Adverse events and medical errors result in thousands of accidental deaths and over one million excess injuries each year. To avoid medical errors in radiation cancer treatment, careful attention needs to be made to ensure accurate implementation of the intended treatment plan. We propose a SmartTool to automatically detect and highlight potential errors in a radiotherapy treatment plan, in real time and before its execution. SmartTool will double check all the treatment parameters in the background against a previously built Predictive Model of a Medical Error (PMME) and flag the operator, [post human QA,] if there is a discrepancy in the treatment plan, by stopping execution, highlighting the outlier treatment parameter and prompting human intervention. To build the PMME we will mine the dataset of previously treated cancer patients, by clustering the data in the groups based on treatment parameter similarity, labeling the clusters and using an innovative algorithm to build a highly accurate anomaly detection tool. PMME will also be dynamically updated [to include new treatment data instances coming in to the system, and updating the model should any treatment flags be identified as false positive or false negative]. The vastly innovative aspect of SmartTool is in the novel use of machine learning techniques to automatically build an anomaly prediction model on unlabeled data (customarily a labeled data is required to build a predictive model) and provide an automatic, real time and unobtrusive intelligent computational treatment checking algorithm. Moreover, having an analytical model of an outlier/anomaly offers the capability to describe the conditions of the outlier being created and is the essential in gaining investigative (and medical) insight in what went wrong and how to improve the process in the future. SmartTool can also be applied in a variety of other medical areas (e.g. predicting errors in pharmacy, laboratory data, and treatment procedure data), to detect anomalies and describe them, offering potential novel medical discoveries and a prospect of saving thousands more lives, with a vast commercialization aspect.
The proposal is aimed at promoting research and development in biomedical computational science and technology that is consistent with the objective of the NIH and NCI to support rapid progress in areas of scientific opportunity in biomedical research, and enhancing the public health. If the project is successfully completed, this proof of concept study will result in a valuable health information technology tool for automatic detection of catastrophic errors in cancer radiotherapy, which adds another safeguard for patient safety.
