Uncertainty aware domain incremental learning for cross domain depression detection
Abstract Deep learning techniques have demonstrated significant promise for detecting Major Depressive Disorder (MDD) from textual data but they still face limitations in real-world scenarios. Specifically, given the limited data availability, some efforts have resorted to aggregating data from diff...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2025-07-01
|
| Series: | Scientific Reports |
| Subjects: | |
| Online Access: | https://doi.org/10.1038/s41598-025-10917-y |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849766961475485696 |
|---|---|
| author | Zita Lifelo Jianguo Ding Huansheng Ning Sahraoui Dhelim |
| author_facet | Zita Lifelo Jianguo Ding Huansheng Ning Sahraoui Dhelim |
| author_sort | Zita Lifelo |
| collection | DOAJ |
| description | Abstract Deep learning techniques have demonstrated significant promise for detecting Major Depressive Disorder (MDD) from textual data but they still face limitations in real-world scenarios. Specifically, given the limited data availability, some efforts have resorted to aggregating data from different domains to expand the data volume. However, these approaches face critical challenges, including data privacy, domain gaps, class imbalance, and uncertainty arising from both the data and the model. To overcome these challenges, we propose an Uncertainty-Aware Domain Incremental Learning framework for Cross-Domain Depression Detection (UDIL-DD), integrating Uncertainty-guided Adaptive Class Threshold Learning (UACTL) and Data-Free Domain Alignment (DFDA). Specifically, our UACTL module measures the discrepancy between predictions across sequential domains and learns adaptive thresholds tailored to each class, incorporating predictive uncertainty to enhance robustness. Subsequently, the DFDA module leverages domain-similar samples identified by UACTL to approximate historical feature distributions without accessing previous domain data, effectively addressing catastrophic forgetting. To validate the effectiveness of the proposed method, we conduct extensive experiments on four benchmark MDD datasets-CMDC, DIAC-WoZ, MODMA and EATD confirming the effectiveness of our method’s potential for reliable depression detection in real-world clinical scenarios. |
| format | Article |
| id | doaj-art-995a8e89333c450da2d95b10a950f799 |
| institution | DOAJ |
| issn | 2045-2322 |
| language | English |
| publishDate | 2025-07-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | Scientific Reports |
| spelling | doaj-art-995a8e89333c450da2d95b10a950f7992025-08-20T03:04:25ZengNature PortfolioScientific Reports2045-23222025-07-0115111710.1038/s41598-025-10917-yUncertainty aware domain incremental learning for cross domain depression detectionZita Lifelo0Jianguo Ding1Huansheng Ning2Sahraoui Dhelim3School of Computer and Communications Engineering, University of Science and Technology BeijingDepartment of Computer Science, Blekinge Institute of TechnologySchool of Computer and Communications Engineering, University of Science and Technology BeijingSchool of Computing, Dublin City UniversityAbstract Deep learning techniques have demonstrated significant promise for detecting Major Depressive Disorder (MDD) from textual data but they still face limitations in real-world scenarios. Specifically, given the limited data availability, some efforts have resorted to aggregating data from different domains to expand the data volume. However, these approaches face critical challenges, including data privacy, domain gaps, class imbalance, and uncertainty arising from both the data and the model. To overcome these challenges, we propose an Uncertainty-Aware Domain Incremental Learning framework for Cross-Domain Depression Detection (UDIL-DD), integrating Uncertainty-guided Adaptive Class Threshold Learning (UACTL) and Data-Free Domain Alignment (DFDA). Specifically, our UACTL module measures the discrepancy between predictions across sequential domains and learns adaptive thresholds tailored to each class, incorporating predictive uncertainty to enhance robustness. Subsequently, the DFDA module leverages domain-similar samples identified by UACTL to approximate historical feature distributions without accessing previous domain data, effectively addressing catastrophic forgetting. To validate the effectiveness of the proposed method, we conduct extensive experiments on four benchmark MDD datasets-CMDC, DIAC-WoZ, MODMA and EATD confirming the effectiveness of our method’s potential for reliable depression detection in real-world clinical scenarios.https://doi.org/10.1038/s41598-025-10917-yMajor depressive disorderData-free domain alignmentIncremental learningClass imbalanceUncertainty estimation |
| spellingShingle | Zita Lifelo Jianguo Ding Huansheng Ning Sahraoui Dhelim Uncertainty aware domain incremental learning for cross domain depression detection Scientific Reports Major depressive disorder Data-free domain alignment Incremental learning Class imbalance Uncertainty estimation |
| title | Uncertainty aware domain incremental learning for cross domain depression detection |
| title_full | Uncertainty aware domain incremental learning for cross domain depression detection |
| title_fullStr | Uncertainty aware domain incremental learning for cross domain depression detection |
| title_full_unstemmed | Uncertainty aware domain incremental learning for cross domain depression detection |
| title_short | Uncertainty aware domain incremental learning for cross domain depression detection |
| title_sort | uncertainty aware domain incremental learning for cross domain depression detection |
| topic | Major depressive disorder Data-free domain alignment Incremental learning Class imbalance Uncertainty estimation |
| url | https://doi.org/10.1038/s41598-025-10917-y |
| work_keys_str_mv | AT zitalifelo uncertaintyawaredomainincrementallearningforcrossdomaindepressiondetection AT jianguoding uncertaintyawaredomainincrementallearningforcrossdomaindepressiondetection AT huanshengning uncertaintyawaredomainincrementallearningforcrossdomaindepressiondetection AT sahraouidhelim uncertaintyawaredomainincrementallearningforcrossdomaindepressiondetection |