Accepted for/Published in: JMIR Medical Informatics
Date Submitted: Aug 13, 2025
Open Peer Review Period: Aug 13, 2025 - Oct 8, 2025
Date Accepted: Mar 6, 2026
(closed for review but you can still tweet)
Predicting Pediatric Urological Surgery Duration Through Multi-modal Patient-Physician Feature Fusion: A Deep Learning Framework Incorporating Clinical Text Embedding
ABSTRACT
Background:
Accurate prediction of surgical duration is critical for optimizing operating room scheduling and resource allocation. Existing models, however, exhibit limited applicability in pediatric urology due to the unique anatomical and developmental characteristics of children.
Objective:
This study aims to develop and validate a specialty-tailored prediction framework for surgery duration predication of pediatric urological surgery.
Methods:
We integrated multi-source heterogeneous data, encompassing patient demographics, surgical details, surgeon-specific features, and electronic medical record narratives, to develop a customized prediction system. Large language model techniques were employed to extract semantic representations from unstructured clinical text, while a multi-head perceptron architecture enabled the efficient fusion of structured and unstructured features. Pediatric-specific clinical variables, such as developmental stage and the severity of urinary tract malformations, were explicitly modeled to capture their impact on surgical duration.
Results:
The proposed approach achieved a mean absolute error (MAE) of 10.74 minutes and a root mean square error (RMSE) of 14.91 minutes, markedly outperforming existing methods. Comparative analyses demonstrated that the Qwen-based structured preprocessing with text embeddings provided superior feature representation, surpassing both traditional LSTM and direct Embedding-3 approaches. Feature importance analysis highlighted the lead surgeon, preoperative diagnosis, primary surgical procedure, and surgical plan as dominant predictive factors.
Conclusions:
By combining innovative feature engineering with a tailored model architecture, the proposed framework substantially improves the accuracy of surgical duration prediction in pediatric urology. These findings offer robust technical support for precision operating room scheduling and hold significant clinical value in enhancing the efficiency of surgical resource utilization.
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.