<!DOCTYPE html>
<html style="font-size: 16px;">
<head>
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<meta charset="utf-8">
<meta name="keywords" content="Intracranial hemorrhage, bleeding that occurs inside the cranium, is a serious health problem requiring
rapid and often intensive medical treatment. Such a condition is traditionally diagnosed by highly
trained specialists analyzing a Computed tomography (CT)">
<meta name="description" content="Intracranial hemorrhage, bleeding that occurs inside the cranium, is a serious health problem requiring
rapid and often intensive medical treatment. Such a condition is traditionally diagnosed by highly
trained specialists analyzing a Computed tomography (CT)">
<meta name="page_type" content="np-template-header-footer-from-plugin">
<title>Intracranial Hemorrhage Detection</title>
<link rel="stylesheet" href="nicepage.css" media="screen">
<link rel="stylesheet" href="Minimalist-Design-1.css" media="screen">
<script class="u-script" type="text/javascript" src="jquery-1.9.1.min.js" defer=""></script>
<script class="u-script" type="text/javascript" src="nicepage.js" defer=""></script>
<meta name="generator" content="Nicepage 3.29.3, nicepage.com">
<link id="u-theme-google-font" rel="stylesheet" href="https://fonts.googleapis.com/css?family=Roboto:100,100i,300,300i,400,400i,500,500i,700,700i,900,900i|Open+Sans:300,300i,400,400i,600,600i,700,700i,800,800i">
<link id="u-page-google-font" rel="stylesheet" href="https://fonts.googleapis.com/css?family=Roboto:100,100i,300,300i,400,400i,500,500i,700,700i,900,900i|Roboto:100,100i,300,300i,400,400i,500,500i,700,700i,900,900i">
<!-- Global site tag (gtag.js) - Google Analytics -->
<script async src="https://www.googletagmanager.com/gtag/js?id=G-XFMEZZCLSR"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag(){dataLayer.push(arguments);}
gtag('js', new Date());
gtag('config', 'G-XFMEZZCLSR');
</script>
<script type="application/ld+json">{
"@context": "http://schema.org",
"@type": "Organization",
"name": ""
}</script>
<meta name="theme-color" content="#478ac9">
<meta property="og:title" content="Intracranial Hemorrhage Detection">
<meta property="og:type" content="website">
</head>
<body class="u-body u-overlap"><header class="u-clearfix u-header u-header" id="sec-3090"><div class="u-clearfix u-sheet u-sheet-1">
<nav class="u-menu u-menu-dropdown u-offcanvas u-menu-1">
<div class="menu-collapse" style="font-size: 1rem; letter-spacing: 0px; font-weight: 700; text-transform: uppercase;">
<a class="u-button-style u-custom-left-right-menu-spacing u-custom-padding-bottom u-custom-text-active-color u-custom-text-color u-custom-text-hover-color u-custom-top-bottom-menu-spacing u-nav-link u-text-active-palette-1-light-2 u-text-body-alt-color u-text-hover-palette-1-light-1" href="#" style="text-decoration: none !important;">
<svg><use xlink:href="#menu-hamburger"></use></svg>
<svg version="1.1" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink"><defs><symbol id="menu-hamburger" viewBox="0 0 16 16" style="width: 16px; height: 16px;"><rect y="1" width="16" height="2"></rect><rect y="7" width="16" height="2"></rect><rect y="13" width="16" height="2"></rect>
</symbol>
</defs></svg>
</a>
</div>
<div class="u-custom-menu u-nav-container">
<ul class="u-nav u-spacing-0 u-unstyled u-nav-1"><li class="u-nav-item"><a class="u-button-style u-nav-link u-text-active-white u-text-hover-white u-text-palette-3-base" href="Page-1.html" style="padding: 10px 20px; text-decoration: none !important;">Page 1</a>
</li></ul>
</div>
<div class="u-custom-menu u-nav-container-collapse">
<div class="u-black u-container-style u-inner-container-layout u-opacity u-opacity-95 u-sidenav">
<div class="u-inner-container-layout u-sidenav-overflow">
<div class="u-menu-close"></div>
<ul class="u-align-center u-nav u-popupmenu-items u-unstyled u-nav-2"><li class="u-nav-item"><a class="u-button-style u-nav-link" href="Page-1.html">Page 1</a>
</li></ul>
</div>
</div>
<div class="u-black u-menu-overlay u-opacity u-opacity-70"></div>
</div>
</nav>
</div></header>
<section class="u-align-right u-clearfix u-image u-section-1" id="carousel_17e5" data-image-width="1920" data-image-height="1080">
<div class="u-clearfix u-sheet u-sheet-1">
<h5 class="u-text u-text-body-alt-color u-text-1"> Artificial intelligence powered </h5>
<h1 class="u-text u-text-body-alt-color u-text-2" spellcheck="false"> intracranial hemorrhage <br>detection
</h1>
</div>
</section>
<section class="u-align-center u-black u-clearfix u-section-2" id="carousel_9866">
<div class="u-clearfix u-sheet u-sheet-1">
<h5 class="u-text u-text-1"> Intracranial hemorrhage detection using neural network based methods</h5>
<h3 class="u-text u-text-2" spellcheck="false">abstract</h3>
<p class="u-text u-text-grey-40 u-text-3" spellcheck="false"> Intracranial hemorrhage, bleeding that occurs inside the cranium, is a serious health problem requiring<br>rapid and often intensive medical treatment. Such a condition is traditionally diagnosed by highly<br>trained specialists analyzing a Computed tomography (CT) scan of the patient and identifying the<br>location and type of hemorrhage if one exists. When a patient shows acute neurological symptoms such<br>as severe headache or loss of consciousness, highly trained specialists review medical images of the<br>patient’s cranium to look for the presence, location and type of hemorrhage. The process is complicated<br>and often time consuming as these scans are essentially a combination of multiple X-Ray scans<br>processed by a computer. To reduce delays which lead to deaths we propose a solution to this problem<br>based on a Machine Vision based approach to automate the detection of intracranial hemorrhaging. We<br>employ a modified version of ResNet 34 trained on a dataset provided by Radiological Society of North<br>America. These convolutional networks are then stacked using a sequence model (such as a recurrent<br>neural network) to preserve temporal information. The system can detect acute intracranial hemorrhage<br>and its subtypes with inference times less than 1 second and accuracy greater than 85%. We aim to<br>provide automated, faster and more accurate diagnosis using computer vision and deep residual learning<br>frameworks for image recognition and classification
</p>
<a href="#" class="u-border-2 u-border-white u-link u-no-underline u-text-body-alt-color u-link-1">read more</a>
</div>
</section>
<section class="u-clearfix u-color-scheme-u10 u-color-style-multicolor-1 u-palette-3-base u-section-3" id="carousel_f130">
<div class="u-clearfix u-sheet u-valign-middle u-sheet-1">
<div class="u-clearfix u-expanded-width u-layout-wrap u-layout-wrap-1">
<div class="u-layout">
<div class="u-layout-row">
<div class="u-container-style u-layout-cell u-left-cell u-size-20 u-layout-cell-1">
<div class="u-container-layout u-container-layout-1">
<h5 class="u-custom-font u-font-roboto u-text u-text-1"> Problem Statement</h5>
<h1 class="u-custom-font u-font-roboto u-text u-text-2" spellcheck="false">fast accurate inference<br>
</h1>
</div>
</div>
<div class="u-container-style u-layout-cell u-size-20 u-layout-cell-2">
<div class="u-container-layout u-valign-top u-container-layout-2">
<p class="u-text u-text-3" spellcheck="false"> The bottleneck , caused by human limitations is lessened by adding the deep learning model.<br>The time spent by the specialist is reduced due to the information provided by the model.<br>Summary reports can be shared digitally for verification as well in extreme cases
</p>
</div>
</div>
<div class="u-align-right u-container-style u-layout-cell u-right-cell u-size-20 u-layout-cell-3">
<div class="u-container-layout u-valign-top u-container-layout-3">
<p class="u-text u-text-4" spellcheck="false"> The raw data never leaves the institutions, which not only<br>adds privacy and reinforce patient confidentiality but also<br>prevents large data transfers on the network.
</p>
<h2 class="u-custom-font u-font-roboto u-text u-text-5" spellcheck="false">privacy</h2>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<section class="u-align-center u-clearfix u-section-4" id="carousel_1d0e">
<div class="u-clearfix u-sheet u-valign-middle u-sheet-1">
<div class="u-clearfix u-gutter-20 u-layout-wrap u-layout-wrap-1">
<div class="u-gutter-0 u-layout">
<div class="u-layout-row">
<div class="u-size-40 u-size-60-md">
<div class="u-layout-col">
<div class="u-size-20">
<div class="u-layout-row">
<div class="u-align-justify u-black u-container-style u-layout-cell u-left-cell u-size-15-md u-size-15-sm u-size-15-xs u-size-16-xl u-size-18-lg u-layout-cell-1">
<div class="u-container-layout u-valign-middle u-container-layout-1">
<h3 class="u-text u-text-1">ai </h3>
</div>
</div>
<div class="u-container-style u-layout-cell u-size-42-lg u-size-44-xl u-size-45-md u-size-45-sm u-size-45-xs u-layout-cell-2">
<div class="u-container-layout u-valign-top u-container-layout-2">
<p class="u-text u-text-2" spellcheck="false">INTRACRANIAL HEMORRHAGE DETECTION</p>
</div>
</div>
</div>
</div>
<div class="u-size-40">
<div class="u-layout-row">
<div class="u-align-left u-container-style u-layout-cell u-left-cell u-size-60 u-layout-cell-3">
<div class="u-container-layout u-container-layout-3">
<div class="u-border-1 u-border-grey-dark-1 u-expanded-width u-line u-line-horizontal u-line-1"></div>
<h1 class="u-text u-text-3" spellcheck="false">the model</h1>
<p class="u-text u-text-palette-2-dark-2 u-text-4" spellcheck="false"> Our work here aims to reduce the computational<br>load of the model via pruning and repurposing the architecture<br>to create a more practical model for inference at even remote<br>locations with limited computation power.<br>
<br> We employ a modified version of DenseNet121 trained<br>on a dataset provided by the Radiological Society of North<br>America. These convolutional networks are then stacked using<br>a sequence model (such as a recurrent neural network) With<br>GRUs to preserve temporal information. The system can<br>detect acute intracranial hemorrhage and its subtypes with<br>accuracy greater than 92%. We aim to provide automated,<br>faster, and more accurate diagnoses using computer vision<br>and deep learning frameworks for image recognition and<br>classification.
</p>
<a href="#" class="u-border-3 u-border-black u-link u-no-underline u-text-black u-link-1">learn more</a>
</div>
</div>
</div>
</div>
</div>
</div>
<div class="u-size-20 u-size-60-md">
<div class="u-layout-col">
<div class="u-container-style u-layout-cell u-right-cell u-size-40 u-layout-cell-4">
<div class="u-container-layout u-valign-middle u-container-layout-4">
<img class="u-image u-image-1" src="images/1.png" data-image-width="581" data-image-height="737">
</div>
</div>
<div class="u-container-style u-layout-cell u-right-cell u-size-20 u-layout-cell-5">
<div class="u-container-layout u-container-layout-5">
<img class="u-expanded-width u-image u-image-2" src="images/fed.png" data-image-width="374" data-image-height="193">
</div>
</div>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<section class="u-align-center u-black u-border-2 u-border-grey-dark-1 u-clearfix u-section-5" id="carousel_d406">
<div class="u-clearfix u-sheet u-valign-middle u-sheet-1">
<h2 class="u-text u-text-1" spellcheck="false">MRI -> INFERENCE</h2>
<p class="u-text u-text-2" spellcheck="false">Our AI-powered model can be integrated with hospital networks to provide lightening fast inferences to save lives</p>
<div class="u-clearfix u-expanded-width u-gutter-0 u-layout-wrap u-layout-wrap-1">
<div class="u-gutter-0 u-layout">
<div class="u-layout-col">
<div class="u-size-30">
<div class="u-layout-row">
<div class="u-align-center u-container-style u-image u-layout-cell u-left-cell u-size-20 u-image-1" data-image-width="1280" data-image-height="1280">
<div class="u-container-layout u-valign-middle u-container-layout-1"></div>
</div>
<div class="u-container-style u-image u-layout-cell u-size-20 u-image-2" data-image-width="1280" data-image-height="1280">
<div class="u-container-layout u-valign-middle u-container-layout-2"></div>
</div>
<div class="u-container-style u-image u-layout-cell u-right-cell u-size-20 u-image-3" data-image-width="630" data-image-height="630">
<div class="u-container-layout u-valign-middle u-container-layout-3"></div>
</div>
</div>
</div>
<div class="u-size-30">
<div class="u-layout-row">
<div class="u-container-style u-image u-layout-cell u-left-cell u-size-20 u-image-4" data-image-width="630" data-image-height="630">
<div class="u-container-layout u-valign-middle u-container-layout-4"></div>
</div>
<div class="u-container-style u-image u-layout-cell u-size-20 u-image-5" data-image-width="630" data-image-height="630">
<div class="u-container-layout u-valign-middle u-container-layout-5"></div>
</div>
<div class="u-container-style u-image u-layout-cell u-right-cell u-size-20 u-image-6" data-image-width="630" data-image-height="630">
<div class="u-container-layout u-valign-middle u-container-layout-6"></div>
</div>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<section class="u-clearfix u-section-6" id="carousel_df45">
<div class="u-expanded-width u-palette-3-base u-shape u-shape-rectangle u-shape-1"></div>
<div class="u-clearfix u-gutter-30 u-layout-wrap u-layout-wrap-1">
<div class="u-gutter-0 u-layout">
<div class="u-layout-row">
<div class="u-size-20 u-size-30-md">
<div class="u-layout-col">
<div class="u-container-style u-image u-layout-cell u-left-cell u-size-30 u-image-1" data-image-width="1280" data-image-height="853">
<div class="u-container-layout u-valign-middle u-container-layout-1"></div>
</div>
<div class="u-container-style u-layout-cell u-left-cell u-size-30 u-layout-cell-2">
<div class="u-container-layout u-container-layout-2">
<h4 class="u-text u-text-1" spellcheck="false">edge computing</h4>
<p class="u-text u-text-2" spellcheck="false"> Our work here aims to reduce the computational<br>load of the model via pruning and repurposing the architecture<br>to create a more practical model for inference at even remote<br>locations with limited computation power.
</p>
</div>
</div>
</div>
</div>
<div class="u-size-20 u-size-30-md">
<div class="u-layout-col">
<div class="u-container-style u-layout-cell u-size-60 u-layout-cell-3">
<div class="u-container-layout u-valign-middle">
<img class="u-image u-image-2" src="images/ge362781010bb53433f8e81e6aa51c3a47f71465cd376c06f0554c2e3d9f60da591e1650b86bba1f22dd58512e777c957ee1addd8817955cfa75182972bc33c4a_1280.png" data-image-width="1280" data-image-height="969">
</div>
</div>
</div>
</div>
<div class="u-size-20 u-size-60-md">
<div class="u-layout-col">
<div class="u-container-style u-layout-cell u-right-cell u-size-30 u-layout-cell-4">
<div class="u-container-layout u-container-layout-4">
<h4 class="u-text u-text-3" spellcheck="false"> Federated Learning</h4>
<p class="u-text u-text-4" spellcheck="false"> The raw data never leaves the institutions, which not only<br>adds privacy and reinforce patient confidentiality but also<br>prevents large data transfers on the network
</p>
</div>
</div>
<div class="u-container-style u-image u-layout-cell u-right-cell u-size-30 u-image-3" data-image-width="970" data-image-height="545">
<div class="u-container-layout u-valign-middle u-container-layout-5"></div>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<section class="u-clearfix u-valign-middle u-section-7" id="carousel_5f44">
<div class="u-clearfix u-expanded-width u-layout-wrap u-layout-wrap-1">
<div class="u-layout">
<div class="u-layout-row">
<div class="u-container-style u-image u-layout-cell u-left-cell u-size-40 u-image-1" data-image-width="1280" data-image-height="853">
<div class="u-container-layout"></div>
</div>
<div class="u-container-style u-image u-layout-cell u-right-cell u-size-20 u-image-2" data-image-width="1280" data-image-height="1089">
<div class="u-container-layout"></div>
</div>
</div>
</div>
</div>
</section>
<section class="u-clearfix u-white u-section-8" id="carousel_76c6">
<img src="images/1_QealTEloaI3XZfI6hkSetQ.png" alt="" class="u-expanded-width u-image u-image-default u-image-1" data-image-width="868" data-image-height="488">
<div class="u-clearfix u-disable-padding u-gutter-0 u-layout-wrap u-layout-wrap-1">
<div class="u-gutter-0 u-layout">
<div class="u-layout-row">
<div class="u-size-36 u-size-60-md">
<div class="u-layout-col">
<div class="u-size-40">
<div class="u-layout-row">
<div class="u-align-left u-container-style u-layout-cell u-left-cell u-size-60 u-layout-cell-1">
<div class="u-container-layout u-valign-middle u-container-layout-1">
<h1 class="u-heading-font u-text u-text-1" spellcheck="false">RESULTS / COnCLUSIONS</h1>
</div>
</div>
</div>
</div>
<div class="u-size-20">
<div class="u-layout-row">
<div class="u-align-left u-container-style u-layout-cell u-left-cell u-palette-3-base u-size-30 u-layout-cell-2">
<div class="u-container-layout u-valign-top u-container-layout-2">
<h2 class="u-heading-font u-text u-text-2" spellcheck="false">BEST IN CLASS ACCURACY</h2>
</div>
</div>
<div class="u-align-left u-black u-container-style u-layout-cell u-size-30 u-layout-cell-3">
<div class="u-container-layout u-valign-top u-container-layout-3">
<h2 class="u-heading-font u-text u-text-3" spellcheck="false">lowest power consuption</h2>
</div>
</div>
</div>
</div>
</div>
</div>
<div class="u-size-24 u-size-60-md">
<div class="u-layout-row">
<div class="u-align-left u-container-style u-layout-cell u-right-cell u-size-60 u-layout-cell-4">
<div class="u-container-layout u-container-layout-4">
<div class="u-border-3 u-border-black u-line u-line-horizontal u-line-1"></div>
</div>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<section class="u-align-center u-clearfix u-image u-shading u-section-9" src="" data-image-width="1280" data-image-height="832" id="sec-be65">
<div class="u-clearfix u-sheet u-valign-middle u-sheet-1">
<h1 class="u-text u-text-default u-title u-text-1" spellcheck="false">Challenges</h1>
<p class="u-large-text u-text u-text-default u-text-variant u-text-2"> The data from the hospitals can be extremely skewed,<br>depending on the specialty of the doctors in question.<br>This works in our favor as it allows us to take advantage<br>of the federated passes and better personal inference<br>scenarios.
</p>
<a href="#" class="u-btn u-button-style u-palette-2-base u-btn-1">Read More</a>
</div>
</section>
<section class="u-clearfix u-section-10" id="sec-c81e">
<div class="u-clearfix u-sheet u-sheet-1">
<div class="u-clearfix u-expanded-width u-layout-wrap u-layout-wrap-1">
<div class="u-layout">
<div class="u-layout-row">
<div class="u-container-style u-image u-layout-cell u-size-30 u-image-1" data-image-width="1280" data-image-height="733">
<div class="u-container-layout u-container-layout-1"></div>
</div>
<div class="u-align-center u-container-style u-layout-cell u-size-30 u-layout-cell-2">
<div class="u-container-layout u-valign-middle u-container-layout-2">
<h2 class="u-text u-text-default u-text-1" spellcheck="false">Contact Us</h2>
<p class="u-text u-text-2" spellcheck="false">to learn more about our research please contact tkrsh.srv.09@gmail.com</p>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<footer class="u-align-center u-clearfix u-footer u-grey-80 u-footer" id="sec-5a76"><div class="u-clearfix u-sheet u-sheet-1">
<p class="u-small-text u-text u-text-variant u-text-1" spellcheck="false">All rights reserved @tkrsh.com</p>
</div></footer>
</body>
</html>
Datasets
Models
