Anil Johny, Dr. Madhusoodanan K. N., Dr. Tom J Nallikuzhy
{"title":"用超参数调整优化CNN模型增强组织病理图像分类的稳健性","authors":"Anil Johny, Dr. Madhusoodanan K. N., Dr. Tom J Nallikuzhy","doi":"10.2139/ssrn.3735831","DOIUrl":null,"url":null,"abstract":"The field of pathology has advanced so rapidly that it is now possible to produce whole slide images (WSI) from glass slides with digital scanners producing high-quality images. Image analysis algorithms applied to such digitized images facilitate automatic diagnostic tasks whilst assisting a medical expert. Successful detection of malignancy in histopathological images largely depends on the expertise of radiologists, though they sometimes disagree with their decisions. Computer-aided diagnosis provides a platform for a second opinion in diagnosis, which can improve the reliability of an expert's opinion. Deep learning provides promising results compared to the conventional approach that relies on manual extraction of features which is time-consuming and labor-intense. Due to the huge size, whole slide images are converted into patches and trained using a Convolutional Neural Network (CNN), a variant of the deep learning model for images. Experimental results show that the proposed native model achieved patch wise classification accuracy of 92.8% and area under ROC curve 0.97 which is close to the values while comparing with the existing pre-trained models.","PeriodicalId":18268,"journal":{"name":"Materials Engineering eJournal","volume":"41 1","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2020-11-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"3","resultStr":"{\"title\":\"Optimization of CNN Model With Hyper Parameter Tuning for Enhancing Sturdiness in Classification of Histopathological Images\",\"authors\":\"Anil Johny, Dr. Madhusoodanan K. N., Dr. Tom J Nallikuzhy\",\"doi\":\"10.2139/ssrn.3735831\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"The field of pathology has advanced so rapidly that it is now possible to produce whole slide images (WSI) from glass slides with digital scanners producing high-quality images. Image analysis algorithms applied to such digitized images facilitate automatic diagnostic tasks whilst assisting a medical expert. Successful detection of malignancy in histopathological images largely depends on the expertise of radiologists, though they sometimes disagree with their decisions. Computer-aided diagnosis provides a platform for a second opinion in diagnosis, which can improve the reliability of an expert's opinion. Deep learning provides promising results compared to the conventional approach that relies on manual extraction of features which is time-consuming and labor-intense. Due to the huge size, whole slide images are converted into patches and trained using a Convolutional Neural Network (CNN), a variant of the deep learning model for images. Experimental results show that the proposed native model achieved patch wise classification accuracy of 92.8% and area under ROC curve 0.97 which is close to the values while comparing with the existing pre-trained models.\",\"PeriodicalId\":18268,\"journal\":{\"name\":\"Materials Engineering eJournal\",\"volume\":\"41 1\",\"pages\":\"\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2020-11-23\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"3\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Materials Engineering eJournal\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.2139/ssrn.3735831\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Materials Engineering eJournal","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.2139/ssrn.3735831","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Optimization of CNN Model With Hyper Parameter Tuning for Enhancing Sturdiness in Classification of Histopathological Images
The field of pathology has advanced so rapidly that it is now possible to produce whole slide images (WSI) from glass slides with digital scanners producing high-quality images. Image analysis algorithms applied to such digitized images facilitate automatic diagnostic tasks whilst assisting a medical expert. Successful detection of malignancy in histopathological images largely depends on the expertise of radiologists, though they sometimes disagree with their decisions. Computer-aided diagnosis provides a platform for a second opinion in diagnosis, which can improve the reliability of an expert's opinion. Deep learning provides promising results compared to the conventional approach that relies on manual extraction of features which is time-consuming and labor-intense. Due to the huge size, whole slide images are converted into patches and trained using a Convolutional Neural Network (CNN), a variant of the deep learning model for images. Experimental results show that the proposed native model achieved patch wise classification accuracy of 92.8% and area under ROC curve 0.97 which is close to the values while comparing with the existing pre-trained models.