-
Notifications
You must be signed in to change notification settings - Fork 0
/
project_generic4.html
136 lines (120 loc) · 6.38 KB
/
project_generic4.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
<!DOCTYPE HTML>
<!--
Stellar by HTML5 UP
html5up.net | @ajlkn
Free for personal and commercial use under the CCA 3.0 license (html5up.net/license)
-->
<html>
<head>
<title>Generic - Stellar by HTML5 UP</title>
<meta charset="utf-8"/>
<meta name="viewport" content="width=device-width, initial-scale=1, user-scalable=no"/>
<link rel="stylesheet" href="assets/css/main.css"/>
<noscript>
<link rel="stylesheet" href="assets/css/noscript.css"/>
</noscript>
</head>
<body class="is-preload">
<div id="wrapper">
<!-- Header -->
<header id="header">
<a href="index.html">
<img style= "padding-top: 0px; padding-bottom: 0px" src="images/TripartiteLogo2.png" alt="" width=600px>
<h1 style="letter-spacing: 7px; margin-bottom: 10px; font-size:5em; text-shadow:1px 1px 10px #fff, 3px 3px 3px #ccc">CRADLE</h1>
<p style="margin-bottom: 10px; font-size:1em">The Cardiovascular and Radiologic Deep Learning Environment</p>
</a>
</header>
<!-- Nav -->
<nav id="nav">
<ul>
<li><a href="index.html#intro">Introduction</a></li>
<li><a href="index.html#first">Projects</a></li>
<li><a href="index.html#second">People</a></li>
<li><a href="index.html#cta">Jobs</a></li>
<li><a href="index.html#footer">Contact</a></li>
</ul>
</nav>
<!-- Main -->
<div id="main">
<!-- Content -->
<section id="content" class="main">
<!-- <span class="image main"><img src="images/pic04.jpg" alt=""/></span>-->
<header class="major">
<h2>ValveNet</h2>
</header>
<p>The routine 12-lead electrocardiogram (ECG) is ubiquitous, inexpensive, and highly useful in the assessment of
heart disease. However, it is limited in its sensitivity and specificity in the detection of serious structural
heart diseases. In this project, we theorized that deep
learning analysis of the ECG could enable the detection of patients with three valvular diseases – aortic
stenosis, aortic regurgitation, and mitral regurgitation – through improved analysis of this feature-rich
data.</p>
<p>We first developed a research database that was comprised of all patients at our medical center who had
undergone an ECG and an echocardiogram. This yielded 77,163 patients who had undergone an ECG within a year
prior to an echocardiogram. This data was split into train, validation, and test sets.</p>
<span class="image story"><img src="ValveNet/Inclusion_Criteria.png" alt="" /></span>
<p> A convolutional neural network was trained using the raw ECG waveform (30,000 data points per ECG) using
multiple ResNet blocks with fusion of tabular data (like age) to detect patients with moderate or severe aortic
stenosis, aortic regurgitation, mitral regurgitation, and a combination of any of the three valvular lesions.
Multiple model architectures were trialed, with the final model providing a high level of accuracy in the
validation set at a low computational cost.</p>
<span class="image story"><img src="ValveNet/Model_Diagram.png" alt=""/></span>
<p>We found that the model could accurately detect each of the three valvular lesions alone and in combination. We
are currently deploying this model in a prospective clinical study to identify patients with undiagnosed
valvular heart disease at Columbia University Irving Medical Center.</p>
<span class="image story"><img src="ValveNet/Model_Performance.jpg" alt=""/></span>
For more details, see our accompanying paper,
<blockquote>
<p><a
href="https://www.jacc.org/doi/10.1016/j.jacc.2022.05.029?utm_medium=email_newsletter&utm_source=jacc&utm_campaign=toc&utm_content=20220801#mmc1"><strong>Deep
Learning Electrocardiographic Analysis for Detection of Left-Sided Valvular Heart Disease</strong></a><br/>
Pierre Elias & Timothy J. Poterucha, et al. <br>Journal of the American College of Cardiology, August 1, 2022.
</p>
</blockquote>
<ul class="actions special">
<li><a href="index.html#first" class="button">Return Home</a></li>
</ul>
</section>
</div>
<!-- Footer -->
<footer id="footer">
<section>
<h2>Interested in starting your own career in ML for healthcare?</h2>
<p>Our white paper walks you through all the best resources we've found to learn about programming, machine learning, and medical AI research!</p>
<ul class="actions">
<li><a href="https://docs.google.com/document/d/1ZXB7j0DNVtBlEVHTQT2VO1EwoWuCYk4BpVeO6O-JNY0/edit#heading=h.49io06jmrp45" class="button">Check It Out</a></li>
</ul>
</section>
<section>
<h2>Contact Us</h2>
<dl class="alt">
<dt>Location</dt>
<dd>Department of Biomedical Informatics (PH20)</dd>
<dt>Address</dt>
<dd>622 W 168th St • New York, NY 10032 • USA</dd>
<!-- <dt>Phone</dt>-->
<!-- <dd>(000) 000-0000 x 0000</dd>-->
<dt>Email</dt>
<dd>pae2115 at dbmi dot columbia dot edu</dd>
</dl>
<ul class="icons">
<li><a href="https://twitter.com/PierreEliasMD" class="icon brands fa-twitter alt"><span class="label">Twitter</span></a></li>
<!-- <li><a href="#" class="icon brands fa-facebook-f alt"><span class="label">Facebook</span></a></li>-->
<!-- <li><a href="#" class="icon brands fa-instagram alt"><span class="label">Instagram</span></a></li>-->
<li><a href="https://github.com/PierreElias/" class="icon brands fa-github alt"><span class="label">GitHub</span></a></li>
<!-- <li><a href="#" class="icon brands fa-dribbble alt"><span class="label">Dribbble</span></a></li>-->
</ul>
</section>
<p class="copyright">© 2022 CRADLE @ Columbia University
<p class="copyright"> <img style= "padding-top: 0px; padding-bottom: 0px" src="images/TripartiteLogo2.png" alt="" width=600px></p>
</footer>
</div>
<!-- Scripts -->
<script src="assets/js/jquery.min.js"></script>
<script src="assets/js/jquery.scrollex.min.js"></script>
<script src="assets/js/jquery.scrolly.min.js"></script>
<script src="assets/js/browser.min.js"></script>
<script src="assets/js/breakpoints.min.js"></script>
<script src="assets/js/util.js"></script>
<script src="assets/js/main.js"></script>
</body>
</html>