ABSTRACT
The early detection of pancreatic ductal adenocarcinoma (PDAC) is essential for optimal treatment of pancreatic cancer patients. We propose a tumor detection framework to improve the detection of pancreatic head tumors on CT scans. In this retrospective research study, CT images of 99 patients with pancreatic head cancer and 98 control cases from the Catharina Hospital Eindhoven were collected. A multi-stage 3D U-Net-based approach was used for PDAC detection including clinically significant secondary features such as pancreatic duct and common bile duct dilation. The developed algorithm was evaluated using a local test set comprising 59 CT scans. The model was externally validated in 28 pancreatic cancer cases of a publicly available medical decathlon dataset. The tumor detection framework achieved a sensitivity of 0.97 and a specificity of 1.00, with an area under the receiver operating curve (AUROC) of 0.99, in detecting pancreatic head cancer in the local test set. In the external test set, we obtained similar results, with a sensitivity of 1.00. The model provided the tumor location with acceptable accuracy obtaining a DICE Similarity Coefficient (DSC) of 0.37. This study shows that a tumor detection framework utilizing CT scans and secondary signs of pancreatic cancer can detect pancreatic tumors with high accuracy.
ABSTRACT
Early Barrett's neoplasia are often missed due to subtle visual features and inexperience of the non-expert endoscopist with such lesions. While promising results have been reported on the automated detection of this type of early cancer in still endoscopic images, video-based detection using the temporal domain is still open. The temporally stable nature of video data in endoscopic examinations enables to develop a framework that can diagnose the imaged tissue class over time, thereby yielding a more robust and improved model for spatial predictions. We show that the introduction of Recurrent Neural Network nodes offers a more stable and accurate model for tissue classification, compared to classification on individual images. We have developed a customized Resnet18 feature extractor with four types of classifiers: Fully Connected (FC), Fully Connected with an averaging filter (FC Avg(n = 5)), Long Short Term Memory (LSTM) and a Gated Recurrent Unit (GRU). Experimental results are based on 82 pullback videos of the esophagus with 46 high-grade dysplasia patients. Our results demonstrate that the LSTM classifier outperforms the FC, FC Avg(n = 5) and GRU classifier with an average accuracy of 85.9% compared to 82.2%, 83.0% and 85.6%, respectively. The benefit of our novel implementation for endoscopic tissue classification is the inclusion of spatio-temporal information for improved and robust decision making, and it is the first step towards full temporal learning of esophageal cancer detection in endoscopic video.
