Biology subjects

Kanavati, F.

Publications and source records attributed to Kanavati, F..

2 recordsLinked to original sources

bioRxiv2021

Breast invasive ductal carcinoma classification on whole slide images with weakly-supervised and transfer learning

AO_SCPLOWBSTRACTC_SCPLOWInvasive ductal carcinoma (IDC) is the most common form of breast cancer. For the non-operative diagnosis of breast carcinoma, core needle biopsy has been widely used in recent years which allows evaluation of both cytologic and tissue architectural features; so that it can provide a definitive diagnosis between IDC and benign lesion (e.g., fibroadenoma). Histopathological diagnosis based on core needle biopsy specimens is currently the cost effective method; therefore, it is an area that could benefit from AI-based tools to aid pathologists in their pathological diagnosis workflows. In this paper, we trained an Invasive Ductal Carcinoma (IDC) Whole Slide Image (WSI) classification model using transfer learning and weakly-supervised learning. We evaluated the model on a core needle biopsy (n=522) test set as well as three surgical test sets (n=1129) obtaining ROC AUCs in the range of 0.95-0.98.

pathology↗

bioRxiv2021

Deep learning models for poorly differentiated colorectal adenocarcinoma classification in whole slide images using transfer learning

Colorectal poorly differentiated adenocarcinoma (ADC) is known to have a poor prognosis as compared with well to moderately differentiated ADC. The frequency of poorly differentiated ADC is relatively low (usually less than 5% among colorectal carcinomas). Histopathological diagnosis based on endoscopic biopsy specimens is currently the most cost effective method to perform as part of colonoscopic screening in average risk patients, and it is an area that could benefit from AI-based tools to aid pathologists in their clinical workflows. In this study, we trained deep learning models to classify poorly differentiated colorectal ADC from Whole Slide Images (WSIs) using a simply transfer learning method. We evaluated the models on a combination of test sets obtained from five distinct sources, achieving receiver operator curve (ROC) area under the curves (AUCs) in the range of 0.94-0.98.

pathology↗