Analysis of gastric cancer with machine learning techniques

In this thesis, stomach cell histopathology images were used. These histopathological images were obtained with the aid of light microscopy. These images are three types: normal, benign and malignant. With these images, higher accuracy stomach cancer detection and diagnosis is targeted. Wavelet transform methods were used for feature extraction of these images, artificial neural networks and decision tree methods were used for classification. Wavelet transform methods are used as haar, daubechies, biorthogonal, coiflets, symlets, inverse biorthogonal methods. For use in the methods, 4 sectional images were taken for each of the tissues taken from 90 patients.Total number of tissue images is 360 pieces. By comparing the accuracy rates obtained from the methods used in this thesis, it was determined which method gave the best results. Thanks to the high accuracy rate obtained for the detection of the gastric cancer; normal, benign and malignant tumors were separated. For this reason, high accuracy rate has been found thanks to the results obtained by passing through various stages for use of stomach images in computer-aided diagnostic systems.