Dados do Resumo

Título

Efficiency of an Artificial Intelligence Platform in the Screening and Prioritization of Suspected Skin Cancer Lesions: A Multicenter Experience in Brazil

Introdução

Skin cancer is the most common neoplasm in Brazil and worldwide, with early detection being critical for better prognoses. Large-scale screening is challenging, primarily due to the need for specialized resources. The use of technologies such as artificial intelligence (AI) has proven to be a promising tool in assisting with the triage and early identification of suspicious lesions, contributing to more efficient case management.

Objetivo

This study aims to report the experience of using a digital platform equipped with AI for the management and screening of suspected skin cancer lesions, facilitating the diagnostic process and prioritizing cases.

Métodos

The platform includes a registration screen for entering sociodemographic data and another for inputting clinical information, including images of lesions taken with a dermatoscope. The integrated AI analyzes the images and suggests percentage-based diagnoses, according to the visual patterns of the lesions. Physicians access the triaged cases through a patient and lesion list. When a lesion is identified as suspicious, the physician assigns its priority (high, medium, or low) for follow-up and treatment. Screening took place in seven Brazilian states, and results obtained up to August 2024 were analyzed for this study. In addition, the study received approval from the Research Ethics Committee of the Hospital de Amor (approval number 6.891.260).

Resultados

A total of 11,662 screenings were performed, evaluating 22,611 lesions. Of these, 64.03% were not suspicious for cancer. Lesions suspected of squamous cell carcinoma accounted for 3.2%, basal cell carcinoma for 7.95%, actinic keratosis for 8.84%, melanoma for 1.81%, while 4.20% were classified as suspicious without a conclusive diagnosis, and 9.96% of the photos did not allow for proper evaluation. After physician analysis, 545 cases (4.45%) were identified as high-risk for skin cancer, 1,450 (11.84%) as medium-risk, and 10,252 (83.71%) as low-risk. The platform enabled the rapid and efficient identification of suspicious lesions, facilitating the prioritization of the most serious cases.

Conclusões

The use of the digital platform with AI proved to be efficient in managing the screening of suspected skin cancer lesions, allowing for the proper classification and prioritization of cases. With more than 22,000 lesions screened, the tool excelled in identifying high- and medium-risk lesions, in addition to providing decisive support to physicians. This highlights the potential of new technologies to assist in large-scale skin cancer screening and management.

Palavras Chave

Screening; Skin cancer; innovation