Page 91 - GTM-4-3

Global Translational Medicine

ORIGINAL RESEARCH ARTICLE
Leveraging convolutional neural networks

to address overfitting and generalizability in
automated bone fracture detection

1
Abel Gallegos , Daniel Nasef 2 , and Milan Toma *
2
1 Department of Osteopathic Manipulative Medicine, College of Osteopathic Medicine, New York
Institute of Technology at Arkansas, Arkansas State University, Jonesboro, Arkansas, United States
of America
2 Department of Osteopathic Manipulative Medicine, College of Osteopathic Medicine, New York
Institute of Technology, Old Westbury, New York, United States of America

Abstract

Bone fractures represent a significant health burden that demands precise and timely
diagnosis to optimize patient outcomes. To address challenges such as data scarcity,
overfitting, and generalizability, this study investigates the use of convolutional
neural networks (CNNs) for automated fracture detection in X-ray images. A dataset
of 4,900 X-ray images was preprocessed and evenly divided into training, validation,
and test subsets. The proposed CNN model directly addressed generalizability
*Corresponding author: and overfitting issues by prioritizing training stability and incorporating advanced
Milan Toma
(tomamil@tomamil.com) techniques. These techniques included batch normalization and dropout to enhance
stability and mitigate overfitting, with five-fold cross-validation yielding an average
Citation: Gallegos A, Nasef D,
Toma M. Leveraging convolutional accuracy of 95%. Validation and held-out test datasets achieved accuracies of
neural networks to address 95.8% and 94.5%, respectively, while external validation on an independent dataset
overfitting and generalizability in confirmed the model’s generalizability at 91.7%. High recall rates across all datasets
automated bone fracture detection.
Global Transl Med. 2025;4(3):83-95. underscore the model’s capacity to minimize missed fracture diagnoses, whereas
doi: 10.36922/gtm.8526 slightly lower precision on external data indicates a need to address false positives.
These findings suggest that artificial intelligence is best deployed as a screening tool,
Received: January 14, 2025
serving as an initial triage mechanism that flags potential cases for further human-
1st revised: February 26, 2025 guided evaluation, thereby enhancing clinical efficiency without replacing the
2nd revised: March 9, 2025 diagnostic expertise of healthcare professionals.
3rd revised: March 21, 2025
Accepted: August 12, 2025 Keywords: Bone fracture detection; Artificial intelligence; Convolutional neural networks;
Medical imaging; Deep learning
Published online: August 29, 2025
Copyright: © 2025 Author(s).
This is an Open-Access article
distributed under the terms of the 1. Introduction
Creative Commons Attribution
License, permitting distribution, Bone fractures are a common and significant health concern, often requiring prompt
and reproduction in any medium,
provided the original work is and accurate diagnosis for effective treatment. In recent years, the integration of
properly cited. artificial intelligence (AI) and machine learning (ML) into radiology has enhanced the
1
Publisher’s Note: AccScience field of medical imaging, particularly in the domain of bone fracture detection. This
Publishing remains neutral with study presents an approach to automated bone fracture detection using convolutional
regard to jurisdictional claims in
published maps and institutional neural networks (CNNs), addressing key challenges in the field and contributing to the
affiliations. advancement of diagnostic accuracy and efficiency in orthopedic imaging.

Volume 4 Issue 3 (2025) 83 doi: 10.36922/gtm.8526