Intelligent evaluation method for inflammatory response after trans-PRK and FS-LASIK based on multimodal feature fusion
- 1Department of Ophthalmology, the First Affiliated Hospital of Henan Medical University
- 2Clinical Medical College, Shandong Second Medical University
Abstract
Assessing postoperative inflammation following trans-PRK and FS-LASIK remains challenging. This study proposes a multimodal fusion framework supported by inflammatory dynamics physical descriptors. The model uses modality-specific encoders for ocular imaging and structured clinical or biomarker sequences, uses explicit observation masks for unavailable modalities, and learns visit-level fusion weights using cross-modal attention modules. In the four de-identified retrospective data partitions evaluated in this study, the proposed fusion model increases the accuracy of the strongest baseline reported in each partition by 2.66 to 4.10% points, and shows similar improvements in F1 score and AUROC. The results indicate that intermediate multimodal fusion can support postoperative inflammation assessment. Before clinical deployment, larger external validation cohorts, calibration analyses, and direct comparisons with medical multimodal fusion methods are still required.
Related
- Comparative Evaluation of Feature Selection Algorithms for Predicting Radiation-Induced Complications in Head and Neck Cancer Using Multimodal CT and MRI Radiomics
- Development and Evaluation of Multimodal Universal CPR AI Assistance and Response Engine (U-CARE)
- Risk prediction models for macrovascular complications in type 2 diabetes mellitus: from traditional risk scores to multimodal intelligent prediction
- Multimodal molecular representation and property prediction based on structured complementary feature fusion
- Multimodal Quantitative Parameter Mapping Characterizes Low-dimensional Feature Organization and Tract-specific Signatures in White Matter
BibTeX
@article{yan2026intelligent,
title = {{Intelligent evaluation method for inflammatory response after trans-PRK and FS-LASIK based on multimodal feature fusion}},
author = {Zeyu Yan and Yan Li and Fan Zhang and Haibo Yan and Shiyang Niu and Hua Yang and Baojun Wang},
journal = {Scientific Reports},
year = {2026},
month = sep,
doi = {10.1038/s41598-026-69257-0},
url = {https://doi.org/10.1038/s41598-026-69257-0},
note = {CC BY},
}
RIS
TY - JOUR
TI - Intelligent evaluation method for inflammatory response after trans-PRK and FS-LASIK based on multimodal feature fusion
AU - Zeyu Yan
AU - Yan Li
AU - Fan Zhang
AU - Haibo Yan
AU - Shiyang Niu
AU - Hua Yang
AU - Baojun Wang
PY - 2026
DA - 2026/09/01
JO - Scientific Reports
DO - 10.1038/s41598-026-69257-0
UR - https://doi.org/10.1038/s41598-026-69257-0
AB - Assessing postoperative inflammation following trans-PRK and FS-LASIK remains challenging. This study proposes a multimodal fusion framework supported by inflammatory dynamics physical descriptors. The model uses modality-specific encoders for ocular imaging and structured clinical or biomarker sequences, uses explicit observation masks for unavailable modalities, and learns visit-level fusion weights using cross-modal attention modules. In the four de-identified retrospective data partitions evaluated in this study, the proposed fusion model increases the accuracy of the strongest baseline reported in each partition by 2.66 to 4.10% points, and shows similar improvements in F1 score and AUROC. The results indicate that intermediate multimodal fusion can support postoperative inflammation assessment. Before clinical deployment, larger external validation cohorts, calibration analyses, and direct comparisons with medical multimodal fusion methods are still required.
SN - 2045-2322
ER -