Objective: The aim: To develop a mathematical model of selecting the extent of surgical intervention in the spinal tumors.
Patients and methods: Materials and methods: The retrospective study included 237 patients with spinal tumors who underwent the following surgeries: vertebroplasty (V); vertebroplasty and spinal fixation (F+V); posterior spinal decompression and spinal fixation (F+F); vertebrectomy and replacement of vertebra by a cage with posterior spinal fixation (F+F+K). The mathematical model is based on the modified Spine Instability Neoplastic Score (SINS). The patients were divided into two clusters. Cluster analysis was used to build a diagnostic decision tree model.
Results: Results: The difference between two clusters is determined by the extent of surgical intervention, the grade of the vertebral lesion, epidural compression, and local kyphosis, and neurological signs as well. The cluster 1 included 115 patients with higher values of SINS compared to the cluster 2. All cases of vertebroplasty belonged to the cluster 2. In the cluster 1 cases of surgery of large extent: F+F; F+V; F+F+K. Analysis of the decision tree model for cluster 1 showed that a type of surgery was determined for 97 patients from 115 that relates to 84.3% of overall accuracy. The decision tree model have a high predictive accuracy for the surgery F+V and better indicators of coverage and predictive accuracy for the surgery F+F+K.
Conclusion: Conclusions: Our study developed a decision tree model to optimize spinal neoplasm surgery, achieving 84.3% accuracy based on significant prognosis criteria. The model considers surgical type, neurological signs, vertebra lesion grade, and stage of epidural compression, potentially improving clinical outcomes.
Keywords: prognosis; spinal neoplasm; cluster analysis; decompression; decision tree.