Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{208553,
        author = {Ayesha Asif Sayyad},
        title = {OA-MultiFuse A Multimodal Multitask Deep Learning Framework for Knee Osteoarthritis Detection and Severity Assessment},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {4},
        pages = {1942-1949},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=208553},
        abstract = {Knee osteoarthritis is a progressive joint disorder in which cartilage deterioration, osteophyte formation, meniscal change, and joint-space loss develop across multiple anatomical structures. Existing computer-aided approaches often use one imaging modality or optimize classification without explicitly linking the prediction to measurable joint anatomy. This paper proposes OA-MultiFuse, a multimodal and multitask framework that combines a two-dimensional X-ray branch with a three-dimensional MRI branch. The X-ray branch uses a transfer-initialized ConvNeXt encoder and attention pooling for radiographic severity cues. The MRI branch uses a three-dimensional attention U-Net with boundary refinement to segment femur, tibia, cartilage, and joint-space regions. A cross-modal gated fusion module combines the learned representations, while three task heads estimate Kellgren-Lawrence grade, joint-space narrowing, and osteoarthritis probability with uncertainty. The proposed objective joins segmentation, ordinal classification, regression, and calibration losses. This design is motivated by the complementary information provided by X-ray and MRI: radiographs are practical for grading, whereas MRI exposes soft-tissue changes that may precede advanced radiographic damage. The paper defines a reproducible evaluation protocol using patient-level splits, external validation, ablation studies, calibration analysis, and explainability checks. The framework is presented as a proposed methodology; new performance claims require validation on the local and Osteoarthritis Initiative datasets.},
        keywords = {Osteoarthritis, knee imaging, MRI segmentation, X-ray grading, joint-space narrowing, multitask learning, multimodal fusion, explainable AI.},
        month = {September},
        }