Artificial Intelligence and High-Resolution Anoscopy: Automatic Identification of Anal Squamous Cell Carcinoma precursors using a Convolutional Neural Network

DIGESTAID is committed to provide innovative AI solutions to enhance the medical approach to several digestive health problems. Anal squamous cell cancer is a malignancy with increasing incidence. The screening of populations with risk factors for this malignancy is pivotal to decrease its burden. High-resolution anoscopy is the gold standard technique to identify precursor lesions (i.e. high-grade dysplasia, also known as HSIL), therefore enabling their treatment prior to the development of invasive lesions. Nevertheless, the proficient performance of high-resolution anoscopy requires intensive training and has a long learning curve. Thus, the limited number of experts in HRA limits the implementation of effective screening programs based on this technique. 

Our group, in cooperation with Groupe Hospitalier Paris Saint-Joseph, have developed, for the first time, a deep learning algorithm for the automatic detection of precursor lesions of anal squamous cell cancer. Our algorithm detected and differentiated HSIL from normal anal mucosa or non-precursor low-grade dysplasia lesions, with an accuracy of 90%,a sensitivity of 91% and a specificity of 90%. We believe that the results of this international cooperation will have a significant impact in the management of patients undergoing screening.

The full-length article can be found here.

In Press // Tech. Coloproctol. 2022 Aug 20, Online ahead of print // August 2022