-
Notifications
You must be signed in to change notification settings - Fork 0
/
cognitivecomputingsystems.html
218 lines (202 loc) · 19.9 KB
/
cognitivecomputingsystems.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
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
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1, shrink-to-fit=no">
<meta name="description" content="">
<meta name="author" content="">
<link rel="icon" href="bootstrap/favicon.ico">
<title>Cognitive Computing Systems</title>
<!-- Bootstrap core CSS -->
<link href="bootstrap/dist/css/bootstrap.min.css" rel="stylesheet">
<!-- Custom styles for this template -->
<link href="jumbotron.css" rel="stylesheet">
</head>
<body>
<nav class="navbar navbar-toggleable-md navbar-inverse fixed-top bg-inverse">
<button class="navbar-toggler navbar-toggler-right" type="button" data-toggle="collapse" data-target="#navbarsExampleDefault" aria-controls="navbarsExampleDefault" aria-expanded="false" aria-label="Toggle navigation">
<span class="navbar-toggler-icon"></span>
</button>
<a class="navbar-brand" href="index.html">AI in Healthcare</a>
<div class="collapse navbar-collapse justify-content-end" id="navbarsExampleDefault">
<ul class="navbar-nav">
<li class="nav-item active">
<a class="nav-link" href="index.html">HOME <span class="sr-only">(current)</span></a>
</li>
<li class="nav-item">
<a class="nav-link" href="about.html">ABOUT</a>
</li>
<li class="nav-item dropdown">
<a class="nav-link dropdown-toggle" id="dropdown01" data-toggle="dropdown" aria-haspopup="true" aria-expanded="false">SMART ASSISTANTS</a>
<div class="dropdown-menu" aria-labelledby="dropdown01">
<a class="dropdown-item" href="cognitivecomputingsystems.html">Cognitive computing systems</a>
<a class="dropdown-item" href="healthapps.html">Health monitoring & diagnosis apps</a>
</div>
</li>
<li class="nav-item dropdown">
<a class="nav-link dropdown-toggle" id="dropdown01" data-toggle="dropdown" aria-haspopup="true" aria-expanded="false">INTELLIGENT ROBOTS</a>
<div class="dropdown-menu" aria-labelledby="dropdown01">
<a class="dropdown-item" href="typesofrobots.html">Types of robots</a>
<a class="dropdown-item" href="roboticsurgery.html">Robotic surgery</a>
</div>
</li>
<li class="nav-item dropdown">
<a class="nav-link dropdown-toggle" id="dropdown01" data-toggle="dropdown" aria-haspopup="true" aria-expanded="false">HEALTHCARE GADGETS</a>
<div class="dropdown-menu" aria-labelledby="dropdown01">
<a class="dropdown-item" href="wearabletechnology.html">Wearable technology</a>
<a class="dropdown-item" href="gadgetsforyourphone.html">Gadgets for your phone</a>
</div>
</li>
</ul>
</div>
</nav>
<div id="main-nav" class="jumbotron jumbotron-cognitivecomputingsystem">
<div class="container">
<h1 class="display-3 text-center">Cognitive Computing Systems</h1>
<p class="text-center">Cognitive computing can be thought of as the the simulation of human thought processes within computers and has the potential to have a profound impact on the world.</p>
<center><a class="scroll-link btn btn-info" href="#" data-id="ibmwatson">IBM Watson</a>
<a class="scroll-link btn btn-info" href="#" data-id="atomnet">AtomNet</a>
<a class="scroll-link btn btn-info" href="#" data-id="medicalsieve">Medical Sieve</a></center>
</div>
</div>
<div id="ibmwatson">
<div class="container-fluid bg-1 text-center">
<h2 class="text-center" id="ibmwatson">IBM Watson</h2>
<center><img src="images/watson.png" class="img-responsive" alt="ibm watson" style="width: 30%; height: 30%"></center>
<p>
<h4 class="text-center">What is IBM Waton?</h3>
<p>Watson is a supercomputer built by IBM to analyse enormous amounts of data that can then be used by researchers. It combines artificial intelligence and complex analytical software to be capable of answering potentially any question you may have. Watson process information at a rate of 80 teraflops and accesses 90 servers storing 200 million pages of information in order to replicate a human brain. Within healthcare, Watson is capable of searching through huge sets of patient records and medical reports and uses hypothesis generation and evidence-based learning to help doctors make better decisions. <sup>[1]</sup> </p>
<button type="button" class="btn btn-outline-info" data-toggle="collapse" data-target="#demo11">Read more</button>
<div id="demo11" class="panel-collapse collapse">
<p>
<h4 class="text-center">What is Watson's healthcare hierarchy?</h3>
<p>Watson's healthcare hierarchy is described as below:</p>
<p>The first level begins with assistance that the computer would be able to give researchers when treating and diagnosing patients. For example, a medical professional would describe symptoms and add them to the system. Watson would then use natural language processing to identify key information, particularly any medical terminology. <sup> [2] </sup> </p>
<p>At the second level, the system combines patient data, current medications and all available data from tests to form hypotheses. Then specific conditions will be mapped to the patient based on treatment guidelines, online medical data and articles. </p>
<p>The third level is composed of decisions on potential diagnoses for the patient, based upon the data inputted earlier and the resulting patterns that have been analysed.</p>
<p>The highest tier is known as discovery and is where Watson searches through huge numbers of papers on and relating to the condition being studied.<br/><a href="#" class="scroll-top back-to-top">↑</a></p>
<center><a href="#ibmrefs" class="btn btn-outline-info" data-toggle="collapse">References</a></center>
<div id="ibmrefs" class="collapse">
<div class="list-group">
<a href="http://www-05.ibm.com/innovation/uk/watson/watson_in_healthcare.shtml" class="list-group-item list-group-item-action">1. How Watson can address healthcare challenges. (accessed February 2017).</a>
<a href="http://www.techrepublic.com/article/ibm-watsons-impressive-healthcare-analytics-capabilities-continue-to-evolve/" class="list-group-item list-group-item-action">2. Mary Shacklett. IBM Watson's impressive healthcare analytics capabilities continue to evolve. (accessed February 2017).</a>
<a href="http://www.atomhibs.com" class="list-group-item list-group-item-action">Photo: atomhibs.com </a>
</div>
</div>
</div>
</div>
</div>
<div id="atomnet">
<div class="container-fluid bg-2 text-center">
<h2 class="text-center" id="atomnet">AtomNet</h2>
<center><img src="images/atomnet.png" class="img-responsive" alt="atomnet" style="width: 50%; height: 50%"></center>
<p>
<h4 class="text-center">What is AtomNet?</h3>
<p>AtomNet is the first deep convolutional neural network for structure-based, “rational” drug design that includes information about the target to make predicitions. It uses artificial intelligence and machine learning to help discover new medicines by predicting the bioactivity of small molecules for drug discovery.</p>
<button type="button" class="btn btn-outline-info" data-toggle="collapse" data-target="#demo12">Read more</button>
<div id="demo12" class="panel-collapse collapse">
<p>
<h4 class="text-center">How are new drugs discovered?</h3>
<p>When a protein plays a crucial role in a disease, it is called a "target", for example, proteins that grow tumours and proteins that help viruses infect human cells. The goal in drug research is to create molecules (called "ligands") that strongly interact with these targets and reduce their effect.</p>
<p>Lock-oriented techniques or, in other words “structure-based” algorithms, identify the structure of the target protein to guide their predictions. This approach works for completely new targets and is therefore very useful. It has led to the development of a variety of software packages, such as Dock and Glide. The accuracy is the main limitation of these technologies, as a high rate of false-positives can be found in these techniques which is why many researchers have remained skeptical about their usefulness.
</p>
<p> Key-oriented techniques or “ligand-based” algorithms have also been explored. Many examples of ligands that are already known to bind to a target are considered and even better ligands can be predicted from that data by technologies such as ROCS (a powerful virtual screening tool capable of quickly identifying active compounds via ashape comparison) and LINGOs (a text-based molecular similarity search). The main limitation is that these technologies require researchers to have already discovered at least some ligands for a target. However, there are very few known ligands for unsolved and challenging drug discovery targets. In this case, these methods are expected to produce poor predictions.</p>
<h4 class="text-center">How does AtomNet discover new drugs?</h3>
<p>AtomNet does this through a structure-based, rational, highly accurate drug design system. Candidate molecules for new drug discovery targets can be predicted by this system and these predictions have a high chance of proving correct and providing researchers with what they need from a virtual drug discovery method.</p>
<p>AtomNet is excellent at understanding complicated concepts by dividing each problem into smaller pieces of information. This property is the reason why the world’s best results for image classification and deep-rooted problems are produced by convolutional networks. For example, a convolutional model can learn to recognise faces by first learning a set of basic features in an image, such as edges. Then, the model can identify different body parts such as eyes, ears and mouth by combining the edges. Finally, the model can learn to recognise faces by combining those parts. AtomNet does something similar except with chemical structures. <sup>[1]</sup> </p>
<center><img src="images/cnn.jpg" class="img-responsive" alt="cnn" style="width: 30%; height: 30%"></center>
<p>
<h4 class="text-center">What are convolutional neural networks?</h3>
<p>Convolutional neural networks (CNNs) are a type of feed-forward neural network which tries to mimic the pattern of neuron connections inside the visual cortex of an animal. CNNs have been used in the field of computer vision for a long time but have CNNs gained more popularity recenyly in a diverse range of applications, from natural language processing to medical image analysis. The power of a CNN is due to its deep architecture, which enables a CNN to extract a list of perceptive features at different levels of abstraction. </p>
<p>Similarly, AtomNet might learn that proteins and ligands are made up of a variety of specialised chemical structures. This would be an exciting result because it suggest that AtomNet was learning fundamental concepts in organic chemistry by itself - no human ever taught AtomNet the building blocks of organic chemistry. AtomNet discovered them itself by studying vast quantities of target and ligand data. The patterns it independently observed are so fascinatingly foundational that they are studied in academic courses. <sup>[2]</sup><br/><a href="#" class="scroll-top back-to-top">↑</a></p>
<center><a href="#atomnetrefs" class="btn btn-outline-info" data-toggle="collapse">References</a></center>
<p>
<div id="atomnetrefs" class="collapse">
<div class="list-group">
<a href="http://www.atomwise.com/introducing-atomnet/" class="list-group-item list-group-item-action">1. Izhar Wallach, Michael Dzamba, Abraham Heifets. Introducing AtomNet – Drug design with convolutional neural networks. (accessed February 2017).</a>
<a href="http://ieeexplore.ieee.org/document/7426826/" class="list-group-item list-group-item-action text-left">2. Nima Tajbakhsh, Jae Y. Shin, Suryakanth R. Gurudu, R. Todd Hurst, Christopher B. Kendall, Michael B. Gotway, Jianming Liang. Convolutional Neural Networks for Medical Image Analysis: Full Training or Fine Tuning? (accessed February 2017).</a>
<a href="http://www.atomwise.com/introducing-atomnet/" class="list-group-item list-group-item-action">Photo: atomwise.com</a>
<a href="https://www.theodysseyonline.com/the-future-with-neural-networks" class="list-group-item list-group-item-action">Photo: theodysseyonline.com</a>
</div>
</div>
</div>
</div>
</div>
<div id="medicalsieve">
<div class="container-fluid bg-1 text-center">
<h2 class="text-center" id="google">Medical Sieve</h2>
<p>
<center><img src="images/medicalsieve.jpg" class="img-responsive" alt="Medical Sieve" style="width: 40%; height: 40%"></center>
<p> </p>
<h4 class="text-center">What is Medical Sieve</h3>
<p>Medical Sieve is a long-term project which aims to construct a cognitive assistant with high-level multimodal analytics and reasoning capabilities able to assist in making clinical decisions in the fields of radiology and cardiology. The sieve will have a fundamental understanding of diseases and how they can be interpreted in modalities such as X-ray, ultrasounds, CT, MRI etc. </p>
<button type="button" class="btn btn-outline-info" data-toggle="collapse" data-target="#demo4">Read more</button>
<div id="demo4" class="panel-collapse collapse">
<p>
<h4 class="text-center">What is the aim of Medical Sieve?</h3>
<p>The aim of this project is to build a sieve that filters vital clinical imaging and diagnostic information to produce summaries that are anomaly-driven and suggestions that greatly decrease the viewing load of clinicians without having a negative impact on diagnosis. Based on statistics, one common issue among radiologists is eye fatigue due to the enourmous number of visual checks they carry out on images every day. For example, around 200 cases may be considered by a radiologist in an emergency room per day, and this number can reach up to 3000 images, particularly in lower body CT angiography. <sup>[1]</sup> </p>
<center><img src="images/mediclsieve.jpg" class="img-responsive" alt="Medical Sieve" style="width: 40%; height: 40%"></center>
<p> </p>
<h4 class="text-center">How does Medical Sieve work?</h3>
<p>Medical Sieve is an image-guided informatics system that acts as a medical sieve, filtering the necessary clinical information needed for physicians to diagnose their patients and find the best treatment plans. Clinical data about patients is gathered from various enterprise systems in hospitals including EMR (electronic medical record) which is the digital version of traditional paper-based medical record for an individual, and radiology/cardiology PACs (picture archiving and communication system) using HL7 and DICOM adapters.</p>
<p>Digital Imaging and Communications in Medicine (DICOM) is used all over the world to exchange medical images, escalate the connectivity of radiological devices and develop modern radiological imaging. Before the DICOM standard became broadly selected, each manufacturer had its own image format and communication protocol. This expansion and growth of formats and protocols made it very difficult to create third-party software to manage or study medical content, or to connect hardware devices from different manufacturers. <sup>[2]</sup> </p>
<p>Health Level-7 or HL7 is a set of international standards which allows clinical and administrative data transaction between software applications. These applications are used by different healthcare providers. Open Systems Interconnection model (OSI) characterises communication functions with no regard for their internal structure and technology. This model has 7 layers and HL7 standards focus on the application layer, the 7th layer of OSI, and specifies the shared protocol and interface methods in a communication network. <sup>[3]</sup> </p>
<p>Then, advanced medical text and image processing pattern recognition and machine learning techniques guided by sophisticated clinical knowledge will be used to process clinical data about the patient and to obtain valid and worthwhile summaries showing the anomalies. Finally, advanced summaries of imaging studies will be created to represent important anomalies that are identified in various point of views.<br/><a href="#" class="scroll-top back-to-top">↑</a></p>
<center><a href="#msrefs" class="btn btn-outline-info" data-toggle="collapse">References</a></center>
<p>
<div id="msrefs" class="collapse">
<div class="list-group">
<a href="http://researcher.watson.ibm.com/researcher/view_group.php?id=4384" class="list-group-item list-group-item-action">1. Medical Sieve, IBM Research. (accessed February 2017).</a>
<a href="https://docs.oracle.com/database/121/IMDCM/ch_intro.htm#IMDCM1100" class="list-group-item list-group-item-action">2. Introduction to Oracle Multimedia DICOM. (accessed February 2017).</a>
<a href="https://www.edifecs.com/downloads/0298_SB-XE-SB-HL7-Adaptor.pdf" class="list-group-item list-group-item-action">3. Health Level 7 (HL7) Adapter. (accessed February 2017).</a>
<a href="https://www.fastcompany.com/3064902/ibm-goes-west-a-73-year-long-saga-from-punch-cards-to-watson" class="list-group-item list-group-item-action">Photo: fastcompany.com</a>
<a href="http://researcher.watson.ibm.com/researcher/view_group_subpage.php?id=5490" class="list-group-item list-group-item-action">Photo: ibm.com</a>
</div>
</div>
</div>
</div> <!-- /container -->
</div>
<!-- Bootstrap core JavaScript
================================================== -->
<!-- Placed at the end of the document so the pages load faster -->
<script src="https://code.jquery.com/jquery-3.1.1.slim.min.js" integrity="sha384-A7FZj7v+d/sdmMqp/nOQwliLvUsJfDHW+k9Omg/a/EheAdgtzNs3hpfag6Ed950n" crossorigin="anonymous"></script>
<script>window.jQuery || document.write('<script src="bootstrap/assets/js/vendor/jquery.min.js"><\/script>')</script>
<script src="https://cdnjs.cloudflare.com/ajax/libs/tether/1.4.0/js/tether.min.js" integrity="sha384-DztdAPBWPRXSA/3eYEEUWrWCy7G5KFbe8fFjk5JAIxUYHKkDx6Qin1DkWx51bBrb" crossorigin="anonymous"></script>
<script src="bootstrap/dist/js/bootstrap.min.js"></script>
<!-- IE10 viewport hack for Surface/desktop Windows 8 bug -->
<script src="bootstrap/assets/js/ie10-viewport-bug-workaround.js"></script>
<!-- Scrolling function -->
<script src="js/jquery-3.1.1.min.js" type="text/javascript"></script>
<script src="js/bootstrap.min.js" type="text/javascript"></script>
<script type="text/javascript">
$(document).ready(function() {
// navigation click actions
$('.scroll-link').on('click', function(event){
event.preventDefault();
var sectionID = $(this).attr("data-id");
scrollToID('#' + sectionID, 1000);
});
// scroll to top action
$('.scroll-top').on('click', function(event) {
event.preventDefault();
$('html, body').animate({scrollTop:0}, 'slow');
});
});
// scroll function
function scrollToID(id, speed){
var offSet = 50;
var targetOffset = $(id).offset().top - offSet;
var mainNav = $('#main-nav');
$('html,body').animate({scrollTop:targetOffset}, speed);
if (mainNav.hasClass("open")) {
mainNav.css("height", "1px").removeClass("in").addClass("collapse");
mainNav.removeClass("open");
}
}
if (typeof console === "undefined") {
console = {
log: function() { }
};
}
</script>
</body>
</html>