Innovative approach for detecting pre-cancerous lesions using large, high-res images

This article from News Medical Life Sciences discusses a Portuguese study that introduces a machine learning solution for automatic detection of cervical dysplasia, a precursor to cervical cancer, using large, high-resolution images. Cervical cancer is preventable with early detection, making screening and lesion detection crucial. The team employed a weakly-supervised machine learning methodology, combining annotated and non-annotated data for model training. The approach utilizes whole-slide images (WSI) containing comprehensive tissue information. Pathology data annotations are challenging due to the large image sizes, making manual annotation time-consuming. The machine learning model grades cervical dysplasia as low or high-grade intraepithelial squamous lesions. The proposed methodology achieved a balanced accuracy of 71.07%, enhancing automatic assessment without manual identification of epithelial areas. The study emphasizes the potential role of computer-aided diagnosis in assisting pathologists and highlights challenges in computational pathology's clinical applicability. Read more here.