# Regression Modeling of ZAP-X Treatment Time

## TL;DR

This paper develops a more accurate model to predict treatment time for the ZAP-X radiosurgery system, improving scheduling and workflow efficiency.

## Contribution

A Ridge Regression model is proposed that outperforms the built-in ZAP-X treatment time estimates with high accuracy.

## Key findings

- Setup time and gantry time were identified as the most influential factors in treatment duration.
- The Ridge Regression model achieved an R² of 0.984 and a mean absolute error of 1.94 minutes.
- The model significantly improves upon the ZAP-X system's internal treatment time estimates.

## Abstract

Accurate prediction of treatment time is critical in stereotactic radiosurgery (SRS) for optimizing patient scheduling and workflow efficiency. The ZAP-X system (ZAP Surgical Systems, Inc., San Carlos, CA), the newest cranial SRS platform, provides a built-in estimate of treatment time but often underestimates the actual treatment duration. This study aimed to develop a robust predictive model for ZAP-X treatment time, improving accuracy and reliability.

Prospective data of 200 patients treated with ZAP-X SRS at Jersey Shore University Medical Center including timing metrics, treatment planning variables, and clinical factors were analyzed. Random Forest regression identified the most critical factors influencing treatment time. Subsequently, Ridge Regression was applied to develop a predictive model, with 10-fold cross-validation used for model tuning and validation.

Random Forest analysis identified setup time and gantry time as the most influential variables affecting the treatment duration, followed by the number of beams, number of isocenters, dose, and target number. The Ridge Regression model demonstrated strong predictive performance (R² = 0.984, MAE = 1.94 minutes, RMSE = 2.49 minutes), significantly surpassing the ZAP-X system’s internal estimates. The model’s accuracy can be attributed to its ability to account for procedural variability, particularly in setup time.

The Ridge Regression model provides a highly accurate and interpretable method for predicting ZAP-X treatment time, outperforming the system’s internal calculations. This model has immediate clinical utility for improving patient scheduling and resource management in SRS. Future work should focus on validating the model across institutions and adapting it to evolving ZAP-X technology.