Galib Muhammad Shahriar Himel, Anusha Achuthan, Yusuf Abas Mohamed, Bee Ee Khoo, Mohd Ezane Bin Aziz
Abstract
Primary malignant bone tumours of the skeleton have a great diversity in their biological behaviour, and the most common in adolescence is osteosarcoma for which the diagnosis and therapeutic management are both a challenge. The use of machine and deep learning in image analysis for MRI, diffusion-weighted imaging (DWI) and dynamic contrast-enhanced MRI (DCE-MRI) to diagnose osteosarcoma has shown major improvements recently. T
Introduction
Osteosarcoma is an aggressive primary bone cancer that mainly affects adolescents and young adults, accounting for roughly one fifth of all primary bone malignancies [1,2]. Because it often spreads early to the lungs and shows heterogeneous responses to treatment, it remains difficult to diagnose accurately and to manage optimally in everyday clinical practice. Careful evaluation of the true extent of the lesion, the response to neoadjuvant chemotherapy, and the risk of metastatic spread is essential for improving prognosis. Magnetic resonance imaging (MRI)—including advanced techniques such as diffusion weighted imaging (DWI) and dynamic contrast enhanced MRI (DCE MRI)—has become central for detailed tumor characterization, surgical planning, and post therapy assessment [3–5].
Materials and Methods
This systematic review was designed to locate and synthesize research from the last six years that uses machine learning (ML) and deep learning (DL) methods on medical imaging—particularly MRI and related techniques—for the diagnosis, segmentation, and prognostic evaluation of osteosarcoma. The review followed the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines, with clearly defined inclusion and exclusion criteria to structure the search and screening workflow.
Results and Discussion
Datasets Descriptions
This study describes 38 different studies which used different medical imaging datasets focused on osteosarcoma research. The datasets primarily include MRI (T1, T2, DWI, DCE-MRI), along with some multi-modality and clinical data. They vary in size, preprocessing methods, and annotations, supporting tasks like tumor segmentation, treatment response analysis, and AI model development.
Conclusion
In recent years, machine learning and deep learning have rapidly advanced osteosarcoma imaging analysis, yet a comparative evaluation reveals distinct performance hierarchies among computational strategies. Quantitative synthesis demonstrates that Lightweight and Multimodal Convolutional Neural Networks (CNNs) currently provide the most robust and consistent spatial segmentation, achieving mean Dice Similarity Coefficients exceeding 94%. While standard CNNs and Vision Transformers also exhibit high baseline accuracy, their broader performance variance underscores a heightened sensitivity to dataset scale and annotation quality.
Acknowledgments
The authors would like to express their gratitude to the School of Computer Sciences at Universiti Sains Malaysia (USM) for providing the institutional resources and research environment that facilitated the completion of this systematic review.
Citation: Himel GMS, Achuthan A, Mohamed YA, Khoo BE, Bin Aziz ME (2026) Deep learning applications in osteosarcoma MRI: A systematic review of recent advances in AI-based osteosarcoma diagnosis. PLoS One 21(8): e0354896.
https://doi.org/10.1371/journal.pone.0354896
Editor: Mario Tortora, Università degli Studi di Napoli Federico II: Universita degli Studi di Napoli Federico II, ITALY
Received: February 13, 2026; Accepted: July 14, 2026; Published: August 18, 2026
Copyright: © 2026 Himel et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the paper and its Supporting information files.
Funding: This work was supported by the Universiti Sains Malaysia, Research University Individual (RUI) Grant Scheme (Grant Number: R502-KR-ARU001-0000002784-K134). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
