TABLE III.
Comparison of Current Methods & Previous Methods Reported in Literature
| Method | Tissue Classes a | Gold Standard | Validation Dataset | Reported Accuracy |
|---|---|---|---|---|
| Zhang et al. [19] | DC, NC, & FT | Expert annotations | 12 IVUS images | Overall: 89.9% |
| Brunenberg et al. [20] | DC, NC, & FT | Expert annotations | 30 IVUS images | Overall: 75.82% |
| Athanasiou et al. [21] | DC, NC, & FT | Experts annotations | 40 IVUS images | Overall: 86.17% |
| Taki et al. (2010) [22] | DC, NC, & FT-FFT | VH-IVUS | 500 VH-IVUS frames | DC: 79%, NC: 52%, & FT-FFT: 81% |
| Taki et al. (2013) [23] | DC, NC, & FT-FFT | VH-IVUS | 500 VH-IVUS frames | DC: 84.87%, NC: 80.57%, FT: 77.4%, & FFT: 63.47% |
| Athanasiou et al. [24] | DC, NC, FT, FFT, & M | VH-IVUS | 300 VH-IVUS frames | DC: 84.87%, NC: 80.57%, FT: 77.4%, & FFT: 63.47% |
| Hwang et al. [25] | DC, NC, FT, & FFT | VH-IVUS | 252 VH-IVUS frames | Overall: 82.8% |
| Kim et al. [26] | DC, NC, FT, & FFT | VH-IVUS | 252 VH-IVUS frames | Overall: 85.1% |
| Naïve Deep Learning Approach | DC, NC, FT, FFT, & M | VH-IVUS | 200 VH-IVUS frames | DC: 98.7%, NC: 87.3%, FT: 87.5%, FFT: 89.5%, & M: 75.9% |
| Domain Enriched Deep Learning Approach | DC, NC, FT, FFT, & M | VH-IVUS | 200 VH-IVUS frames | DC: 98.5%, NC: 88.6%, FT: 91.1%, FFT: 90.0%, & M: 99.4% |
| VH-IVUS (Validation by Nair et al. [27]) | DC, NC, FT, FFT, & M | Histology | 889 histological regions | DC: 96.7%, NC: 95.8%, FT: 93.5%, & FFT: 94.1% (M not reported) |
a
DC: Dense Calcium; NC: Necrotic Core; FT: Fibrous Tissue; FFT: Fibro-Fatty Tissue; M: Media/Non-Pathological