<p>Despite recent surge of interest in deploying colon capsule endoscopy (CCE) for early diagnosis of colorectal diseases, there remains a large gap between the current state of CCE in clinical practice, and the state of its counterpart optical colonoscopy (OC). This is due to several factors, such as low quality bowel cleansing, logistical challenges around both delivery and collection of the capsule, and most importantly, the tedious manual assessment of images after retrieval. Our study, built on the “Danish CareForColon2015 trial (cfc2015)” is aimed at closing this gap, by focusing on the full integration of AI in CCE’s pathway, where image processing steps linked to the detection, localization and characterisation of important findings are carried out autonomously using various AI algorithms. We developed a family of algorithms based on explainable deep neural networks (DNN) that detect polyps within a sequence of images, feed only those images containing polyps into two parallel independent networks to characterize, and estimate the size of important findings. Our recognition DNN to detect colorectal polyps was trained and validated (<InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="41598_2025_89648_Article_IEq1.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="31" /> </InlineMediaObject> <EquationSource Format="TEX">\(70\%\)</EquationSource> </InlineEquation>) and tested (<InlineEquation ID="IEq2"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="41598_2025_89648_Article_IEq2.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="31" /> </InlineMediaObject> <EquationSource Format="TEX">\(30\%\)</EquationSource> </InlineEquation>) on an unaugmented database of 1751 images containing colorectal polyps and 1672 images of normal mucosa reached an impressive sensitivity of <InlineEquation ID="IEq3"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="41598_2025_89648_Article_IEq3.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="43" /> </InlineMediaObject> <EquationSource Format="TEX">\(99.9\%\)</EquationSource> </InlineEquation>, a specificity of <InlineEquation ID="IEq4"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="41598_2025_89648_Article_IEq4.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="43" /> </InlineMediaObject> <EquationSource Format="TEX">\(99.4\%\)</EquationSource> </InlineEquation>, and a negative predictive value (NPV) of <InlineEquation ID="IEq5"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="41598_2025_89648_Article_IEq5.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="43" /> </InlineMediaObject> <EquationSource Format="TEX">\(99.8\%\)</EquationSource> </InlineEquation>. The characterisation DNN trained on an unaugmented database of 317 images featuring neoplastic polyps and 162 images of non-neoplastic polyps reached a sensitivity of <InlineEquation ID="IEq6"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="41598_2025_89648_Article_IEq6.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="43" /> </InlineMediaObject> <EquationSource Format="TEX">\(84.3\%\)</EquationSource> </InlineEquation> and a specificity of <InlineEquation ID="IEq7"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="41598_2025_89648_Article_IEq7.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="43" /> </InlineMediaObject> <EquationSource Format="TEX">\(81.5\%\)</EquationSource> </InlineEquation> in classifying polyps. The size estimation DNN trained on an unaugmented database of 280 images reached an accuracy of <InlineEquation ID="IEq8"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="41598_2025_89648_Article_IEq8.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="31" /> </InlineMediaObject> <EquationSource Format="TEX">\(88\%\)</EquationSource> </InlineEquation> in correctly segmenting the polyps. By automatically incorporating important information including size, location and pathology of the findings into CCE’s pathway, we moved a step closer towards the full integration of explainable AI (XAI) in CCE’s routine clinical practice. This translates into a fewer number of unnecessary investigations and resection of diminutive, insignificant colorectal polyps.</p>

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Towards full integration of explainable artificial intelligence in colon capsule endoscopy’s pathway

  • Esmaeil S. Nadimi,
  • Jan-Matthias Braun,
  • Benedicte Schelde-Olesen,
  • Smith Khare,
  • Vinay C. Gogineni,
  • Victoria Blanes-Vidal,
  • Gunnar Baatrup

摘要

Despite recent surge of interest in deploying colon capsule endoscopy (CCE) for early diagnosis of colorectal diseases, there remains a large gap between the current state of CCE in clinical practice, and the state of its counterpart optical colonoscopy (OC). This is due to several factors, such as low quality bowel cleansing, logistical challenges around both delivery and collection of the capsule, and most importantly, the tedious manual assessment of images after retrieval. Our study, built on the “Danish CareForColon2015 trial (cfc2015)” is aimed at closing this gap, by focusing on the full integration of AI in CCE’s pathway, where image processing steps linked to the detection, localization and characterisation of important findings are carried out autonomously using various AI algorithms. We developed a family of algorithms based on explainable deep neural networks (DNN) that detect polyps within a sequence of images, feed only those images containing polyps into two parallel independent networks to characterize, and estimate the size of important findings. Our recognition DNN to detect colorectal polyps was trained and validated ( \(70\%\) ) and tested ( \(30\%\) ) on an unaugmented database of 1751 images containing colorectal polyps and 1672 images of normal mucosa reached an impressive sensitivity of \(99.9\%\) , a specificity of \(99.4\%\) , and a negative predictive value (NPV) of \(99.8\%\) . The characterisation DNN trained on an unaugmented database of 317 images featuring neoplastic polyps and 162 images of non-neoplastic polyps reached a sensitivity of \(84.3\%\) and a specificity of \(81.5\%\) in classifying polyps. The size estimation DNN trained on an unaugmented database of 280 images reached an accuracy of \(88\%\) in correctly segmenting the polyps. By automatically incorporating important information including size, location and pathology of the findings into CCE’s pathway, we moved a step closer towards the full integration of explainable AI (XAI) in CCE’s routine clinical practice. This translates into a fewer number of unnecessary investigations and resection of diminutive, insignificant colorectal polyps.