using Different Ensemble Methods, Evaluated by Quality Score (), Overlap Rate (), Precision (), and Recall ()
Ensemble Methods
Precision
Recall
Baseline
Polar domain fusion
Result-level fusion
MM-net
Table 6.
Comparison of TCFA Detection Using MM-Net with Different Constraints, Evaluated by Quality Score (), Overlap Rate (), Precision (), and Recall ()