Linear Accelerator Multi-Collimator Leaves' displacement Data

One of the industries that uses Machine Learning is Radiation Oncology where it’s used to tackle automation of quality assurance checks in order to reduce, avoid and/or eliminate frequent mistakes and faults occurrence during Treatment by Linear Accelerator with Multi-leaf Collimator (Linac MLC). Most of Machine Learning applications have been implemented using supervised learning algorithms rather than unsupervised ones. But, most of the time, there’s an obvious bias in choosing which supervised machine learning algorithm to use. And this bias may lead to lower accuracy in prediction results.  So In this paper, we have proposed by evidence the newly use of Logistic Regression algorithm to predict Linac MLC’s positioning accuracy problem which achieved 98.68% prediction accuracy with robust and stable performance across different sets of Linac MLC’s positioning displacement data. We have got this evidence by evaluating the performance for different supervised machine learning algorithms (i.e. Logistic Regression, Decision Tree, Support Vector Machine, Random Forest, Naïve Bayes, and K-Nearest Neighbor) in the prediction of Linac MLC’s positioning accuracy problem utilizing leaves’ positioning displacement dataset that are already have labeled results as training and test dataset. The Performance have been evaluated by two factors; measuring the prediction accuracy and Receiver operating characteristic curve per each algorithm [2]. Based on that evaluation, and in order to get near-optimum performance prediction in the prediction of Linac MLC’s leaf positioning accuracy, we have successfully proposed the proper selection order for supervised Machine Learning algorithms.  And thus the selection bias have been successfully avoided as well as the negative side effects (i.e. ineffective preventive maintenance plan for Linac MLC to avoid and solve cause behind inaccurate leaf displacement such as motor fatigue and stuck problems) that could have happened if such selection bias occurred.