Early detection of gastric dysplasia using deep learning methods

The cancer rate around the world varies by country and region. Cancer rate refers to the probability that a group of people will be diagnosed with cancer. This rate is usually expressed as cancer cases per 100,000 people and is reported annually. Cancer is the uncontrolled proliferation and growth of cells anywhere in our body. Cancer cells, unlike normal cells, do not know how to die and continue to multiply constantly. This situation causes damage to tissues and organs, leading to cancer. There are many factors that affect cancer rates, including lifestyle choices, genetic predisposition, environmental factors and early diagnosis. Smoking, alcohol consumption, obesity, unhealthy diet, sedentary lifestyle and exposure to carcinogenic substances are effective on cancer. Gastric dysplasia is when the cells lining the inner surface of the stomach stop growing and developing normally. These cells begin to grow and multiply uncontrollably and eventually turn into stomach cancer. The main goal is to minimize this disease before it turns into cancer. For this reason, in this thesis study, deep learning and Convolutional Neural Networks (ESA) techniques were used to detect gastric dysplasia at an early stage. In the study, feature extraction was performed on histopathological images using more than one ESA model. In the study, the features obtained as a result of feature extraction using VGG16, VGG19, ResNet, MobileNet, NasNet and EfficientNet models were divided into three separate classes using Support Vector Machine (SVM), Nearest Neighbor (KNN) and Multi-Layer Perceptrons (MCA) classifiers. Additionally, the ESA model was used. The results obtained proved that convolutional neural networks are a successful method in classifying sick and healthy images.