Detection and Diagnosis of Congenital Heart Disease from Chest X-Rays with Deep Learning Models

Abstract

Congenital heart disease (CHD) is a leading cause of morbidity and mortality in children, requiring early and accurate diagnosis for effective management. In this study, we employed advanced deep learning models—EfficientNetV2, ResNeSt, and MobileNetV4—to classify CHD using chest X-ray (CXR) images. A dataset comprising 828 images, categorized into normal, atrial septal defect (ASD), ventricular septal defect (VSD), and patent ductus arteriosus (PDA), was utilized. The dataset was split into training, validation, and test sets in a stratified manner to ensure balanced evaluation. EfficientNetV2 achieved the best performance, with an accuracy of 92.5%, precision of 90.9%, sensitivity of 84.1%, and specificity of 88.6%, demonstrating its reliability for CHD diagnosis. ResNeSt closely followed, with an accuracy of 91.7% and precision of 90.4%, while MobileNetV4, though slightly less accurate at 88.6%, offered a lightweight alternative for resource-constrained environments.

References

[1] Hoffman JI, Kaplan S. The incidence of congenital heart disease. Journal of the American college of cardiology. 2002 Jun 19;39(12):1890-900.

[2] Hoffman JI, Kaplan S, Liberthson RR. Prevalence of congenital heart disease. American heart journal. 2004 Mar 1;147(3):425-39.

[3] Sun R, Liu M, Lu L, Zheng Y, Zhang P. Congenital heart disease: causes, diagnosis, symptoms, and treatments. Cell biochemistry and biophysics. 2015 Jul;72:857-60.

[4] Leusveld EM, Kauling RM, Geenen LW, Roos-Hesselink JW. Heart failure in congenital heart disease: management options and clinical challenges. Expert review of cardiovascular therapy. 2020 Aug 2;18(8):503-16.

[5] Afrazeh F. Advances in Imaging Analysis for Understanding Brain Disorders. International Journal of Scientific and Applied Research (IJSAR), eISSN: 2583-0279. 2024 Nov 24;4(8):1-8.

[6] Huda W, Abrahams RB. X-ray-based medical imaging and resolution. American Journal of Roentgenology. 2015 Apr;204(4):W393-7.

[7] Kasban H, El-Bendary MA, Salama DH. A comparative study of medical imaging techniques. International Journal of Information Science and Intelligent System. 2015 Apr;4(2):37-58.

[8] Afrazeh F, Ghasemi Y, Shomalzadeh M, Rostamian S. The Role of Imaging Data from Different Radiologic Modalities During the Previous Global Pandemic. International Journal of Applied Data Science in Engineering and Health. 2024 Jul 27;1(1):9-17.

[9] Taylor AM, Thorne SA, Rubens MB, Jhooti P, Keegan J, Gatehouse PD, Wiesmann F, Grothues F, Somerville J, Pennell DJ. Coronary artery imaging in grown up congenital heart disease: complementary role of magnetic resonance and x-ray coronary angiography. Circulation. 2000 Apr 11;101(14):1670-8.

[10] Perloff JK. The Chest X‐Ray in Congenital Heart Disease. Adult Congenital Heart Disease. 2009 Apr 17:208-20.

[11] Orouskhani M, Zhu C, Rostamian S, Zadeh FS, Shafiei M, Orouskhani Y. Alzheimer's disease detection from structural MRI using conditional deep triplet network. Neuroscience Informatics. 2022 Dec 1;2(4):100066.

[12] Minoo S, Ghasemi F. A Narrative Review: Dental Radiology with Deep Learning. International Research in Medical and Health Sciences. 2024 Oct 25;7(5):23-36.

[13] El Asnaoui K. Design ensemble deep learning model for pneumonia disease classification. International Journal of Multimedia Information Retrieval. 2021 Mar;10(1):55-68.

[14] Abbasi H. Transfer Learning and Advanced CNN Models for Detecting Brain Tumors using MRI. International Journal of Scientific and Applied Research (IJSAR), eISSN: 2583-0279. 2024 Dec 13;4(9):92-103.

[15] Norouzi F, Machado BL, Nematzadeh S. Schizophrenia Diagnosis and Prediction with Machine Learning Models. International Journal of Scientific and Applied Research (IJSAR), eISSN: 2583-0279. 2024 Dec 13;4(9):113-22.

[16] Mzoughi O, Yahiaoui I. Deep learning-based segmentation for disease identification. Ecological Informatics. 2023 Jul 1;75:102000.

[17] Livne M, Rieger J, Aydin OU, Taha AA, Akay EM, Kossen T, Sobesky J, Kelleher JD, Hildebrand K, Frey D, Madai VI. A U-Net deep learning framework for high performance vessel segmentation in patients with cerebrovascular disease. Frontiers in neuroscience. 2019 Feb 28;13:97.

[18] Abbasi H, Afrazeh F, Ghasemi Y, Ghasemi F. A Shallow Review of Artificial Intelligence Applications in Brain Disease: Stroke, Alzheimer's, and Aneurysm. International Journal of Applied Data Science in Engineering and Health. 2024 Oct 5;1(2):32-43.

[19] Pang G, Shen C, Cao L, Hengel AV. Deep learning for anomaly detection: A review. ACM computing surveys (CSUR). 2021 Mar 5;54(2):1-38.

[20] Mahmoudiandehkordi S, Yeganegi M. Transforming Gynecology with Artificial Intelligence: Advances in Clinical Practice. International Journal of Scientific and Applied Research (IJSAR), eISSN: 2583-0279. 2024 Dec 13;4(9):104-12.

[21] https://kaggle.com/competitions/congenital-heart-disease, 2023. Kaggle.

[22] Zhang H, Wu C, Zhang Z, Zhu Y, Lin H, Zhang Z, Sun Y, He T, Mueller J, Manmatha R, Li M. Resnest: Split-attention networks. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition 2022 (pp. 2736-2746).

[23] Howard A, Sandler M, Chu G, Chen LC, Chen B, Tan M, Wang W, Zhu Y, Pang R, Vasudevan V, Le QV. Searching for mobilenetv3. InProceedings of the IEEE/CVF international conference on computer vision 2019 (pp. 1314-1324).

[24] Tan M, Le Q. Efficientnetv2: Smaller models and faster training. InInternational conference on machine learning 2021 Jul 1 (pp. 10096-10106). PMLR.