Dear Editor,
We read with great interest the editorial titled “Clinical applicability of machine learning in family medicine” recently published in the
Korean Journal of Family Medicine [
1]. This editorial provides an overview of the potential applications of artificial intelligence (AI) and machine learning (ML) techniques in osteoporosis and fracture risk prediction. We appreciate the authors for highlighting this important and rapidly evolving area of research.
As noted in the editorial, Kang et al. [
2] presented a significant advancement in leveraging ML algorithms to predict fracture risk. By utilizing a large-scale cohort study that incorporates various risk factors such as bone mineral density and trabecular bone score data, their ML models demonstrated promising performance in predicting osteoporotic fractures compared with conventional risk assessment tools.
While the editorial provides a comprehensive overview, we suggest additional perspectives and future directions in this field.
1. Multimodal data integration: Although Kang et al. [
2] focused on clinical and imaging data, further improvements in fracture prediction accuracy can be achieved by integrating multimodal data sources. These include genetic factors, lifestyle, environmental exposures, and biomarkers. ML algorithms identify complex patterns and interactions across diverse data types, potentially uncovering novel risk factors and enhancing predictive performance [
3].
2. Explainable AI: As ML models grow in complexity, interpretability becomes crucial for clinical acceptance and trust. Developing explainable AI (XAI) techniques for transparent and understandable rationales in model predictions is invaluable for both clinicians and patients [
4]. This could facilitate shared decision-making and personalized treatment strategies.
3. Longitudinal modeling: Kang et al. [
2] focused on crosssectional data, incorporating longitudinal data into ML models to capture temporal changes in risk factors and disease progression. This enables more accurate and dynamic risk predictions, potentially providing timely interventions and personalized monitoring strategies [
5].
4. Transfer learning and domain adaptation: As previously mentioned, the availability and quality of data pose challenges for ML models. Transfer learning and domain adaptation techniques leverage knowledge from related domains or larger datasets, potentially improving the model performance and generalizability, even with limited domain-specific data [
6].
To demonstrate the potential of ML in this field, we propose a conceptual figure depicting a multimodal fracture risk prediction pipeline (
Figure 1). This pipeline integrates diverse data sources, including clinical, imaging, genetic, and lifestyle factors using an ensemble of ML models. These models are designed to capture complex interactions and temporal patterns while incorporating XAI techniques for interpretability. The resulting risk predictions can inform personalized treatment strategies and monitoring plans, ultimately improving patient outcomes.
In conclusion, this editorial highlighted the promising applications of AI and ML in osteoporosis and fracture risk prediction. As this field advances, embracing multimodal data integration, XAI, longitudinal modeling, and transfer learning techniques will be crucial in unlocking the full potential of these powerful tools. Collaborative efforts among clinicians, researchers, and data scientists are essential for translating these advances into clinical practice, improving patient care and outcomes.
Thank you for the opportunity to contribute to this discussion. We eagerly anticipate further developments and insights into this dynamic field.
Notes
Figure. 1.Multimodal fracture risk prediction pipeline. This figure outlines a comprehensive pipeline for fracture-risk prediction utilizing machine learning (ML) techniques and multimodal data integration. It begins with the aggregation of diverse data sources including clinical, imaging, genetic, and lifestyle data. These data were preprocessed to extract the relevant features and normalize them for ML modeling. An ensemble of ML models supported by explainable artificial intelligence (XAI) techniques generates personalized fracture risk predictions for patients. These predictions, along with interpretable model explanations, inform the development of personalized treatment and monitoring plans. The feedback loop continuously integrates patient outcomes and new data, facilitating iterative model refinement and improvements. Overall, the pipeline visualized the seamless integration of data-driven approaches to enhance fracture risk assessment and patient care. BMD, bone mineral density; TBS, trabecular bone score; SNPs, single nucleotide polymorphisms; SHAP, Shapley Additive Explanations; LIME, Local Interpretable Model-Agnostic Explanations.
References
- 1. Lee J. Clinical applicability of machine learning in family medicine. Korean J Fam Med 2024;45:123-4.
- 2. Kang SJ, Kim MJ, Hur YI, Haam JH, Kim YS. Application of machine learning algorithms to predict osteoporotic fractures in women. Korean J Fam Med 2024;45:144-8.
- 3. Vora LK, Gholap AD, Jetha K, Thakur RR, Solanki HK, Chavda VP. Artificial intelligence in pharmaceutical technology and drug delivery design. Pharmaceutics 2023;15:1916.
- 4. Metta C, Beretta A, Pellungrini R, Rinzivillo S, Giannotti F. Towards transparent healthcare: advancing local explanation methods in explainable artificial intelligence. Bioengineering (Basel) 2024;11:369.
- 5. Wang Y, Liu L, Wang C. Trends in using deep learning algorithms in biomedical prediction systems. Front Neurosci 2023;17:1256351.
- 6. Gu C, Lee M. Deep transfer learning using real-world image features for medical image classification, with a case study on pneumonia Xray images. Bioengineering (Basel) 2024;11:406.
Figure & Data
Citations
Citations to this article as recorded by
- Osteocytes: master orchestrators of skeletal homeostasis, remodeling, and osteoporosis pathogenesis
Yan Wu, Donghao Gan, Zhikang Liu, Daodi Qiu, Guoqing Tan, Zhanwang Xu, Haipeng Xue
Frontiers in Cell and Developmental Biology.2025;[Epub] CrossRef - AI-driven Technologies for Wrist Fracture Prediction: A Narrative Review of Emerging Approaches
Stefania Briano, Maria Cesarina May, Giacomo Demontis, Giulia Pachera, Vittoria Mazzola, Federico Vitali, Alessandra Galuppi, Emanuela Dapelo, Andrea Zanirato, Matteo Formica
Journal of Wrist Surgery.2025;[Epub] CrossRef
