An Enhanced Residual Architecture for Automated Detection of Bone Metastasis on Whole-Body SPECT Images
An Enhanced Residual Architecture for Automated Detection of Bone Metastasis on Whole-Body SPECT Images
Dr. Krishna Gudi 1, Bhoomika N2
1Associate Professor, Department of Computer Science and Engineering, KSIT, India
2 Student, Department of Computer Science and Engineering, KSIT, India
Abstract:
Bone Advanced malignancies frequently progress to bone metastasis, necessitating early and precise identification to optimize therapeutic strategies. This paper introduces an advanced residual deep learning framework designed to automate the identification of osseous metastatic lesions using whole-body Single- Photon Emission Computed Tomography (SPECT) scans. By embedding residual learning blocks, the architecture optimizes feature propagation, mitigates gradient degradation, and minimizes critical data loss during deep feature extraction. The model effectively differentiates malignant focal uptakes from normal physiological bone activity by evaluating complex functional imaging profiles. Empirical evaluations indicate that the proposed framework achieves superior diagnostic accuracy, sensitivity, and consistency over traditional computational techniques. Ultimately, this clinical decision-support tool offers a robust mechanism for accelerated and precise diagnosis, thereby improving patient triage and therapeutic management.
Keywords:
Bone metastasis, Single-Photon Emission Computed Tomography (SPECT), Clinical decision-support tool,Feature extraction.