Back to search

McMaster University

Osteoporotic Hip Fracture Risk Assessment: SSAM-Based Machine Learning Models for Improved Fracture Prediction

Abstract

dc:description.abstract

Due to the severe health impacts of osteoporotic hip fracture, identifying patients at elevated risk is essential for physicians. Conventional diagnostics using bone mineral density (BMD) and clinical risk factors (CRF) have shown limited accuracy predicting acute fracture risk. Machine Learning (ML) offers improvements, particularly using Statistical Shape and Appearance Models (SSAM) to extract data from dual x-ray absorptiometry (DXA) scans. This thesis centred on an SSAM-based fracture prediction tool herein titled the “Osteoporotic Hip Fracture Risk Assessment”. There were three phases conducted developing and validating this tool, with DXA scans and CRF data being provided by the Canadian Multicentre Osteoporosis Study (CaMos) and Amgen FREEDOM trial. First, five model types were compared: Logistic Regression (LR), LR with Least Absolute Shrinkage and Selection Operator (LASSO) regularization, LR with ElasticNet regularization, Random Forest (RF), and Adaptive Boosting (AdaBoost). All displayed high accuracy, while RF and AdaBoost models displayed elevated sensitivity in select datasets. LR, RF, and AdaBoost models were selected for further investigation. Alternate cross-validation methods were tested during the second phase to assess methodological robustness, with models validated using Leave-One-Out Cross-validation (LOOCV) 10-fold cross-validation, and 5-fold cross-validation. LOOCV showcased elevated performance metrics, with minimal difference between 5-fold and 10-fold. Finally, modified 10-fold cross-validations were performed to determine model ability to detect differences in fracture risk progression in patients treated with either placebo or the anti-osteoporosis drug denosumab (Prolia®, Amgen, Thousand Oaks, CA). None of the models detected significant differences between medication cohorts across 12 months, suggesting a need for future development identifying short term medication-driven risk change. This work has provided developments as well as insight into the current abilities of the Osteoporotic Hip Fracture Risk Assessment. These advancements have positive implications for improved hip fracture risk assessment and will help to direct future investigations.

Degree

thesis:*
Department dc:contributor.department
Biomedical Engineering
Year dc:date.issued
2027

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Delaney, Aiden
Advisor dc:contributor.advisor
  • Quenneville, Cheryl

Rights

dc:rights
Statement dc:rights
  • Attribution-NonCommercial-NoDerivs 2.5 Canada
Language dc:language.iso
en_US

Identifiers

dc:identifier.*

Chain of custody

source
Harvested from
McMaster University
Base URL
macsphere.mcmaster.ca/server/oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
related terms
citation

Delaney, Aiden. Osteoporotic Hip Fracture Risk Assessment: SSAM-Based Machine Learning Models for Improved Fracture Prediction. 2027. https://hdl.handle.net/11375/33211