Table 2 Summary of tests to evaluate the segmentation performance of deep learning networks.

From: Deep learning segmentation of major vessels in X-ray coronary angiography

Test name

Dataset

Major Vessel

Evaluation area

Training

Validation

Test

Hyperparameter

Fold 1–3

Fold 4

Fold 5

RCA + LAD + LCX

512 × 512

Combined dataset

Cross validation (Fold ratio = 3:1:1)

RCA + LAD + LCX

512 × 512

128 × 128†

Separate dataset

Cross validation (Fold ratio = 3:1:1)

RCA

512 × 512

LAD

LCX

Data size

Fold 1–4* (Image ratio = 3:1)

Fold 5

RCA + LAD + LCX

512 × 512

External validation

Fold 1–3

Fold 4

External

RCA + LAD + LCX

512 × 512

  1. RCA, right coronary artery; LAD, left anterior descending artery; LCX, left circumflex artery; †The center of the evaluation area was set at the midpoint of the most narrowed location in a major vessel; *Dataset size was gradually increased from fold 1 with increment of a fold.
Back to article page