Related Experiment Video
Updated: Oct 1, 2025
14:14
Targeting Neuronal Fiber Tracts for Deep Brain Stimulation Therapy Using Interactive, Patient-Specific Models
Published on: August 12, 2018
9.0K
Neuroimaging-ITM: A Text Mining Pipeline Combining Deep Adversarial Learning with Interaction Based Topic Modeling
Jianzhuo Yan1,2, Lihong Chen1,2, Yongchuan Yu1,2
1Faculty of Information Technology, Beijing University of Technology, Beijing, 100124, China.
Neuroinformatics
|March 2, 2022
Summary
This study introduces a novel text mining pipeline to enhance FAIR neuroscience by automatically extracting neuroimaging provenance and identifying research topics, improving data sharing and reproducibility.
Area of Science:
- Neuroscience
- Data Science
- Computational Biology
Background:
- Sharing neuroimaging data is crucial for FAIR (Findable, Accessible, Interoperable, Reusable) neuroscience.
- Characterizing neuroimaging provenance is essential for understanding brain cognition, enabling replication, and fostering open cooperation.
- Existing neuroimaging text mining faces challenges like fragmented information, limited labeled data, and ambiguous topics.
Purpose of the Study:
- To propose a text mining pipeline for enabling FAIR neuroimaging studies.
- To systematically capture provenance requests and transform them into text mining tasks.
- To automatically extract neuroimaging provenance and identify research topics, especially in few-shot scenarios.
Main Methods:
- Redesigned the Brain Informatics provenance model based on the Neuroimaging Data Model (NIDM) and FAIR facets.
- Developed a neuroimaging text mining pipeline combining deep adversarial learning with interaction-based topic modeling (Neuroimaging-ITM).
- Utilized real data from PloS One for experimental validation.
Main Results:
- Neuroimaging-ITM systematically and accurately extracts provenance information from neuroimaging articles.
- Achieved high-quality research topic identification with mean F1 values for provenance extraction exceeding 0.9.
- Demonstrated superior performance compared to baseline methods, with topic coherence reaching 9.95 and KL divergence 0.96.
Conclusions:
- The proposed Neuroimaging-ITM pipeline effectively addresses challenges in neuroimaging text mining.
- Enables systematic and accurate extraction of provenance information, crucial for FAIR data principles.
- Facilitates high-quality research topic identification, supporting reproducible and collaborative neuroscience research.
