Application of the CenterNet Network for Acne Detection

According to a study published in Frontiers in Medicine, a multi-task acne detection method utilizing a CenterNet-based training paradigm represents advancements in acne detection, and offers a comprehensive tool for classification, localization, counting, and precise segmentation.

Researchers aimed to enhance acne detection by assessing image quality, accurately segmenting lesions, and precisely grading acne types. Their method incorporated multi-task learning, enabling simultaneous detection of various acne types, including noninflammatory acne, papules, pustules, nodules, cysts, and post-acne scars. By leveraging smartphone-captured images, the system offers a convenient and accessible approach to acne detection.

In clinical diagnostics, the multi-task framework achieved promising results. It demonstrated an 83% accuracy in lesion categorization, surpassing existing models by 12%, and a 76% precision in lesion stratification, outperforming dermatologists by 16%. These findings highlight the effectiveness of this approach in accurately identifying and categorizing acne lesions.

“It not only enhances the accuracy of remote acne lesion identification by doctors but also clarifies grading logic and criteria, facilitating easier grading judgments,” the authors concluded.

Reference
Zhang D, Li H, Shi J, et al. Advancements in acne detection: application of the CenterNet network in smart dermatology. Front Med (Lausanne). Published online March 25, 2024. doi:10.3389/fmed.2024.1344314