Detecting Diseases in Gastrointestinal Biopsy Images using deep learning
Machine learning and computer vision have found applications in medical science and, recently, pathology. In particular, deep learning methods for medical diagnostic imaging can reduce delays in diagnosis and give improved accuracy rates over other analysis techniques.
This project focuses on methods with applicability to the automated diagnosis of images obtained from gastrointestinal biopsies. These deep learning techniques for biopsy images may help detect distinguishing features in tissues affected by enteropathies.
Learning from different areas of an image, or looking for similar patterns in new images, allow for the development of potential classification or clustering models Techniques like these provide a cutting-edge solution to detecting anomalies.
In this project, we explore state-of-the-art deep learning architectures used for the visual recognition of natural images and assess their applicability in medical image analysis of digitized human gastrointestinal biopsy slides.
Clinical biopsy images are used for diagnosing various diseases. However, it can become difficult to distinguish patterns when associated conditions share histological features. This can lead to a wrong diagnosis and, hence, wrong treatment. Recent advances in technology can help distinguish between closely related conditions, reduce misdiagnoses and reduce false positives. An approach that can learn from data rather than rely on pre-defined features can potentially pick up on visually challenging differences. Deep learning techniques for computer vision like Convolutional Neural Networks (CNNs) can learn such complex features and help solve the problem. CNN’s are used primarily for image classification and object detection problems. Applying CNNs to find differences in histologically similar issues that might be indistinguishable under a microscope is the primary focus of this paper.
- K-means clustering
- Gabor Wavelet
- Normal threshold
- SVM classifier
- Difficult to get accurate results
- Not applicable for multiple images for disease detection in a short time
- Medical Resonance images contain a noise caused by operator performance which can lead to serious inaccuracies in classification
The deep learning techniques for biopsy images may help detect distinguishing features in tissues affected by enteropathies. Learning from different areas of an image, or looking for similar patterns in new images, allow for the development of potential classification or clustering models
- DE noising complexity is low
- It is useful to classify the cancer images for accurate detection.
- CNN training is faster and has high compatibility
- Clinical Decision Support System
- Matlab 2018a and above
- Deep learning toolbox
 Rubio-Tapia A, Ludvigsson JF, Brantner TL, Murray JA, Everhart JE., “The prevalence of celiac disease in the United States.” The American Journal of Gastroenterology 107(10):1538-1544; quiz 1537, 1545, 2012.
 Husby S, Koletzko S, Korponay-Szabo IR, et al. European Society for Pediatric Gastroenterology, “Hepatology, and Nutrition guidelines for the diagnosis of coeliac disease.”, Journal of Pediatric Gastroenterology and Nutrition 54(1):136-160, 2012.
 Ali A, Iqbal NT, Sadiq K. “Environmental enteropathy”, Current Opinion in Gastroenterology;32(1):12-17, 2012.
 Uddin MI, Islam S, Nishat NS, et al. “Biomarkers of Environmental Enteropathy are Positively Associated with Immune Responses to an Oral Cholera Vaccine in Bangladeshi Children.”, PLOS Neglected Tropical Diseases 10(11): e0005039, 2016.
 Campbell DI, Elia M, Lunn PG., “Growth faltering in rural Gambian infants is associated with impaired small intestinal barrier function, leading to endotoxemia and systemic inflammation”, Journal of Nutrition 133(5):1332-1338, 2016.