nih-gov/www.ncbi.nlm.nih.gov/research/trialgpt/faq/index.html

488 lines
No EOL
22 KiB
HTML
Raw Permalink Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

<!DOCTYPE html>
<html lang="en" >
<head >
<meta charset="UTF-8">
<meta http-equiv="X-UA-Compatible" content="IE=edge">
<!-- Mobile properties -->
<meta name="HandheldFriendly" content="True">
<meta name="MobileOptimized" content="320">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<!-- Stylesheets -->
<link href="/research/trialgpt/static/django_uswds/uswds/css/uswds.css" rel="stylesheet" />
<link rel="stylesheet" href="/research/trialgpt/static/nwds/css/nwds.css" />
<link rel="stylesheet" href="/research/trialgpt/static/nwds/css/header.css" />
<link rel="stylesheet" href="/research/trialgpt/static/nwds/css/footer.css" />
<link rel="stylesheet" href="/research/trialgpt/static/nwds/css/form.css" />
<link rel="stylesheet" href="https://maxcdn.bootstrapcdn.com/font-awesome/4.7.0/css/font-awesome.min.css"/>
<link rel="stylesheet" href="/research/trialgpt/static/main/css/styles.css"/>
<title>TrialGPT - NLM/NIH</title>
<!-- Favicons -->
<link rel="shortcut icon" type="image/ico" href="https://www.ncbi.nlm.nih.gov/coreutils/nwds/img/favicons/favicon.ico" />
<link rel="icon" type="image/png" href="https://www.ncbi.nlm.nih.gov/coreutils/nwds/img/favicons/favicon.png" />
<!-- 192x192, as recommended for Android
http://updates.html5rocks.com/2014/11/Support-for-theme-color-in-Chrome-39-for-Android
-->
<link rel="icon" type="image/png" sizes="192x192" href="https://www.ncbi.nlm.nih.gov/coreutils/nwds/img/favicons/favicon-192.png" />
<!-- 57x57 (precomposed) for iPhone 3GS, pre-2011 iPod Touch and older Android devices -->
<link rel="apple-touch-icon-precomposed" href="https://www.ncbi.nlm.nih.gov/coreutils/nwds/img/favicons/favicon-57.png">
<!-- 72x72 (precomposed) for 1st generation iPad, iPad 2 and iPad mini -->
<link rel="apple-touch-icon-precomposed" sizes="72x72" href="https://www.ncbi.nlm.nih.gov/coreutils/nwds/img/favicons/favicon-72.png">
<!-- 114x114 (precomposed) for iPhone 4, 4S, 5 and post-2011 iPod Touch -->
<link rel="apple-touch-icon-precomposed" sizes="114x114" href="https://www.ncbi.nlm.nih.gov/coreutils/nwds/img/favicons/favicon-114.png">
<!-- 144x144 (precomposed) for iPad 3rd and 4th generation -->
<link rel="apple-touch-icon-precomposed" sizes="144x144" href="https://www.ncbi.nlm.nih.gov/coreutils/nwds/img/favicons/favicon-144.png">
<meta name="robots" content="index, follow"/>
<meta name="google-site-verification" content="iCMo85IH9lnJfVT9AlenEwbwc1BAkJToy4Gy1UgOa8Y"/>
<meta name="description"
content="TrialGPT is an AI powered tool for matching patients to clinical trials">
<!-- Logging params: Pinger defaults -->
<meta name="ncbi_app" content="cbb-research-trialgpt" />
<meta name="ncbi_pdid" content="faq" />
<meta name="ncbi_phid" content="939BC1CF9DCD1EB5000039B5A13400F0.1.m_1" />
<meta name="ncbi_page" content="qa"/>
<meta name="ncbi_op" content="page_view"/>
<meta name="description" content="Django project trialgpt"/>
</head>
<body >
<a class="usa-skipnav" href="#main_content">Skip to main page content</a>
<!-- ========== BEGIN HEADER ========== -->
<section class="usa-banner">
<div class="usa-accordion">
<header class="usa-banner-header">
<div class="usa-grid usa-banner-inner">
<img src="https://www.ncbi.nlm.nih.gov/coreutils/uswds/img/favicons/favicon-57.png" alt="U.S. flag" />
<p>An official website of the United States government</p>
<button
class="usa-accordion-button usa-banner-button"
aria-expanded="false"
aria-controls="gov-banner-top"
>
<span class="usa-banner-button-text">Here's how you know</span>
</button>
</div>
</header>
<div
class="usa-banner-content usa-grid usa-accordion-content"
id="gov-banner-top"
>
<div class="usa-banner-guidance-gov usa-width-one-half">
<img
class="usa-banner-icon usa-media_block-img"
src="https://www.ncbi.nlm.nih.gov/coreutils/uswds/img/icon-dot-gov.svg"
alt="Dot gov"
/>
<div class="usa-media_block-body">
<p>
<strong>The .gov means its official.</strong>
<br />
Federal government websites often end in .gov or .mil. Before
sharing sensitive information, make sure youre on a federal
government site.
</p>
</div>
</div>
<div class="usa-banner-guidance-ssl usa-width-one-half">
<img
class="usa-banner-icon usa-media_block-img"
src="https://www.ncbi.nlm.nih.gov/coreutils/uswds/img/icon-https.svg"
alt="Https"
/>
<div class="usa-media_block-body">
<p>
<strong>The site is secure.</strong>
<br />
The <strong>https://</strong> ensures that you are connecting to the
official website and that any information you provide is encrypted
and transmitted securely.
</p>
</div>
</div>
</div>
</div>
</section>
<div class="usa-overlay"></div>
<header class="ncbi-header" role="banner" data-section="Header">
<div class="usa-grid">
<div class="usa-width-one-whole">
<div class="ncbi-header__logo">
<a href="https://www.nlm.nih.gov/" class="logo" aria-label="NLM Logo" data-ga-action="click_image" data-ga-label="NIH NLM Logo" target="_blank">
<img src="https://assets.nlm.nih.gov/uswds/img/NLM_White.png" alt="NIH NLM Logo" />
<!-- <img src="https://www.ncbi.nlm.nih.gov/coreutils/nwds/img/logos/AgencyLogo.svg" alt="NIH NLM Logo" /> -->
</a>
</div>
<div class="ncbi-header__account">
<a id="account_login" href="/research/trialgpt/login/" class="usa-button header-button" style="display:none" data-ga-action="open_menu" data-ga-label="account_menu">Log in</a>
</div>
<!-- <div class="ncbi-popup-anchor">
<div class="ncbi-popup account-popup" id="account_popup" aria-hidden="true">
<div class="ncbi-popup-head">
<button class="ncbi-close-button" data-ga-action="close_menu" data-ga-label="account_menu"><span class="fa fa-times"></span><span class="usa-sr-only">Close</span></button>
<h4>Account</h4>
</div>
<div class="account-user-info">
Logged in as:<br/>
<b><span class="username" id="uname_long">username</span></b>
</div>
<div class="account-links">
<ul class="usa-unstyled-list">
<li><a id="account_myncbi" href="/myncbi/" class="set-base-url" data-ga-action="click_menu_item" data-ga-label="account_myncbi">Dashboard</a></li>
<li><a id="account_pubs" href="/myncbi/collections/bibliography/" class="set-base-url" data-ga-action="click_menu_item" data-ga-label="account_pubs">Publications</a></li>
<li><a id="account_settings" href="/account/settings/" class="set-base-url" data-ga-action="click_menu_item" data-ga-label="account_settings">Account settings</a></li>
<li><a id="account_logout" href="/account/signout/" class="set-base-url" data-ga-action="click_menu_item" data-ga-label="account_logout">Log out</a></li>
</ul>
</div>
</div>
</div> -->
</div>
</div>
</header>
<!-- <div role="navigation" aria-label="access keys">
<a id="nws_header_accesskey_0" href="https://www.ncbi.nlm.nih.gov/guide/browsers/#ncbi_accesskeys" class="usa-sr-only" accesskey="0" tabindex="-1">Access keys</a>
<a id="nws_header_accesskey_1" href="https://www.ncbi.nlm.nih.gov" class="usa-sr-only" accesskey="1" tabindex="-1">NCBI Homepage</a>
<a id="nws_header_accesskey_2" href="/myncbi/" class="set-base-url usa-sr-only" accesskey="2" tabindex="-1">MyNCBI Homepage</a>
<a id="nws_header_accesskey_3" href="#maincontent" class="usa-sr-only" accesskey="3" tabindex="-1">Main Content</a>
<a id="nws_header_accesskey_4" href="#" class="usa-sr-only" accesskey="4" tabindex="-1">Main Navigation</a>
</div> -->
<section data-section="Alerts">
<div class="ncbi-alerts-placeholder"></div>
</section>
<!-- ========== END HEADER ========== -->
<div class="nav usa-header usa-header--basic">
<div class="usa-nav-container nav-container">
<div class="usa-navbar ">
<button class="usa-menu-btn">Menu</button>
</div>
<nav class="usa-nav" aria-label="Primary navigation">
<button class="usa-nav-close"><img src="https://assets.nlm.nih.gov/uswds/img/close.svg" alt="close"></button>
<ul class="usa-nav-primary usa-accordion">
<li class="usa-nav-primary-item">
<a class="usa-nav-link" href="/research/trialgpt/" title="Home" aria-label="Home">
<span>Home</span>
</a>
</li>
<!-- <li class="usa-nav-primary-item">
<a class="usa-nav-link" href="/research/trialgpt/news/" title="Press Release" aria-label="Press Release">
<span>Press Release</span>
</a>
</li> -->
<li class="usa-nav-primary-item">
<a class="usa-nav-link" href="/research/trialgpt/faq/" title="FAQ" aria-label="FAQ">
<span>FAQ</span>
</a>
</li>
<li class="usa-nav-primary-item">
<a class="usa-nav-link" href="https://github.com/ncbi-nlp/TrialGPT" title="Code & Data" aria-label="Code & Data" target="_blank">
<span>Code & Data</span>
</a>
</li>
<li class="usa-nav-primary-item">
<a class="usa-nav-link" href="/research/trialgpt/about/" title="About" aria-label="About">
<span>About</span>
</a>
</li>
</ul>
</nav>
</div>
</div>
<main id="main_content" accesskey="3">
<section class="usa-section">
<div class="usa-grid">
<h1>TrialGPT FAQ</h1>
<h2>What is TrialGPT?</h2>
<div class="publication faq">
TrialGPT is an end-to-end framework designed for patient-to-trial matching, leveraging large language models (LLMs).
It consists of three main modules: TrialGPT-Retrieval for filtering and retrieving relevant clinical trials,
TrialGPT-Matching for determining patient eligibility against trial criteria with truthful explanations,
and TrialGPT-Ranking for scoring and ranking the trials based on patient suitability. Given a patient summary,
TrialGPT can return a list of clinical trial recommendations from a large collection of candidates,
based on the patients eligibility.
</div>
<h2>How does TrialGPT work?</h2>
<div class="publication faq">
TrialGPT operates through three sequential components:
<ul>
<li>
<strong>TrialGPT-Retrieval:</strong> This module uses LLMs to generate keywords from patient summaries to
filter out the most irrelevant clinical trials from an initial, large pool using a combination of lexical and
semantic retrieval methods.
</li>
<li>
<strong>TrialGPT-Matching:</strong> Analyzes patient-trial pairs on a criterion-by-criterion basis,
providing explanations of patient eligibility and locating relevant sentences in patient notes for each criterion.
</li>
<li>
<strong>TrialGPT-Ranking:</strong> Aggregates criterion-level predictions to compute trial-level scores,
which are used to rank the clinical trials by the extent of patient eligibility.
</li>
</ul>
</div>
<h2>Who are targeted users of TrialGPT?</h2>
<div class="publication faq">
The targeted users of TrialGPT include clinical trial recruiters, research teams involved in trial matching,
and healthcare professionals who facilitate patient recruitment. It is designed to assist domain experts by
enhancing efficiency and reducing time spent on patient-trial matching.
</div>
<h2>Is TrialGPT also for patients?</h2>
<div class="publication faq">
While TrialGPT can empower individual patients to identify potential clinical trials they may qualify for,
its primary design focuses on supporting clinical and research staff to facilitate patient matching.
Patients might access its benefits through healthcare intermediaries at this point.
</div>
<h2>Is the TrialGPT software freely available?</h2>
<div class="publication faq">
Yes, the TrialGPT framework is publicly accessible. The code and related resources can be found on
its GitHub repository:
<a href="https://github.com/ncbi-nlp/TrialGPT" target="_blank">https://github.com/ncbi-nlp/TrialGPT</a>.
</div>
<h2>Do you have a web-based system for TrialGPT?</h2>
<div class="publication faq">
No. Although one can potentially use the official code to deploy a web-based system for TrialGPT,
we do not provide an official implementation.
</div>
<h2>Who built TrialGPT?</h2>
<div class="publication faq">
TrialGPT is developed by researchers from the Division of Intramural Research,
National Library of Medicine, National Institutes of Health.
</div>
<h2>How is TrialGPT different from other similar systems or tools?</h2>
<div class="publication faq">
Unlike traditional embedding-based methods such as DeepEnroll <a href="#ref1">[1]</a>
and COMPOSE <a href="#ref2">[2]</a> which require massive training data and are not explainable, TrialGPT can perform
zero-shot and explainable patient-to-trial matching without requiring any training data.
Unlike structuring-based methods such as Criteria2Query <a href="#ref3">[3]</a>, TrialGPT does not depend on
the ontology mapping of clinical trial criteria, which is a non-trivial task itself.
<br><br>
<strong>References</strong>
<ol>
<li id="ref1">
Zhang X, Xiao C, Glass LM, et al. DeepEnroll: Patient-Trial Matching with Deep Embedding and Entailment Prediction.
<a href="https://doi.org/10.1145/3366423.3380181" target="_blank">
Proceedings of The Web Conference 2020.
New York, NY, USA: Association for Computing Machinery 2020:102937
</a>
</li>
<li id="ref2">
Gao J, Xiao C, Glass LM, et al. COMPOSE: Cross-Modal Pseudo-Siamese Network for Patient Trial Matching.
<a href="https://dl.acm.org/doi/10.1145/3394486.3403123" target="_blank">
Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining.
New York, NY, USA: Association for Computing Machinery 2020:80312
</a>
</li>
<li id="ref3">
Fang Y, Idnay B, Sun Y, et al. Combining human and machine intelligence for clinical trial eligibility querying.
<a href="https://doi.org/10.1093/jamia/ocac051" target="_blank">
Journal of the American Medical Informatics Association. 2022;29:116171. doi: 10.1093/jamia/ocac051
</a>
</li>
</ol>
</div>
<h2>How to cite TrialGPT?</h2>
<div class="publication faq">
Please cite TrialGPT as: <br><br>
Qiao Jin, Zifeng Wang, Charalampos S. Floudas, Fangyuan Chen,
Changlin Gong, Dara Bracken-Clarke, Elisabetta Xue, Yifan Yang, Jimeng Sun, Zhiyong Lu.
Matching Patients to Clinical Trials with Large Language Models.
<a href="https://www.nature.com/articles/s41467-024-53081-z" target="_blank" rel="noreferrer">
Nat Commun. 2024;15:9074. doi: 10.1038/s41467-024-53081-z
</a><br><br>
</div>
<h2>I heard you are working on TrialGPT 2.0. What are your goals for that?</h2>
<div class="publication faq">
Given the promising benchmarking results, the research team was recently selected for The
<a href="https://oir.nih.gov/sourcebook/awards-fellowships-grant-opportunities/directors-challenge-innovation-award-program/2024-directors-challenge-awards#2024-directors-lu" target="_blank">
Directors Challenge Innovation Award
</a>
to further assess the models performance and effectiveness in real-world clinical settings.
The researchers anticipate that this work could make clinical trial recruitment more effective
and help reduce barriers to participation for patients seeking to join clinical research.
</div>
</div>
</section>
</main>
<!-- ========== BEGIN FOOTER ========== -->
<footer>
<section class="container-fluid bg-primary pt-5" style="padding-bottom: 2.0rem;">
<div class="usa-grid pt-5">
<div class="row mt-3">
<div class="col-lg-3 col-12">
<p><a class="text-white" href="https://www.nlm.nih.gov/socialmedia/index.html">Connect with NLM</a></p>
<ul class="social_media add-list-reset">
<li class="margin-right-05"><a href="https://www.facebook.com/nationallibraryofmedicine"><img class="bg-secondary" src="https://www.nlm.nih.gov/images/facebook.svg" alt="Facebook"></a></li>
<li class="margin-right-05"><a title="External link: please review our privacy policy." href="https://www.linkedin.com/company/national-library-of-medicine-nlm/"><img class="bg-secondary" src="//www.nlm.nih.gov/images/linkedin.svg" alt="LinkedIn"></a></li>
<li class="margin-right-05"><a title="External link: please review our privacy policy." href="https://twitter.com/NLM_NIH"><img src="https://www.nlm.nih.gov/images/twitter.svg" class="bg-secondary" style="padding: 7px;" alt="Twitter"></a></li>
<li class="margin-right-05"><a title="External link: please review our privacy policy." href="https://www.youtube.com/user/NLMNIH"><img src="//www.nlm.nih.gov/images/youtube.svg" class="bg-secondary" alt="You Tube"></a></li>
<li class="margin-right-05"><a title="External link: please review our privacy policy." href="https://public.govdelivery.com/accounts/USNLMOCPL/subscriber/new?preferences=true"><img src="//www.nlm.nih.gov/images/mail.svg" class=" bg-secondary" alt="Government Delivery"></a></li>
</ul>
</div>
<div class="col-lg-3 col-12">
<p class="address_footer text-white">National Library of Medicine<br>
<a href="https://www.google.com/maps/place/8600+Rockville+Pike,+Bethesda,+MD+20894/@38.9959508,-77.101021,17z/data=!3m1!4b1!4m5!3m4!1s0x89b7c95e25765ddb:0x19156f88b27635b8!8m2!3d38.9959508!4d-77.0988323"
class="text-white" target="_blank" rel="noopener noreferrer">8600 Rockville Pike<br>
Bethesda, MD 20894</a></p>
</div>
<div class="col-lg-3 col-12 centered-lg">
<p><a href="https://www.nlm.nih.gov/web_policies.html" class="text-white">Web Policies</a><br>
<a href="https://www.nih.gov/institutes-nih/nih-office-director/office-communications-public-liaison/freedom-information-act-office"
class="text-white">FOIA</a><br>
<a href="https://www.hhs.gov/vulnerability-disclosure-policy/index.html" class="text-white" id="vdp">HHS Vulnerability Disclosure</a></p>
</div>
<div class="col-lg-3 col-12 centered-lg">
<p><a class="supportLink text-white" href="https://support.nlm.nih.gov/">Help</a><br>
<a href="https://www.nlm.nih.gov/accessibility.html" class="text-white">Accessibility</a><br>
<a href="https://www.nlm.nih.gov/careers/careers.html" class="text-white">Careers</a></p>
</div>
</div>
<div class="row">
<div class="col-lg-12 centered-lg">
<nav class="bottom-links">
<ul class="mt-3">
<li>
<a class="text-white" href="//www.nlm.nih.gov/">NLM</a>
</li>
<li>
<a class="text-white"
href="https://www.nih.gov/">NIH</a>
</li>
<li>
<a class="text-white" href="https://www.hhs.gov/">HHS</a>
</li>
<li>
<a
class="text-white" href="https://www.usa.gov/">USA.gov</a>
</li>
</ul>
</nav>
</div>
</div>
</div>
</section>
</footer>
<!-- ========== END FOOTER ========== -->
<!-- javascript to inject NWDS meta tags. Note: value of nwds_version is updated by "npm version" command -->
<!-- <script type="text/javascript">
var nwds_version = "1.2.3";
var meta_nwds_ver = document.createElement('meta');
meta_nwds_ver.name = 'ncbi_nwds_ver';
meta_nwds_ver.content = nwds_version;
document.getElementsByTagName('head')[0].appendChild(meta_nwds_ver);
var meta_nwds = document.createElement('meta');
meta_nwds.name = 'ncbi_nwds';
meta_nwds.content = 'yes';
document.getElementsByTagName('head')[0].appendChild(meta_nwds);
var alertsUrl = "/core/alerts/alerts.js";
if (typeof ncbiBaseUrl !== 'undefined') {
alertsUrl = ncbiBaseUrl + alertsUrl;
}
</script> -->
<!-- JavaScript -->
<script src="/research/trialgpt/static/django_uswds/uswds/js/uswds.js"></script>
<script src="https://code.jquery.com/jquery-3.5.0.min.js"
integrity="sha256-xNzN2a4ltkB44Mc/Jz3pT4iU1cmeR0FkXs4pru/JxaQ="
crossorigin="anonymous">
</script>
<script>
var fallbackJquery = "/research/trialgpt/static/base/js/jquery-3.5.0.min.js";
window.jQuery || document.write("<script src=" + fallbackJquery + ">\x3C/script>")
</script>
<script src="/research/trialgpt/static/nwds/js/nwds.js" type="text/javascript"> </script>
<script src="/research/trialgpt/static/nwds/js/header.js" type="text/javascript"> </script>
<script src="/research/trialgpt/static/nwds/js/ncbipopup.js" type="text/javascript"> </script>
<script src="/research/trialgpt/static/nwds/js/ncbiclearbutton.js" type="text/javascript"> </script>
<script src="/research/trialgpt/static/nwds/js/override-uswds.js" type="text/javascript"> </script>
<script src="/research/trialgpt/static/nwds/js/ncbifeedback.js" type="text/javascript"> </script>
<script type="text/javascript" src="https://www.ncbi.nlm.nih.gov/core/pinger/pinger.js"> </script>
</body>
</html>