Automated In Vivo High-Resolution Imaging to Detect Human Papillomavirus-Associated Anal Precancer in Persons Living...

Area of Science:

• Gastroenterology and Hepatology
• Oncology
• Medical Imaging

Background:

• Anal cancer screening effectiveness is limited by a shortage of trained professionals and high patient dropout rates.
• Early detection of anal intraepithelial neoplasia grade 2 or more severe (AIN 2+) is crucial for preventing anal cancer.
• Novel optical imaging offers potential for point-of-care diagnoses to improve current screening methods.

Purpose of the Study:

• To demonstrate the potential of high-resolution microendoscopy (HRME) combined with a machine learning algorithm for automated, in vivo diagnosis of anal precancer.
• To assess the system's accuracy in detecting AIN 2+ in people living with HIV.

Main Methods:

• High-resolution microendoscopy (HRME) captured real-time images of anal squamous epithelial nuclei after topical staining with proflavine.
• A multitask deep learning network (MTN) analyzed HRME images to compute AIN 2+ probabilities.
• Histopathology served as the gold standard for comparison.

Main Results:

• The MTN achieved an area under the receiver operating curve of 0.84 for AIN 2+ detection.
• At a cutoff of 0.212, the MTN demonstrated sensitivity of 0.92 and specificity of 0.60, comparable to expert high-resolution anoscopy (HRA) impression.
• The study included 77 people living with HIV.

Conclusions:

• The HRME system with machine learning shows potential for automated, in vivo diagnosis of anal precancer.
• This technology could enhance the efficiency and accessibility of anal cancer screening.
• Integration with HRA may enable more selective biopsies and same-day treatment decisions.