TABLE IRELATED WORK SUMMARYNumber o

TABLE I
RELATED WORK SUMMARY
Number of
Genetic Disorders
Number of Training
Samples (Syndromic) Evaluation Method Accuracy
(top-1-accuracy)
Problem 1: Single syndrome vs. other population
Saraydemir et al. [22] 1 15 3,4-Fold Cross-Validation 97.34%
Burccin et al. [23] 1 10 51 images in a test set 95.30%
Zhao et al. [24] 1 50 Leave-One-Out 96.70%
Basel-Vanagaite et al. [4] 1 134 7 images in test set 94%
Kruszka et al. [25] 1 129 Leave-One-Out 94.30%
Kruszka et al. [26] 1 156 Leave-One-Out 94.90%
Liehr et al. [18] 2 173 10-Fold Cross-Validation 100%
Shukla et al. [21] 6 1126 5-Fold Cross-Validation 94.93 (mAP)1
Ferry et al. [3] 8 1363 Leave-One-Out 94.90%
Problem 2: Syndromic vs. normal
Zhao et al. [24] 14 24 Leave-One-Out 97%
Cerrolaza et al. [28] 15 73 Leave-One-Out 95%
Shukla et al. [21] 6 1126 5-Fold Cross-Validation 98.80%
Problem 3: Multiple syndromes classification
Loos et al. [30] 5 55 Leave-One-Out 76%
Kuru et al. [29] 15 92 Leave-One-Out 53%
Boehringer et al. [32] 10 147 10-Fold Cross-Validation 75.70%
Boehringer et al. [31] 14 202 91 images in a test set 21%
Ferry et al. [3] 8 1363 Leave-One-Out 75.60%2
Shukla et al. [21] 6 1126 5-Fold Cross-Validation 48%2