Spine Oncology Detection Net: a Data-Centric Deep Neural Model for Diagnosis
Year: 2025
Type: conference-paper
Abstract: Many malignant and benign tumors that occur in the spinal column and cord are included under spine oncology, challenging doctors because of their complex features and unique pathologies. Finding and correctly classifying these tumors early is important for providing patients with effective care. The field of spine oncology has seen great potential from the use of ML which can now automate, scale and reproduce traditional diagnostic procedures. This study investigates the ability of different ML techniques-CNNs, SVMs, random forests and ensemble methods-to find and classify spinal tumors using images (from MRIs and CTs), health records and examination of tissue samples. Image segmentation and tumor localization have been nicely handled by CNN models, whereas traditional ML models work well with data that has structure. Based on accuracy, precision, recall and AUC measures, XGBoost models accurately classifies with good results over traditional models. Machine learning can make spine tumor detection and classification easier, faster and more precise, helping clinicians treat patients earlier and design more individualized approaches.
Language: English
Topic: Medical Imaging and Analysis
Subfield: Biomedical Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7125596554