Comparison of Machine Learning Methods to Predict Incomplete Atypical Femoral Fracture After Bisphosphonate Use in Postmenopausal Women

Year 2023, Volume: 24 Issue: 2, 161 - 167, 30.06.2023

Sultan Turhan , Tuğba Dübektaş Canbek , Umut Canbek , Eralp Doğu

Abstract

Objective: Long-term use of bisphosphonates (BP) for treating osteoporosis may cause incomplete atypical femoral fracture. In this study, we compared the classification and risk estimation of incomplete atypical femoral fractures, which is an alternative approach to clinical risk assessment.
Materials and Methods: A data set was randomly selected from women using postmenopausal BP. We identified a class imbalance problem in the population and created a balanced structure using the density-based synthetic minority over-sampling technique. We compared machine learning algorithms and conducted a case study. Results: We solved the class imbalance problem with the density-based synthetic minority over-sampling technique and found that the random forest and adaboost methods achieved the highest performance in the classification step.
Conclusion: It is recommended to apply resampling methods in cases where there is an unbalanced class problem such as incomplete atypical femoral fracture. Ensemble methods perform better than traditional methods in this study.

Keywords

Classification, unbalanced data, disease diagnosis, orthopedics, incomplete atypical fractures

References

• 1. Fischer S, Kapinos KA, Mulcahy A, Pinto L, Hayden O, Barron R. Estimating the long-term functional burden of osteoporosis-related fractures. Osteoporos Int 2017; 28: 2843-51.
• 2. Canbek U, Hazer DB, Rosberg H, Akgün U, Canbek TD, Cömert A, et al. The Effect of Bisphosphonates on Lumbar Vertebral Disc Height. Ege Klinikleri Tıp Dergisi 2019; 57: 52-6.
• 3. Hopkins RB, Tarride JE, Leslie WD, Metge C, Lix LM, Morin S, Finlayson G, et al. Estimating the excess costs for patients with incident fractures, prevalent fractures, and nonfracture osteoporosis. Osteoporos Int 2013; 24: 581-93.
• 4. Silverman S, Christiansen C. Individualizing osteoporosis therapy. Osteoporos Int 2012; 23: 797-809.
• 5. Nishino T, Hyodo K, Matsumoto Y, Yanagisawa Y, Yoshizawa T, Yamazaki M. Surgical results of atypical femoral fractures in long-term bisphosphonate and glucocorticoid users - Relationship between fracture reduction and bone union. J. Orthop 2020; 19: 143-9.
• 6. Shane E, Burr D, Abrahamsen B, Adler RA, Brown TD, Cheung AM, et al. Atypical subtrochanteric and diaphyseal femoral fractures: second report of a task force of the American Society for Bone and Mineral Research. J Bone Miner Res 2014; 29: 1-23.
• 7. WSG on the P. and M. of Osteoporosis, Prevention and management of osteoporosis : report of a WHO scientific group. World Health Organization. Geneva PP, 2003 (Online). Available: https://apps. who.int/iris/handle/10665/42841.
• 8. Obermeyer Z, Emanuel EJ. Predicting the Future - Big Data, Machine Learning, and Clinical Medicine 2016; 375: 1216-9.
• 9. Cabitza F, Locoro A, Banfi G. Machine Learning in Orthopedics: A Literature Review. 2018; 6.
• 10. Canbek U, Akgun U, Soylemez D, Canbek TD, Aydogan NH. Incomplete atypical femoral fractures after bisphosphonate use in postmenopausal women. J Orthop Surg 2019; 27: 1-10.
• 11. Mitchell TM, Machine Learning. 1st. New York: McGraw-Hill; 1997:414.
• 12. Öztürk H, Türe M, Kıylıoğlu N, Kurt Ömürlü İ. The Comparison of Different Dimension Reduction and Classification Methods in Electroencephalogram Signals. Meandros Med Dent J 2018; 19: 336-44.
• 13. Kruse C, Eiken P, Vestergaard P. Machine Learning Principles Can Improve Hip Fracture Prediction. Calcif Tissue Int 2017; 100: 348- 60.
• 14. Engels A, Reber KC, Lindlbauer I, Rapp K, Büchele G, Klenk J, et al. Osteoporotic hip fracture prediction from risk factors available in administrative claims data-A machine learning approach. PLoS One 2020; 15: 1-14.
• 15. Berkson J. Application of the logistic function to bio-assay. J Am Stat Assoc 1944; 39: 357-65.
• 16. Tu PL, Chung JY. A new decision-tree classification algorithm for machine learning. Proc - Int Conf Tools with Artif Intell ICTAI 1992; 370-7.
• 17. Breiman L. Random Forests. Mach Learn 2001; 45: 5-32.
• 18. Freund Y, Schapire RRE. Experiments with a New Boosting Algorithm. Machine Learning: Proceedings of the Thirteenth International Conference 1996; 148-56.
• 19. Rusdah DA, Murfi H. XGBoost in handling missing values for life insurance risk prediction. SN Appl Sci 2020; 8: 1-10.
• 20. Bunkhumpornpat C, Sinapiromsaran K, Lursinsap C. DBSMOTE: Density-based synthetic minority over-sampling technique. Appl Intell 2012; 36: 664-84.
• 21. Pun S, Thapa S, Timilsina S, Customer Churn Prediction Using ADASYN Sampling Technique and Ensemble Model. Proc IOE Grad Conf 2019; 6: 513-8.
• 22. Wang JB, Zou CA, Fu GH. AWSMOTE: An SVM-Based Adaptive Weighted SMOTE for Class-Imbalance Learning, Sci. Program 2021; 9947621:1-9947621:18.
• 23. Turhan S, Tunç M, Doğu E, Balcı Y. Machine learning in forensic science and forensic medicine: Research on the literature. Adli Tıp Dergisi 2022; 36: 1-7.
• 24. Wu H, Meng FJ. Review on evaluation criteria of machine learning based on big data. In Journal of Physics: Conference Series 2020; 1486: 5; 052026.
• 25. Turhan S, Özkan Y, Yürekli BS, Suner A, Doğu E. Comparison of Ensemble Learning Methods for Disease Diagnosis in Presence of Class Unbalanced: Case of Diabetes. Turkiye Klin J Biostat 2020; 12: 16-26.