-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathdoppler.html
More file actions
372 lines (311 loc) · 16.8 KB
/
Copy pathdoppler.html
File metadata and controls
372 lines (311 loc) · 16.8 KB
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
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
<!DOCTYPE html>
<html lang="en">
<!-- ======= Head Section - DON'T CHANGE ======= -->
<head>
<meta charset="utf-8">
<meta content="width=device-width, initial-scale=1.0" name="viewport">
<title>Mitral Inflow</title>
<link href="assets/img/logo.png" rel="icon">
<!-- Google Fonts -->
<link
href="https://fonts.googleapis.com/css?family=Open+Sans:300,300i,400,400i,600,600i,700,700i|Raleway:300,300i,400,400i,500,500i,600,600i,700,700i|Poppins:300,300i,400,400i,500,500i,600,600i,700,700i"
rel="stylesheet">
<!-- Vendor CSS Files -->
<link href="assets/vendor/aos/aos.css" rel="stylesheet">
<link href="assets/vendor/bootstrap/css/bootstrap.min.css" rel="stylesheet">
<link href="assets/vendor/bootstrap-icons/bootstrap-icons.css" rel="stylesheet">
<link href="assets/vendor/boxicons/css/boxicons.min.css" rel="stylesheet">
<link href="assets/vendor/glightbox/css/glightbox.min.css" rel="stylesheet">
<link href="assets/vendor/remixicon/remixicon.css" rel="stylesheet">
<link href="assets/vendor/swiper/swiper-bundle.min.css" rel="stylesheet">
<!-- Main CSS File -->
<link href="assets/css/style.css" rel="stylesheet">
</head>
<!-- ======= Body Section ======= -->
<body>
<!-- ======= Navigation Bar - DON'T CHANGE ======= -->
<header id="header" class="fixed-top">
<div class="container d-flex align-items-center justify-content-between">
<a href="index.html" class="logo"><img src="assets/img/logo.png" alt="" class="img-fluid"></a>
<nav id="navbar" class="navbar">
<ul>
<li><a class="nav-link scrollto" href="index.html">Home</a></li>
<li><a class="nav-link scrollto" href="index.html#projects">Projects</a></li>
<li><a class="nav-link scrollto" href="index.html#team">People</a></li>
<li><a class="nav-link scrollto" href="index.html#resources">Resources</a></li>
<li><a class="nav-link scrollto" href="index.html#funding">Funding</a></li>
<li><a class="nav-link scrollto" href="index.html#partners">Partners</a></li>
<li><a class="nav-link scrollto" href="index.html#workwithus">Vacancies</a></li>
<li><a class="nav-link scrollto" href="index.html#contact">Contact</a></li>
</ul>
<i class="bi bi-list mobile-nav-toggle"></i>
</nav>
</div>
</header><!-- End Navigation Bar -->
<!-- ======= Title Section ======= -->
<section id="project-title" class="d-flex align-items-center">
<div class="container position-relative" data-aos="fade-up" data-aos-delay="100">
<div class="row">
<div class="col-12 text-left">
<h1>Automated Analysis of Mitral Inflow Doppler using Convolutional Neural Networks</h1>
</div>
</div>
</div>
</section><!-- End Title -->
<!-- Main Webpage Section -->
<main id="project-main">
<!-- ======= Project Body Section ======= -->
<section id="project-body">
<!-- ======= Project Introduction ======= -->
<div class="project-intro">
<div class="container">
<div class="row justify-content-left">
<div class="col-12">
<hr>
<h1><b>Abstract</b> </h1><br>
<h1>
Doppler echocardiography is commonly used for functional assessment of heart valves such as mitral
valve.
Currently, the measurements are made manually which is a laborious and subjective process.
We have demonstrated the feasibility of using neural networks to fully automate the process of mitral v
alve inflow measurements. Experiments show that the automated system yields comparable performance to
the experts.</h1>
<br>
</div>
</div>
<!-- <div class="text-left">
<a class="btn btn-primary" href="https://link.springer.com/article/10.1007/s10554-017-1092-4" role="button">Download the paper</a>
</div>-->
</div>
</div>
<!-- ======= Project Dataset ======= -->
<div class="project-dataset">
<div class="container project-container">
<div class="row justify-content-left d-flex flex-wrap align-items-center">
<div class="col-12 col-md-12 col-lg-6 col-xl-6">
<img class="img-fluid" src="assets/img/projects/Mitral_Inflow/dataset_sample2.png" alt="Mitral inflow example image">
</div>
<div class="col-12 col-md-12 col-lg-6 col-xl-6">
<h2>Introduction</h2>
<h3>
Transmitral Doppler echocardiography has become the primary mode of noninvasive assessment of
diastolic filling and function. The physiology of ventricular filling is reflected by the transmitral
Doppler velocity profile which is comprised of two distinct components (waves); the E-wave (early phase
ventricular filling) and the A-wave (atrial phase of ventric- ular filling)). Accurate analysis of
Doppler echocardiographic images is therefore of crucial importance. Current clinical practice requires
sonographers to manually perform velocity measurements on Doppler traces , which often results in high
intra- and inter-observer variability. Automated systems could help
standardise measurement protocol and reduce the amount of time spent on these measurements, thereby
improving clinical workflow.
</h3>
<br>
<div class="text-left">
<a class="btn btn-primary" href="#contact" role="button">Download the dataset</a>
</div>
</div>
</div>
</div>
<!-- ======= Project Architecture ======= -->
<div class="project-architecture">
<div class="container">
<hr>
<div class="row justify-content-left">
<div class="col-12">
<h2>Method</h2>
</div>
<div class="col-12 col-md-12 col-lg-6 col-xl-8">
<h3>
A random sample of 500 echocardiographic examinations of different patients was extracted from
Imperial College Healthcare NHS Trust’s echocardiogram database. The acquisition of the images was
performed by experienced echocardiographers and according to standard protocols, using ultrasound
equipment from GE and Philips manufacturers. Ethical ap- proval was obtained from the Health
Regulatory Agency (Integrated Research Application System identifier 243023). Mitral inflow Doppler
images were then annotated manually by expert cardiologists, selecting the E- and A-wave peak
coordinates. DICOM-formatted images of varying image sizes images were zero-padded to the size of
1024x1024 pixels. The cardiologist’s annotations of the images were used as the ground truth (400 for
training, and 100 for testing).
<br>
A schematic diagram of the full deep learning pipeline is illustrated in Figure 1. In order to
detect/isolate heartbeats, Faster-RCNN object-detection architecture was used. All over- lapping
bounding box predictions are processed using non-maximum suppression method. Predicted bounding boxes
with probability over 70% are then passed to a landmark detec- tion model based on architecture
proposed in (Wei et al., 2016). The predicted heatmaps are then used to extract the location of the E-
and A-waves for all the heartbeats present in the Doppler image. TensorFlow was used to implement the
models, running on Nvidia RTX3090 GPU. <br>
<br>
<br>
<i>
Shih-En Wei, Varun Ramakrishna, Takeo Kanade, and Yaser Sheikh. Convolutional pose machines. In
Proceedings of the IEEE conference on Computer Vision and Pattern Recog- nition, pages 4724–4732,
2016.
</i>
</h3>
</div>
<div class="col-12 col-md-12 col-lg-6 col-xl-4">
<br>
<figure>
<img class="img-fluid" src="assets/img/projects/Mitral_Inflow/Pipeline_new.png" alt="Entire Pipeline">
<figcaption>Figure 1: Full pipeline. a) Original Doppler image as input with no pre-processing. b)
Faster-RCNN detects/seperates every heartbeat present in the image. c) Isolated heartbeats are
cropped. d) Landmark detection model predicts heatmaps. e) E- and A-wave peak points are obtained
from the heatmaps. f) Peak velocities from all heartbeats are put back on the Doppler image.
</figcaption>
</figure>
<!-- <img class="img-fluid" src="assets/img/projects/Mitral_Inflow/Pipeline_new.png" alt="Entire Pipeline"> -->
</div>
</div>
</div>
</div>
<!-- ======= Project Implementation ======= -->
<div class="project-implementation">
<div class="container">
<div class="row justify-content-left">
<div class="col-12">
<hr>
<h2>Results</h2>
<h3>Figure 2 upper row shows Bland-Altman plots for the pool of heartbeats across all patients. Bias for
the automated measurements versus the expert measurements was less than 1 cm/s for both E- and A-wave
peak velocities. The lower row shows the same plots, but for a patient-by-patient analysis, where all
heartbeats present in a Doppler image for each patient are averaged. The standard deviation in both
beat-by-beat and patient-by-patient analyses for the A-wave was larger; indicating that this peak was
harder to detect/measure than the E-wave. The image datasets, code, and trained networks
generated/developed in this study are made freely-available for future benchmarking.</h3>
<br>
<figure>
<img class="img-fluid" src="assets/img/projects/Mitral_Inflow/BAplots.png" alt="Plots" width="800px"
height="700px">
<figcaption>
Figure 2: Bland-Altman plots for automated vs. manual measurements.
</figcaption>
</figure>
<!-- <img class="img-fluid" src="assets/img/projects/Mitral_Inflow/BAplots.png" alt="Plots" width="800px" height="700px"> -->
<!-- <div class="text-left">
<br>
<a class="btn btn-primary"
href="URL HERE"
role="button">Download the code</a>
</div> -->
</div>
</div>
</div>
</div>
<!-- ======= Contact Section ======= -->
<section id="contact" class="contact">
<div class="container">
<div class="section-title">
<hr>
<h2>Request Access to the project dataset</h2>
<h3>We aim to make the dataset available for research purposes soon.</h3>
<div class="container">
<div class="section-title">
<h3><b>If you wish to request access, please complete the form below:<b></h3>
<br>
</div>
<div class="row">
<div class="col-12">
<!-- due to GitHub being static, you need to create a form project on https://formspree.io -->
<form action="https://formspree.io/f/xvojdbaa" method="POST" role="form">
<div class="row justify-content-center align-middle">
<div class="col-6 form-group">
<input type="text" name="name" class="form-control" id="name" placeholder="Your Name"
required>
</div>
</div>
<div class="row justify-content-center align-middle">
<div class="col-6 form-group">
<input type="email" class="form-control" name="email" id="email" placeholder="Your Email"
required>
</div>
</div>
<div class="row justify-content-center align-middle">
<div class="col-6 form-group ">
<input type="text" class="form-control" name="institution" id="subject"
placeholder="Institution/Workplace" required>
</div>
</div>
<div class="row justify-content-center align-middle">
<div class="col-6 form-group ">
<input type="text" class="form-control" name="dataset" id="dataset" value="Mitral Inflow Doppler" readonly>
</div>
</div>
<div class="row justify-content-center align-middle">
<div class="col-9 form-group">
<textarea class="form-control" name="message" rows="5"
placeholder="Please let us know why you would like access to this dataset and what you intend to use it for"
required></textarea>
</div>
</div>
<br>
<div class="text-center"><button type="submit">Submit</button></div>
</form>
</div>
</div>
</div>
</div>
<div class="container">
<hr>
<div class="row justify-content-left">
<div class="col-12">
<div class="text-left">
<h1> <b> Project Team <b></h1> <br>
<h6> <a href="https://twitter.com/intsav_?lang=en-gb" target="_blank" rel="noopener noreferrer">
Jevgeni Jevsikov</a></h6>
<h6> <a href="https://elisabethlane.github.io/" target="_blank" rel="noopener noreferrer"> Elisabeth
S
Lane </a></h6>
<h6> <a href="https://www.uwl.ac.uk/staff/massoud-zolgharni" target="_blank"
rel="noopener noreferrer"> Massoud Zolgharni </a></h6>
</div>
</div>
</div>
</div>
<!--
<div class="container">
<hr>
<div class="row justify-content-left">
<div class="col-12">
<div class="text-left">
<h1> <b> References </b></h1>
<br>
<h3> <a href="https://link.springer.com/article/10.1007/s10554-017-1092-4">TDI Beat and Keypoint
Detection</a></h3>
</div>
</div>
</div>
</div> -->
</div>
</section><!-- End Contact Section -->
</section><!-- End project-body -->
</main><!-- End #main -->
<!-- ======= Footer - DONT CHANGE ======= -->
<footer id="footer">
<div class="container d-md-flex py-4">
<div class="me-md-auto text-center text-md-start">
<div class="copyright">
© Copyright <strong><span>IntSaV</span></strong>. All Rights Reserved
</div>
</div>
<div class="social-links text-center text-md-right pt-3 pt-md-0">
<a href="#" class="twitter"><i class="bx bxl-twitter"></i></a>
<a href="#" class="facebook"><i class="bx bxl-facebook"></i></a>
<a href="#" class="instagram"><i class="bx bxl-instagram"></i></a>
<a href="#" class="google-plus"><i class="bx bxl-skype"></i></a>
<a href="#" class="linkedin"><i class="bx bxl-linkedin"></i></a>
</div>
</div>
</footer><!-- End Footer -->
<div id="preloader"></div>
<a href="#" class="back-to-top d-flex align-items-center justify-content-center"><i
class="bi bi-arrow-up-short"></i></a>
<!-- Vendor JS Files -->
<script src="assets/vendor/aos/aos.js"></script>
<script src="assets/vendor/bootstrap/js/bootstrap.bundle.min.js"></script>
<script src="assets/vendor/glightbox/js/glightbox.min.js"></script>
<script src="assets/vendor/isotope-layout/isotope.pkgd.min.js"></script>
<script src="assets/vendor/php-email-form/validate.js"></script>
<script src="assets/vendor/purecounter/purecounter.js"></script>
<script src="assets/vendor/swiper/swiper-bundle.min.js"></script>
<!-- Template Main JS File -->
<script src="assets/js/main.js"></script>
</body>
</html>