Researchers at Asan Medical Center (AMC) said Tuesday that they have confirmed the excellent efficacy of a deep learning-based artificial intelligence algorithm in detecting about 43 types of skin cancer and diseases.

According to AMC, the AI model has similar levels to dermatologists in diagnosing skin diseases. It showed 66.9 percent sensitivity, 87.4 percent specificity, equivalent to 65.8 and 85.7 percent made by dermatologists, respectively.

Professor Chang Sung-eun of the Department of Dermatology at Asan Medical Center said she and her research team have verified the effectiveness of a deep learning-based artificial intelligence (AI) algorithm for detecting skin diseases. (AMC)
Professor Chang Sung-eun of the Department of Dermatology at Asan Medical Center said she and her research team have verified the effectiveness of a deep learning-based artificial intelligence (AI) algorithm for detecting skin diseases. (AMC)

The technology provides a clue in screening skin cancers with only a picture of the area affected by the lesion.

Slowly growing skin tumor is often misdiagnosed as benign, leading to delayed diagnosis. When patients miss the appropriate treatment period, cancer can spread to other organs such as the liver or lungs and threaten their lives. Most of the existing diagnostic studies used AI to distinguish whether or not the cancer is benign or malignant just by reading the image.

However, the images used in this study were verified with all types of malignant and benign skin cancer data that Severance Hospital has tested over the past decade to train the AI model.

In this study, doctors who diagnosed the patients showed higher accuracy than the algorithm with a sensitivity of 70.2 percent and specificity of 95.6 percent.

“Malignant melanoma is a fatal disease that has shown a five-year survival rate below 20 percent if the tumor spreads to other organs including lung and liver,” AMC dermatologist Chang Sung-eun said. “We expect this AI algorithm can help in diagnosing and treating patients with skin cancers via periodic self-examination.”

The study results were published in the international journal, Public Library of Science Medicine.

Copyright © KBR Unauthorized reproduction, redistribution prohibited