Efficient Gastric Cancer Detection Using Deep Belief Network Compared Over Support Vector Machine With Improved Accuracy

The aim of this study is to develop an efficient method for gastric cancer detection utilizing a novel Deep Belief Network (DBN) architecture, comparing its performance with the traditional Support Vector Machine (SVM) approach to achieve improved accuracy. The GasHisSDB dataset, central to this study, comprises 410 pathological images meticulously curated to represent a wide range of gastric cancer manifestations. The sample size of 40 (20 samples per group) is determined to be sufficient for this study, considering the statistical parameters and the expected effect size. SVM model achieved an accuracy of $86.95 \%$, while the DBN model demonstrated a superior performance with an accuracy of $\mathbf{9 6. 0 6 \%}$. A two-tailed significance test was conducted to assess the statistical significance of the difference in accuracy between the two models. The test yielded a p-value of 0.001, which is below the conventional threshold of $P\lt 0.05$. The research conducted provides compelling evidence of the efficacy of novel Deep Belief Networks (DBN) in the detection of gastric cancer, surpassing the performance of traditional Support Vector Machines (SVM)