Researchers at Washington University in St. Louis are developing a deep learning-based pattern recognition (PR)-OCT system that will automate image processing and provide accurate, computer-aided diagnosis of colorectal cancer potentially in real time. The technique combines OCT and deep learning to detect patterns in the images of normal and abnormal tissue samples. The researchers began using OCT in 2017 as a research tool to image samples of colorectal tissue removed from patients at the Washington University School of Medicine. Yifeng Zeng, a biomedical engineering doctoral student, observed that the healthy colorectal tissue had a pattern that looked similar to teeth. However, the precancerous and cancerous tissues rarely showed this pattern. The teeth pattern was caused by light attenuation of the healthy mucosa microstructures of the colorectal tissue. Zeng began working with Shiqi Xu, also a graduate student, to train RetinaNet, a neural network model of the brain, to capture and learn the structural patterns in human colon OCT images. The researchers trained and tested the network using about 26,000 OCT images acquired from 20 tumor areas, 16 benign areas, and six other abnormal areas in patient tissue samples. The trained network successfully detected the patterns that identified normal and neoplastic colorectal tissue. Experimental diagnoses predicted by the PR-OCT system were compared to evaluations of the tissue specimens performed using standard histology. The PR-OCT system was able to identify tumors with 100% accuracy in this pilot study, compared with pathology reports. A sensitivity of 100% and specificity of 99.7% were achieved. The PR-OCT imaging approach detected images of colon cancer (top photo) and of normal colon tissue. The green boxes indicate the scores of probability of the predicted “teeth” patterns in the tissue. Courtesy of the Zhu Lab. “The unique part of our system is that we could detect a structural pattern within the image,” Zeng said. “Using OCT, we are imaging something that we can find a pattern across all normal tissues. Then we can use this pattern to classify abnormal and cancerous tissue for accurate diagnosis.” The researchers are developing a catheter that could be used simultaneously with a colonoscopy endoscope to analyze the teeth-like pattern on the surface of the colon tissue and to provide a score of probability of cancer from RetinaNet to the surgeon. “Right now, we can obtain the feedback in 4 seconds,” Zeng said. “With further development of computation speed and the catheter, we can provide the feedback to surgeons in real time.” OCT can be used to detect the difference in the way healthy and diseased tissues refract light and is highly sensitive to precancerous and early cancer morphological changes. When the PR-OCT system is further developed, the researchers believe that the technique could be used to assist with screening deeply seated precancerous polyps and early-stage colon cancers. “We think this technology, combined with the colonoscopy endoscope, will be very helpful to surgeons in diagnosing colorectal cancer,” professor Quing Zhu said. “More research is necessary, but the idea is that when the surgeons use colonoscopy to examine the colon surface, this technology could be zoomed in locally to help make a more accurate diagnosis of deeper precancerous polyps and early-stage cancers versus normal tissue.” The research was accepted by Theranostics and published as an advance article: “Real-time colorectal cancer diagnosis using PR-OCT with deep learning,” Y. Zeng et al. Theranostics, 2019 (doi:10.7150/thno.40099).