Table 1 Representative literature on the screening of congenital heart disease

Plana et al. (2018) [4]

Meta-analysis (n = 457,202)

Pulse oximetry

Sensitivity: 76.3%

Specificity: 99.9%

Detected only critical cyanotic heart disease

Lv et al. (2021) [9]

1,362 CHD children requiring surgery

AI based heart sound analysis

Accuracy: 98%

Sensitivity: 91%

Specificity: 97%

Unable to identify CHDs without significant murmur

Xu et al. (2022) [10]

Children aged 2 days to 12 years (408 CHD, 553 controls)

AI-based heart sound analysis

Accuracy: 95%

Sensitivity: 94%

Specificity: 96%

Detected CHD with heart murmurs, while overlooking CHDs without murmur

Liu et al. (2022) [11]

Children (475 CHD, 409 controls)

AI-based heart sound analysis

Accuracy: 83%

ASD detection accuracy: 65%

Poor performance in ASD detection, due to variations in ECG caused by factors, such as right heart pressure and defect size

Alkahtani et al. (2024) [7]

583 PCG from local database and 23 ECG from public database

AI-based heart sound analysis

Accuracy: 98.6%

Sensitivity: 99.0%

Specificity: 98.0%

Binary classifications (Normal vs. abnormal). Patient age not specified

Du et al. (2020) [12]

68,969 ECGs (58,624 Non-CHD and 10,345 CHD)

AI-based ECG analysis

Sensitivity: 74.7%

Specificity: 94.1%

Demographic of CHD patients not specified. Patient age not specified

Mori et al. (2021) [8]

Children aged 6–18 (364 ASD, 828 normal)

Deep learning-based ECG analysis

Accuracy: 0.89

Specificity: 0.96

F1 Score: 0.81

Focus on school-aged children

Included only ASD patients

Liu et al. (2023) [13]

Adults (1,196 ASD, 21,430 controls)

AI-based ECG analysis

Accuracy: 0.86

Specificity: 0.87

AUC: 0.88

Focus on adults with no data from infants or young children. Did not address hemodynamics