Table 3.
Comparison of the proposed method with related work.
| Author | Year | Method | Dataset | # | Automation | Precision (%) | Run time (s) |
|---|---|---|---|---|---|---|---|
| Oliveira et al. [7] | 2011 | RG | Sliver07 | 20 | Auto | — | — |
| Luu et al. [4] | 2015 | RG | Clinical CTA | 51 | Auto | ACC = 86.2; SEN = 85.1; SPE = 92.3 | — |
| Esneault et al. [10] | 2010 | GC | Clinical CTA | 1 | Auto | — | 10–100 |
| Zeng et al. [12] | 2017 | GC | Clinical CTA | 6 | Auto | ACC = 97.7; SEN = 79.8; SPE = 98.6 | 390 |
| Sangsefidi et al. [11] | 2018 | GC | 3Dircadb/Clinical CTA | 7 | Auto | DICE = 74.0 | 560 |
| Shang et al. [9] | 2011 | LS | Clinical CTA | 20 | Auto | SEN∗ = 91.0 | 480 |
| Ahmadi et al. [28] | 2016 | FCC | Sliver07 | 20 | Auto | ACC = 91.0; SEN = 94.1; SPE = 83.6 | 27.1 |
| Zeng et al. [13] | 2016 | ML | Clinical CTA | 6 | Auto | ACC = 98.1; SEN = 74.2; SPE = 99.3 | 0.05–0.1 |
| Guo et al. [15] | 2015 | FC | Clinical CTA | 4 | Semi | — | 112.5 |
| Wang et al. [16] | 2016 | FC | Clinical CTA | 3 | Semi | — | 22 |
| Huang et al. [14] | 2018 | DL | 3Dircadb | 20 | Auto | ACC = 97.1; SEN = 74.3; SPE = 98.3; DICE = 67.5 | 230 |
| Ours | 2018 | IFC | 3Dircadb Sliver07 |
20 20 |
Auto Auto |
ACC = 96.4; SEN = 73.7; SPE = 97.4; DICE = 67.3 ACC = 96.8; SEN = 84.4; SPE = 97.6; DICE = 71.4 |
200 210 |
∗Evaluation by the number of vascular nodes; CTA = computed tomography angiography; RG = region growing; GC = graph cuts; LS = level set; FCC = fuzzy C-means clustering; ML = machine learning; FC = fuzzy connectedness; DL = deep learning; IFC = improved fuzzy connectedness; ACC = accuracy; SEN = sensitivity; SPE = specificity; DICE = Dice coefficient; Auto = automatic; Semi = semiautomatic.