{"id":249,"date":"2023-10-05T19:00:19","date_gmt":"2023-10-05T19:00:19","guid":{"rendered":"https:\/\/groups.cs.umass.edu\/infofusion\/?page_id=249"},"modified":"2023-10-09T00:59:58","modified_gmt":"2023-10-09T00:59:58","slug":"detecting-mitral-regurgitation-with-cmr","status":"publish","type":"page","link":"https:\/\/groups.cs.umass.edu\/infofusion\/detecting-mitral-regurgitation-with-cmr\/","title":{"rendered":"Machine Learning for Automated Mitral Regurgitation Detection from Cardiac Imaging"},"content":{"rendered":"\n<p><a href=\"https:\/\/link.springer.com\/chapter\/10.1007\/978-3-031-43990-2_23#auth-Ke-Xiao\">Ke <\/a><a href=\"https:\/\/people.cs.umass.edu\/~kexiao\/\">Xiao<\/a>,&nbsp;<a href=\"https:\/\/people.cs.umass.edu\/~elm\/\">Erik Learned-Miller<\/a>,&nbsp;<a href=\"https:\/\/people.cs.umass.edu\/~kalo\/\">Evangelos Kalogerakis<\/a>,&nbsp;<a href=\"https:\/\/scholar.google.com\/citations?user=FuHC3zYAAAAJ&amp;hl=en\">James Priest<\/a>&nbsp;&amp;&nbsp;<a href=\"https:\/\/people.cs.umass.edu\/~mfiterau\/\">Madalina Fiterau<\/a><\/p>\n\n\n\n<p><a href=\"https:\/\/link.springer.com\/chapter\/10.1007\/978-3-031-43990-2_23\">MICCAI 2023 Paper<\/a>, <a href=\"https:\/\/youtu.be\/YCHVoWmi3hY\">MR vs non-MR video<\/a>, <a href=\"https:\/\/youtu.be\/DARZU0QTBmU\">Patch Selection video<\/a>, <a href=\"https:\/\/github.com\/Information-Fusion-Lab-Umass\/CUSSP_UKB_MR\">Github repo<\/a><\/p>\n\n\n\n<p>Mitral regurgitation (MR) is a heart valve disease with potentially fatal consequences that can only be forestalled through timely diagnosis and treatment. Traditional diagnosis methods are expensive, labor-intensive and require clinical expertise, posing a barrier to screening for MR. To overcome this impediment, we propose an automated five stage pipeline named <strong>Cardio-vascular magnetic resonance U-Net localized Self-Supervised Predictor (CUSSP)<\/strong>. CUSSP operates on cardiac magnetic resonance (CMR) imaging slices of the 4-chamber view of the heart. It uses standard computer vision techniques and contrastive models to learn from large amounts of unlabeled data, in conjunction with specialized classifiers to establish the <strong><em>first ever automated MR classification system using CMR imaging sequences<\/em><\/strong>. Evaluated on a test set of 179 labeled &#8212; 154 non-MR and 25 MR &#8212; sequences, CUSSP attains an F1 score of 0.69 and a ROC-AUC score of 0.88, setting the first benchmark result for detecting MR from CMR imaging sequences.<\/p>\n\n\n\n<figure class=\"wp-block-gallery has-nested-images columns-default is-cropped wp-block-gallery-1 is-layout-flex wp-block-gallery-is-layout-flex\">\n<figure class=\"wp-block-image size-large\"><img fetchpriority=\"high\" decoding=\"async\" width=\"600\" height=\"500\" data-id=\"269\" src=\"https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/demo_mr0_4Ch_labeled_cropped.png\" alt=\"\" class=\"wp-image-269\" srcset=\"https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/demo_mr0_4Ch_labeled_cropped.png 600w, https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/demo_mr0_4Ch_labeled_cropped-300x250.png 300w\" sizes=\"(max-width: 600px) 100vw, 600px\" \/><\/figure>\n\n\n\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" width=\"600\" height=\"500\" data-id=\"271\" src=\"https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/demo_mr0_4Ch_open_labeled_cropped.png\" alt=\"\" class=\"wp-image-271\" srcset=\"https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/demo_mr0_4Ch_open_labeled_cropped.png 600w, https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/demo_mr0_4Ch_open_labeled_cropped-300x250.png 300w\" sizes=\"(max-width: 600px) 100vw, 600px\" \/><\/figure>\n\n\n\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" width=\"600\" height=\"500\" data-id=\"272\" src=\"https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/demo_mr_4Ch_jet_label_cropped.png\" alt=\"Illustration of the mitral valve and mitral regurgitation.\" class=\"wp-image-272\" srcset=\"https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/demo_mr_4Ch_jet_label_cropped.png 600w, https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/demo_mr_4Ch_jet_label_cropped-300x250.png 300w\" sizes=\"(max-width: 600px) 100vw, 600px\" \/><\/figure>\n<figcaption class=\"blocks-gallery-caption wp-element-caption\"><strong>Figure 1<\/strong>. Three CMR images showing the long-axis four-chamber view of the heart. Left: a heart with normal mitral valve. Middle: a heart with normal mitral valve when the valve leaflets are open. Right: a heart with mitral regurgitation. The red dotted line indicates the mitral valve<\/figcaption><\/figure>\n\n\n\n<figure class=\"wp-block-embed is-type-video is-provider-youtube wp-block-embed-youtube wp-embed-aspect-4-3 wp-has-aspect-ratio\"><div class=\"wp-block-embed__wrapper\">\n<iframe title=\"MR vs non-MR comparison in 4CH CMR\" width=\"600\" height=\"450\" src=\"https:\/\/www.youtube.com\/embed\/YCHVoWmi3hY?feature=oembed\" frameborder=\"0\" allow=\"accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share\" referrerpolicy=\"strict-origin-when-cross-origin\" allowfullscreen><\/iframe>\n<\/div><\/figure>\n\n\n\n<p><strong>Challenges:<\/strong> <\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Requires multiple video frames, and tracking blood flow.<\/li>\n\n\n\n<li>The regurgitant jet can be faint in the video, and thus difficult to detect.<\/li>\n\n\n\n<li>Blood flow has high turbulence<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">Segmentation<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Segmentation for CMR views including long-axis 2-chamber (LAX-2CH) view, 4-chamber (LAX-4CH) view and short-axis (SA) view.<\/li>\n\n\n\n<li><strong>For patch selection<\/strong>:<br>Semantic segmentation of 4CH CMR imaging sequences.<\/li>\n\n\n\n<li><strong>For tabular data<\/strong>:<br>Semantic segmentation of LAX-2CH, LAX-4CH and SA CMR imaging<br>sequences (to estimate the cardiac structure and functions in Table 1).<\/li>\n<\/ul>\n\n\n\n<figure class=\"wp-block-gallery has-nested-images columns-default is-cropped wp-block-gallery-2 is-layout-flex wp-block-gallery-is-layout-flex\">\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"960\" height=\"720\" data-id=\"276\" src=\"https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/example_4Ch_seg-edited.png\" alt=\"\" class=\"wp-image-276\" srcset=\"https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/example_4Ch_seg-edited.png 960w, https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/example_4Ch_seg-edited-300x225.png 300w, https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/example_4Ch_seg-edited-768x576.png 768w\" sizes=\"(max-width: 960px) 100vw, 960px\" \/><\/figure>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"768\" data-id=\"277\" src=\"https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/example_2Ch_seg-edited-1024x768.png\" alt=\"\" class=\"wp-image-277\" srcset=\"https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/example_2Ch_seg-edited-1024x768.png 1024w, https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/example_2Ch_seg-edited-300x225.png 300w, https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/example_2Ch_seg-edited-768x576.png 768w, https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/example_2Ch_seg-edited.png 1040w\" sizes=\"(max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"999\" height=\"750\" data-id=\"278\" src=\"https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/example_sa_seg-edited.png\" alt=\"Example of the segmentation outputs\" class=\"wp-image-278\" srcset=\"https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/example_sa_seg-edited.png 999w, https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/example_sa_seg-edited-300x225.png 300w, https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/example_sa_seg-edited-768x577.png 768w\" sizes=\"(max-width: 999px) 100vw, 999px\" \/><\/figure>\n<figcaption class=\"blocks-gallery-caption wp-element-caption\"><strong>Figure 2<\/strong>. Example of the segmentation outputs of the long axis 4CH (left), 2CH (middle) CMR<br>view imaging data and the short axis (right) CMR imaging data<\/figcaption><\/figure>\n\n\n\n<p><\/p>\n\n\n\n<figure class=\"wp-block-table\"><table><tbody><tr><td class=\"has-text-align-center\" data-align=\"center\">Left Atrium<\/td><td class=\"has-text-align-center\" data-align=\"center\">Right Atrium<\/td><td class=\"has-text-align-center\" data-align=\"center\">Left Ventricle<\/td><td class=\"has-text-align-center\" data-align=\"center\">Right Ventricle<\/td><\/tr><tr><td class=\"has-text-align-center\" data-align=\"center\">Vol Max (mL)<br>Vol Min (mL)<br>Stroke Vol (mL)<br>Ejection Fraction (%)<\/td><td class=\"has-text-align-center\" data-align=\"center\">Vol Max (mL)<br>Vol Min (mL)<br>Stroke Vol (mL)<br>Ejection Fraction (%)<\/td><td class=\"has-text-align-center\" data-align=\"center\">End Systolic Vol (mL)<br>End Diastolic Vol (mL)<br>Stroke Vol (mL)<br>Ejection Fraction (%)<br>Cardiac Output (L\/min)<br>Mass (g)<\/td><td class=\"has-text-align-center\" data-align=\"center\">End Systolic Vol (mL)<br>End Diastolic Vol (mL)<br>Stroke Vol (mL)<br>Ejection Fraction (%)<\/td><\/tr><\/tbody><\/table><figcaption class=\"wp-element-caption\"><strong>Table 1<\/strong>. Cardiac measurements derived from the semantic segmentation of the CMR.<\/figcaption><\/figure>\n\n\n\n<h2 class=\"wp-block-heading\">Our Prior Attempts: Random Forests and CNN-LSTM<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Random forest (RF) classifier trained on the tabular heart measurements shown in Table 1.<\/li>\n\n\n\n<li>A deep learning model using a CNN-LSTM pipeline, shown in Figure 3.<\/li>\n<\/ul>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"458\" src=\"https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/CNN_LSTM_pipeline_simplied-2-1024x458.png\" alt=\"\" class=\"wp-image-282\" srcset=\"https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/CNN_LSTM_pipeline_simplied-2-1024x458.png 1024w, https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/CNN_LSTM_pipeline_simplied-2-300x134.png 300w, https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/CNN_LSTM_pipeline_simplied-2-768x344.png 768w, https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/CNN_LSTM_pipeline_simplied-2-1536x687.png 1536w, https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/CNN_LSTM_pipeline_simplied-2-2048x916.png 2048w, https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/CNN_LSTM_pipeline_simplied-2-1200x537.png 1200w\" sizes=\"(max-width: 1024px) 100vw, 1024px\" \/><figcaption class=\"wp-element-caption\"><strong>Figure 3<\/strong>. Overview of the CNN-LSTM method pipeline for MR classification. The pipeline includes segmentation, masking and cropping for pre-processing, and DenseNet as the frame encoder, bi-directional LSTM as the sequence encoder, with an injected attention layer after the first convolutional layer of the DenseNet block.<\/figcaption><\/figure>\n\n\n\n<h2 class=\"wp-block-heading\">Patch Selection<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Localization:<\/strong> localizing the region in the heart where mitral regurgitation occurs.<\/li>\n\n\n\n<li><strong>Equalization:<\/strong> increasing contrast for the regurgitant jet.<\/li>\n\n\n\n<li><strong>Selection:<\/strong> cropping the patch for the ML system.<\/li>\n<\/ul>\n\n\n<div class=\"wp-block-image\">\n<figure class=\"aligncenter size-full is-resized\"><img loading=\"lazy\" decoding=\"async\" src=\"https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/barlow-data-prep-edited.png\" alt=\"\" class=\"wp-image-285\" width=\"694\" height=\"434\" srcset=\"https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/barlow-data-prep-edited.png 512w, https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/barlow-data-prep-edited-300x188.png 300w\" sizes=\"(max-width: 694px) 100vw, 694px\" \/><figcaption class=\"wp-element-caption\"><strong>Figure 4<\/strong>.Top: the 4CH CMR image in the original contrast (left), after histogram equalization based on the whole image (middle), and after histogram equalization based on the patch (right). Blue contours outline the left atrium, and the red square boxes outline the selected patch. Bottom: Example of the cropped mitral valve patch as outlined in the red square boxes above.<\/figcaption><\/figure>\n<\/div>\n\n\n<figure class=\"wp-block-embed is-type-video is-provider-youtube wp-block-embed-youtube wp-embed-aspect-4-3 wp-has-aspect-ratio\"><div class=\"wp-block-embed__wrapper\">\n<iframe title=\"CUSSP Data Patch Selection\" width=\"600\" height=\"450\" src=\"https:\/\/www.youtube.com\/embed\/DARZU0QTBmU?feature=oembed\" frameborder=\"0\" allow=\"accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share\" referrerpolicy=\"strict-origin-when-cross-origin\" allowfullscreen><\/iframe>\n<\/div><figcaption class=\"wp-element-caption\">Patch selection process for CUSSP pipeline. <\/figcaption><\/figure>\n\n\n\n<h2 class=\"wp-block-heading\">CUSSP Pipeline Stages<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>CUSSP pipeline:<\/strong> (1) segmentation, (2) localization, (3) cropping, (4) equalization, and (5) prediction<\/li>\n\n\n\n<li><strong>Prediction steps:<\/strong> (i) Barlow-Twins network, (ii) Siamese network, (iii) Multilayer perceptron (MLP).<\/li>\n<\/ul>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"605\" src=\"https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/MR_pipeline_v3-1-1024x605.png\" alt=\"\" class=\"wp-image-253\" srcset=\"https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/MR_pipeline_v3-1-1024x605.png 1024w, https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/MR_pipeline_v3-1-300x177.png 300w, https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/MR_pipeline_v3-1-768x454.png 768w, https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/MR_pipeline_v3-1-1536x908.png 1536w, https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/MR_pipeline_v3-1-2048x1211.png 2048w, https:\/\/groups.cs.umass.edu\/infofusion\/wp-content\/uploads\/sites\/26\/2023\/10\/MR_pipeline_v3-1-1200x709.png 1200w\" sizes=\"(max-width: 1024px) 100vw, 1024px\" \/><figcaption class=\"wp-element-caption\"><strong>Figure 5<\/strong>. Overview of the CUSSP pipeline for MR classification<\/figcaption><\/figure>\n","protected":false},"excerpt":{"rendered":"<p>Ke Xiao,&nbsp;Erik Learned-Miller,&nbsp;Evangelos Kalogerakis,&nbsp;James Priest&nbsp;&amp;&nbsp;Madalina Fiterau MICCAI 2023 Paper, MR vs non-MR video, Patch Selection video, Github repo Mitral regurgitation (MR) is a heart valve disease with potentially fatal consequences that can only be forestalled through timely diagnosis and treatment. Traditional diagnosis methods are expensive, labor-intensive and require clinical expertise, posing a barrier to screening &hellip; <a href=\"https:\/\/groups.cs.umass.edu\/infofusion\/detecting-mitral-regurgitation-with-cmr\/\" class=\"more-link\">Continue reading<span class=\"screen-reader-text\"> &#8220;Machine Learning for Automated Mitral Regurgitation Detection from Cardiac Imaging&#8221;<\/span><\/a><\/p>\n","protected":false},"author":35,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"footnotes":""},"class_list":["post-249","page","type-page","status-publish","hentry","group-blog","no-sidebar","hfeed"],"_links":{"self":[{"href":"https:\/\/groups.cs.umass.edu\/infofusion\/wp-json\/wp\/v2\/pages\/249","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/groups.cs.umass.edu\/infofusion\/wp-json\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/groups.cs.umass.edu\/infofusion\/wp-json\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/groups.cs.umass.edu\/infofusion\/wp-json\/wp\/v2\/users\/35"}],"replies":[{"embeddable":true,"href":"https:\/\/groups.cs.umass.edu\/infofusion\/wp-json\/wp\/v2\/comments?post=249"}],"version-history":[{"count":13,"href":"https:\/\/groups.cs.umass.edu\/infofusion\/wp-json\/wp\/v2\/pages\/249\/revisions"}],"predecessor-version":[{"id":293,"href":"https:\/\/groups.cs.umass.edu\/infofusion\/wp-json\/wp\/v2\/pages\/249\/revisions\/293"}],"wp:attachment":[{"href":"https:\/\/groups.cs.umass.edu\/infofusion\/wp-json\/wp\/v2\/media?parent=249"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}