{"title":"A statistical framework for evaluating convolutional neural networks. Application to colon cancer","authors":"L. Popa","doi":"10.52846/ami.v48i1.1449","DOIUrl":null,"url":null,"abstract":"\"Purpose: Explore the efficiency of two convolutional neural networks in helping physicians in establishing colon cancer diagnosis from histopathological image scans. Methods: The dataset used in this study contains 357 histopathological image slides that ranged from benign cases to colon cancer grade three. The slides were collected by doctors at the Emergency Hospital of Craiova, Romania. The study proposes a statistical framework that studies the performances of two convolutional neural networks AlexNet and GoogleNet. Results: AlexNet has revealed a competitive accuracy in comparison with GoogleNet. To prove the robustness of the AlexNet in fair terms, we have performed a thorough statistical analysis of its performance. Conclusions: On this particular dataset which contains histopathological image scans regarding colon cancer, the convolutional neural network AlexNet proved to be superior to GoogleNet. \"","PeriodicalId":43654,"journal":{"name":"Annals of the University of Craiova-Mathematics and Computer Science Series","volume":"5 1","pages":""},"PeriodicalIF":0.5000,"publicationDate":"2021-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"2","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Annals of the University of Craiova-Mathematics and Computer Science Series","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.52846/ami.v48i1.1449","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"MATHEMATICS","Score":null,"Total":0}
引用次数: 2
Abstract
"Purpose: Explore the efficiency of two convolutional neural networks in helping physicians in establishing colon cancer diagnosis from histopathological image scans. Methods: The dataset used in this study contains 357 histopathological image slides that ranged from benign cases to colon cancer grade three. The slides were collected by doctors at the Emergency Hospital of Craiova, Romania. The study proposes a statistical framework that studies the performances of two convolutional neural networks AlexNet and GoogleNet. Results: AlexNet has revealed a competitive accuracy in comparison with GoogleNet. To prove the robustness of the AlexNet in fair terms, we have performed a thorough statistical analysis of its performance. Conclusions: On this particular dataset which contains histopathological image scans regarding colon cancer, the convolutional neural network AlexNet proved to be superior to GoogleNet. "