Articles Vol. 67 No. CĐ2-NCKH 05/03/2026

COMPARING THE DIAGNOSTIC SUPPORT VALUE OF DERMOSCOPY AND ARTIFICIAL INTELLIGENCE IN BASAL CELL CARCINOMA

Phan Nu Thuc Hien1,2,3,4, Nguyen Huu Sau5,6, Nguyen Long Giang7,8
1 Hanoi Medical University
2 Bach Mai Hospital
3 Trường Đại học Y Hà Nội
4 Bệnh viện Bạch Mai
5 Central Dermatology Hospital
6 Bệnh viện Da liễu Trung ương
7 Institute of Information Technology, Vietnam Academy of Science and Technology
8 Viện Công nghệ thông tin, Viện Hàn Lâm Khoa học và Công nghệ Việt Nam
DOI: 10.52163/yhc.v67iCD2.4433
15 Views
14 Downloads
Abstract

Objective: To compare the diagnostic support value of dermoscopy and artificial intelligence (AI) using a convolutional neural network (StackNet model) in the diagnosis of basal cell carcinoma (BCC).

Materials and Methods: A prospective cross-sectional descriptive study was conducted on 110 patients presenting with clinically suspected BCC lesions at the National Hospital of Dermatology and Venereology.

Results: Dermoscopy demonstrated a sensitivity of 91.4% and a specificity of 73.7% for the diagnosis of BCC, improving diagnostic accuracy compared with clinical examination alone. Among 124 clinically suspected lesions, dermoscopy correctly identified 96 of 105 histopathologically confirmed BCC lesions. The AI model (StackNet) achieved a sensitivity of 94.3% and a specificity of 78.9%, correctly identifying 99 of 105 true BCC lesions, indicating superior diagnostic support compared with visual clinical assessment. The diagnostic performance of dermoscopy (AUROC = 0.826) and AI (AUROC = 0.866) was rated as good. The mean difference in AUROC between the two methods was 0.04; DeLong’s test showed no statistically significant difference between the methods (p > 0.05).

Conclusion: Both dermoscopy and artificial intelligence significantly enhance the diagnostic accuracy of basal cell carcinoma compared with naked-eye clinical examination. The sensitivity and specificity of AI in diagnosing BCC were high. The diagnostic support value of dermoscopy and AI was comparable and rated as good, with no statistically significant difference between the two approaches.

References
[1]
Geller AC, Annas GD. Epidemiology of melanoma and nonmelanoma skin cancer. Seminars in Oncology Nursing. 2003;19(1):2-11. doi:10.1053/sonu.2003.50000 Google Scholar
[2]
Rubin AI, Chen EH, Ratner D. Basal-Cell Carcinoma. New England Journal of Medicine. 2005; 353(21):2262-2269. doi:10.1056/NEJMra044151 Google Scholar
[3]
Nguyễn Văn Thường. Ung thư biểu mô tế bào đáy. In: Hình Ảnh Lâm Sàng, Chẩn Đoán và Điều Trị Trong Chuyên Ngành Da Liễu. Vol 2. ; 2019:241-250 Google Scholar
[4]
Phạm Thị Minh Phương, Đặng Thị Lương, Nguyễn Văn Thường. Độ nhạy, độ đặc hiệu của Dermoscopy trong chẩn đoán ung thư biểu mô tế bào đáy tại bệnh viện Da liễu Trung ương. Tạp chí Y học Việt Nam. 2020:250-253 Google Scholar
[5]
Hekler A, Utikal JS, Enk AH, et al. Deep learning outperformed 11 pathologists in the classification of histopathological melanoma images. European Journal of Cancer. 2019; 118:91-96. doi:10.1016/j.ejca.2019.06.012 Google Scholar
[6]
Jiang F, Jiang Y, Zhi H, et al. Artificial intelligence in healthcare: past, present and future. Stroke Vasc Neurol. 2017;2(4):230-243. doi:10.1136/svn-2017-000101 Google Scholar
[7]
Chan G, Ho H. A study of dermoscopic features of pigmented basal cell carcinoma in Hong Kong Chinese. Hong Kong Journal of Dermatology and Venereology. 2008; 16(4):189-196 Google Scholar
[8]
Reiter O, Mimouni I, Gdalevich M, et al. The diagnostic accuracy of dermoscopy for basal cell carcinoma: A systematic review and meta-analysis. Journal of the American Academy of Dermatology. 2019;80(5):1380-1388. doi:10.1016/j.jaad.2018.12.026 Google Scholar
[9]
Esteva A, Kuprel B, Novoa RA, et al. Dermatologist-level classification of skin cancer with deep neural networks. Nature. 2017;542(7639):115-118. doi:10.1038/nature21056 Google Scholar
[10]
Assessment of Accuracy of an Artificial Intelligence Algorithm to Detect Melanoma in Images of Skin Lesions | Dermatology | JAMA Network Open | JAMA Network. Accessed July 11, 2022. https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2752995 Google Scholar