The Development of Machine Learning Models for Fracture Detection in Healthcare - TREO Talk Paper

Aug 5, 2026

Upload by Asher

Artificial intelligence and advanced analytics in healthcare have triggered debate, especially about whether AI could displace clinical work. This study develops a cost-effective AI support solution aimed at augmenting clinicians while respecting ethical standards. Two machine learning models are trained on X-ray images: a classification model separating fractured from non-fractured cases and an object detection model localizing fracture regions. The models’ measured accuracy, precision, and detection metrics indicate feasibility as a diagnostic assistant. Results suggest improved imaging workflow support for both entry-level and experienced professionals, with future work refining fracture-pattern understanding and exploring hybrid or advanced imaging approaches.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages2

Reading Time5 min

AccessFree to download

Abstract

Study objective

Methods: classification and object detection

Results and evaluation metrics

Implications and future work

Association for Information Systems AIS Electronic Library (AISeL) | AMCIS 2024 TREOs | AIS TREO Papers | | --- | --- | | 8-16-2024
The Development of Machine Learning Models for Fracture Detection in Healthcare
Alex M. Campbell
Western Michigan University, [alexander.m68.campbell@wmich.edu](alexander.m68.campbell@wmich.edu)
Jiahe Song
Western Michigan University, [carrie.song@wmich.edu](carrie.song@wmich.edu)
Follow this and additional works at: [https://aisel.aisnet.org/treos_amcis2024](https://aisel.aisnet.org/treos_amcis2024) | | Recommended Citation Campbell, Alex M. and Song, Jiahe, "The Development of Machine Learning Models for Fracture Detection in Healthcare" (2024) . AMCIS 2024 TREOs. 135. [https://aisel.aisnet.org/treos_amcis2024/135](https://aisel.aisnet.org/treos_amcis2024/135) This material is brought to you by the AIS TREO Papers at AIS Electronic Library (AISeL) . It has been accepted for inclusion in AMCIS 2024 TREOs by an authorized administrator of AIS Electronic Library (AISeL) . For more information, please [contact](contact elibrary@aisnet.org)[ elibrary@aisnet.org](contact elibrary@aisnet.org). The Development of Machine Learning Models for Fracture Detection in Healthcare TREO Talk Paper Alex M. Campbell Western Michigan University [alexander.m68.campbell@wmich.edu](alexander.m68.campbell@wmich.edu) Jiahe Song Western Michigan University [carrie.song@wmich.edu](carrie.song@wmich.edu) Abstract The integration of artificial intelligence (AI) and advanced analytics in healthcare has sparked debate in both practice and academia, particularly concerning its potential to displace and/or replace jobs (Lee and Yoon, 2021; Thompson et al., 2020). However, the inherent complexity of diagnostic imaging examinations does make AI a set of appealing technologies to be equipped in healthcare facilities (Zeltzer et al. 2023) to improve care quality. This study aims to develop a cost-effective AI solution that augments rather than replaces the existing medical professionals, adhering to the established ethical standards. Specifically, we developed two machine learning models trained on X-ray images: (1) an image classification model to distinguish between fractured and non-fractured images and (2) an object detection model to pinpoint fracture locations within the images. The classification model, trained on 612 fractured and 620 nonfractured images, achieved an accuracy of 64.3%, with a precision of 66% for fractured images and 62% for non-fractured ones. The object detection model, utilizing the same dataset with additional bounding box annotations for fractures, showed a box precision of 57.1% and a mean average precision of 47.7% at the 50th percentile Intersection over Union (IoU) . These results demonstrate the feasibility of AI as a supportive tool in medical diagnostics, offering a promising avenue for enhancing healthcare services without increasing resource demand on the existing medical workforce. This study offers key implications for healthcare practice. First, the observed accuracies and precisions of our models can add ease to the imaging diagnostics process for both entry-level and experienced medical professionals. Second, our machine-learning model training operations help provide solutions to the nuanced challenges in automating AI diagnostics for healthcare practitioners. Our future work will focus on refining these models to better navigate the intricate patterns of fractures, potentially integrating more advanced imaging techniques or hybrid AI approaches. References Lee, D., and Yoon, S. N. 2021.“Application of Artificial Intelligence-Based Technologies in the Healthcare Industry: Opportunities and Challenges,” International Journal of Environmental Research and Public Health (18:1), pp.271. Thompson, S., Whitaker, J., Kohli, R., and Jones, C. 2020.“Chronic Disease Management: How IT and Analytics Create Healthcare Value through the Tempo

FAQs

What is the main goal of this study?

To develop a cost-effective AI solution that supports and augments existing medical professionals for fracture diagnosis rather than replacing them.

What models were developed for fracture detection?

An X-ray image classification model distinguishes fractured vs. non-fractured images, and an object detection model pinpoints fracture locations using bounding box annotations.

How did the models perform?

The classification model achieved 64.3% accuracy with precision of 66% for fractured images and 62% for non-fractured ones. The object detection model reached box precision of 57.1% and mean average precision of 47.7% at the 50th percentile IoU.

Aug 5, 2026

Upload by Asher