Development of Gastric Cancer Classification System using Deep Learning algorithm

Cancer is a disease in which the normal growth of cells is not regulated and they may spread to other parts of the body. The third most universal cause of cancer deaths is gastric cancer that leads to colossal number of patients affected with in the urban areas. Traditional diagnostic inventions such as endoscopy and surgery are highly invasive technique due to the excision of tissues from gastrointestinal tract. Invasive procedures cause patient discomfort, increased risk of complications and the availability is only on a limited extent. Currently the need of early diagnosis helps to predict the disease. In this work, a gastric cancer classification system based on deep learning algorithms specifically CNN architectures were proposed. It was analysed using different CNN models namely DenseNet121, VGG16, VGG19 and InceptionV3. The 98.3% accuracy was achieved in VGG16 when compared three models of classification. VGG16 models show the potential to make a totally non-invasive process with almost the highest accuracy in the gastric cancer diagnosis. It would promote diagnostic capabilities even in under-served areas.