T58. DEVELOPMENT OF AN AL-BASED WEB DIAGNOSTIC SYSTEM FOR PHENOTYPING PSYCHIATRIC DISORDERS
Abstract Background The technique of phenotyping psychiatric disorders with neuroimaging data (e.g. MRI, PET) allows physicians to not only better diagnose but also introduce early interventions if necessary than traditional approaches. Recently, there is an increasing interest initiating AI-based m...
Saved in:
| Published in: | Schizophrenia bulletin Vol. 45; no. Supplement_2; p. S226 |
|---|---|
| Main Authors: | , , |
| Format: | Journal Article |
| Language: | English |
| Published: |
US
Oxford University Press
09.04.2019
|
| Subjects: | |
| ISSN: | 0586-7614, 1745-1701 |
| Online Access: | Get full text |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract
Background
The technique of phenotyping psychiatric disorders with neuroimaging data (e.g. MRI, PET) allows physicians to not only better diagnose but also introduce early interventions if necessary than traditional approaches. Recently, there is an increasing interest initiating AI-based medical diagnostic applications for oncology and pathology. One distinguished explainable DNN (EDNN) framework is currently presented for identifying key structural deficits related to the known structural pathology of schizophrenia [1].
As a consequence, this project aims to develop an Al-based web diagnostic system under this latest EDNN framework for diagnosing probability of schizophrenia with 3D visualization of subjects’ neuroimaging dataset.
Methods
The Al-based web diagnostic system consists of three main components: the website, the server and the database.
The website is served up as HTML and JavaScript (js) files, both of which have become enormously popular in web development. All the data will be converted for graphical preparations and visualizations at this level through user’s local computing resources and WebGL for the graphical abilities.
The server is constructed with Node.js, a platform on Chrome’s JavaScript runtime for building scalable network applications. Node.js has been tested to yield better efficiency than PHP and Python-Web [2]. On server side, the EDNN framework is deployed to communicate with the database.
The database is the data storage for all the dataset to be viewed and to be added. For the verification of the applied EDNN framework, the diagnostic system is validated in respect of accuracy with structural brain magnetic resonance (MR) images. The structural MR images were obtained from 200 schizophrenic patients and 200 age- and sex-matched healthy control subjects recruited at Taipei Veterans General Hospital, Taiwan.
Results
Using the EDNN algorithm, our Al-based web diagnostic system achieves 80% accuracy rate in schizophrenia classification and is capable of predicting a schizophrenia probability for reference. In addition, our system can display subjects’ 3D MRI image with specifically highlighted brain voxels identified by EDNN framework, enabling users to efficiently evaluate the imaging data and to phenotype schizophrenia.
Discussion
The current developing Al-based web diagnostic system makes the EDNN framework really accessible to both scientific and clinical community. Our next step is to extend the applicability of this diagnostic system from schizophrenia to other major psychiatric disorders (e.g. schizoaffective disorder, psychotic bipolar disorder, and Alzheimer’s disease), making it a powerful and practical diagnosis tool in future medical applications. |
|---|---|
| Bibliography: | ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 14 |
| ISSN: | 0586-7614 1745-1701 |
| DOI: | 10.1093/schbul/sbz019.338 |