The early detection of polyps could help prevent colorectal cancer. The automated detection of polyps on the colon walls could reduce the number of false negatives that occur due to manual examination errors or polyps being hidden behind folds, and could also help doctors locate polyps from screening tests such as colonoscopy and wireless capsule endoscopy. Losing polyps may result in lesions evolving badly. In this paper, we propose a modified region-based convolutional neural network (R-CNN) by generating masks around polyps detected from still frames. The locations of the polyps in the image are marked, which assists the doctors examining the polyps. The features from the polyp images are extracted using pre-trained Resnet-50 and Resnet-101 models through feature extraction and fine-tuning techniques. Various publicly available polyp datasets are analyzed with various pertained weights. It is interesting to notice that fine-tuning with balloon data (polyp-like natural images) improved the polyp detection rate. The optimum CNN models on colonoscopy datasets including CVC-ColonDB, CVC-PolypHD, and ETIS-Larib produced values (F1 score, F2 score) of (90.73, 91.27), (80.65, 79.11), and (76.43, 78.70) respectively. The best model on the wireless capsule endoscopy dataset gave a performance of (96.67, 96.10). The experimental results indicate the better localization of polyps compared to recent traditional and deep learning methods.

In this paper, we present a 3D surface imaging technique to re-engineer traditional endoscopy into a quantitatively capable instrument. In our design concept, we demonstrate that, by utilizing two fiber bundle channels and structured light based 3D reconstruction principle, we can obtain the 3D information on a GI tract surface in all most real time fashion. The two fiber bundles are used for pattern light projection and image capture. The implied significance of our design and experiment includes: (a) It is possible to convert the traditional 2D video based endoscope into a 3D endoscope through minimum modifications; and (b) The proposed 3D endoscope allows clinicians to obtain the actual size information on any target of interest during the procedures.

Currently, clinical assessments on wound tissues are all done through visual inspection only, and wound size measurements (width, length and depth) are commonly done through either a disposable ruler, transparent tape based edge tracing, liquid filling, or even a dipstick swab. Since it is very subjective to select reference points to measure the wound’s dimension, the measured results are often non-repeatable and crude with up to 44% error in area measurement reported. We present a new measurement solution or app on mobile platforms to directly tackle the challenges facing current chronic wound care. The app uses computer vision algorithms for quantitative measurement in 3D digital space. Since our app utilizes a client-server framework to allow mobile app based simple wound video uploading, easy patient data access and analysis, it is designed to significantly reduce the inconvenience and high cost barriers for chronic wound care.

In this paper, we present a thermal imaging based app to augment traditional appearance based wound growth monitoring. Accurate diagnose and track of wound healing enables physicians to effectively assess, document, and individualize the treatment plan given to each wound patient. Currently, wounds are primarily examined by physicians through visual appearance and wound area. However, visual information alone cannot present a complete picture on a wound’s condition. In this paper, we use a smartphone attached thermal imager and evaluate its effectiveness on augmenting visual appearance based wound diagnosis. Instead of only monitoring wound temperature changes on a wound, our app presents physicians a comprehensive measurements including relative temperature, wound healing thermal index, and wound blood flow. Through the rat wound experiments and by monitoring the integrated thermal measurements over 3 weeks of time frame, our app is able to show the underlying healing process through the blood flow. The implied significance of our app design and experiment includes: (a) It is possible to use a low cost smartphone attached thermal imager for added value on wound assessment, tracking, and treatment; and (b) Thermal mobile app can be used for remote wound healing assessment for mobile health based solution.

In this paper, we propose an innovative mobile app for remote wound growth measurement, tracking, and diagnosis. Through the collected 150 videos and over 400 measurements, we found that the proposed app showed an average relative area error of 1.87%, and was also able to accurately distinguish between the three typical wound tissues. Once the app is fully automated and integrated with other mHealth sensors, we expect it will significantly reduce the treatment costs, the mortality rate, and amputation rate.

In this paper, we present a novel and clinically valuable software platform for automatic detection on gastrointestinal (GI) tract from Capsule Endoscopy (CE) videos. Typical CE videos take about 8 hours. They have to be reviewed manually by physicians to detect and locate diseases such as ulcers and bleedings. The process is time consuming. Moreover, because of the long-time manual review, it is easy to lead to miss-finding. Working with our collaborators, we were focusing on developing a software platform called GISentinel, which can fully automated GI tract detection and classification. This software includes 3 parts: the frequency based Log-Gabor filter regions of interest (ROI) extraction, the unique feature selection and validation method (e.g. illumination invariant feature, color independent features, and symmetrical texture features), and the cascade SVM classification for handling ulcer vs. non-ulcer cases. After the experiments, this SW gave descent results. In frame-wise, the detection rate is 69.65% (319/458). In instance-wise, the detection rate is 82.35%(28/34).The false alarm rate is 16.43% (34/207). This work is a part of our innovative 2D/3D based GI tract disease detection software platform. The final goal of this SW is to find and classification of major GI tract diseases intelligently, such as bleeding, ulcer, and polyp from the CE videos. This paper will mainly describe the automatic detection functional module.