XNAT Clinical Analytics Platform (XCAP): An Informatics Platform for Machine Learning Applications in Clinical Research
|
|
- Melinda Dickerson
- 5 years ago
- Views:
Transcription
1 XNAT Clinical Analytics Platform (XCAP): An Informatics Platform for Machine Learning Applications in Clinical Research Daniel Marcus, PhD Associate Professor of Radiology Washington University School of Medicine
2 Disclosures This work was supported in part by funding from the National Institutes of Health (4R01EB , 1U24CA , 1P30NS ) and the McDonnell Center for Higher Brain Function. Dr. Marcus has an ownership interest in Radiologics, Inc. and may financially benefit if the company is successful in marketing its products that are related to this research. Dr. Marcus has a financial interest in White Rabbit and may financially benefit if the company is successful in marketing its products that are related to this research.
3 Overview The XNAT imaging informatics platform XNAT analytics XNAT clinical workflows Case study: Connectomes for surgical planning
4 XNAT is a feature rich Archive, manage, process, view, and share imaging and related data. open Open source Open API Free (though commercial support is available) Used by organizations around the world platform. Clinical/translational research Institutional repositories Multi-center studies Data sharing #SIIM17
5 Top 10 XNAT features User Access Control Integrated Search & Reporting Automated Analytics DICOM Integration Audit Trail #SIIM17 Bulletproof Security Electronic Signatures Case Report Forms Dashboards Extensibility
6 Top 10 XNAT features User Access Control Integrated Search & Reporting Automated Analytics DICOM Integration Audit Trail #SIIM17 Bulletproof Security Electronic Signatures Case Report Forms Dashboards Extensibility
7 The XNAT Architecture #SIIM17
8 XNAT Open Source Community #SIIM17
9 Who uses XNAT? #SIIM17 100s imaging centers, 1000s of studies rely on XNAT.
10 XNAT Analytics Our goal is to enable users to easily: package analytic routines deploy analytic routines execute analytic routines automate analytic routines share analytic routines monitor analytic routines develop new analytic routines
11 XNAT Analytics What are containers? Docker containers wrap a piece of software in a complete filesystem that contains everything needed to run: code, runtime, system tools, system libraries anything that can be installed on a server. This guarantees that the software will always run the same, regardless of its environment.
12 XNAT Analytics Step 1. Get a container image
13 XNAT Analytics Step 2. Configure analytic routine
14 XNAT Analytics Step 3. Configure data I/O
15 XNAT Analytics Step 4. Run routine
16 XNAT Analytics Step 5. XNAT manages the rest #SIIM17 NIAC Compute Clusters
17 XNAT Analytics Example Containers Data format conversion (e.g. DICOM NIFTI) Compute image statistics (e.g. image histogram) Transform image (e.g. register to atlas) Group operations (e.g. build atlas) Train deep learning model (e.g. TensorFlow integration)
18 XNAT Analytics
19 XNAT Analytics
20 XNAT Analytics
21 XNAT Analytics
22 XNAT Analytics
23 XNAT Analytics
24 XNAT Analytics
25 XNAT Analytics
26 XNAT Analytics
27 XNAT Analytics
28 XNAT Analytics Enables sharing of analytic routines Promotes reproducible research Facilitates automation Can be run on demand or on event XNAT orchestrates file I/O Execute on single objects or groups of objects
29 XNAT Clinical Workflows 1. Large patient cohorts Build data sets, train algorithms 2. Individual patient Apply algorithm to patient exam
30 Large Patient Cohorts #SIIM17
31 Large Patient Cohorts #SIIM17
32 Individual patient workflow #SIIM17
33 Individual patient workflow Query the PACS Run Container Export to PACS
34 Case Study: Connectomes The Cirrus deep neural network generates maps of individual patient brains, localizing functional brain areas like vision, motor, memory, and attention. #SIIM17 Hacker, et al, 2013
35 Case Study: Connectomes XNAT interfaces with PACS to exchange images, executes the Cirrus application, manages workflows with neuroradiologist, and generates QC metrics. Hacker, et al, 2013 #SIIM17
36 Case Study: Connectomes The brain maps generated by Cirrus are used by neurosurgeons to navigate around critical brain areas during surgical section.
37 Case Study: Connectomes
38 Case Study: Connectomes
39 Case Study: Connectomes
40 Case Study: Connectomes #SIIM17 The brain maps are loaded onto the Medtronic Stealth system for precise intraoperative mapping and surgical navigation.
41 Case Study: Connectomes
42 Thanks! Learn more at Reach out to