MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel
  1. Home
  2. Meeting Program
  3. All
  4. Predicting Linac Failure Risks From Machine Performance Check Application Using An Integration of Statistical Process Control and Machine Learning

Predicting Linac Failure Risks From Machine Performance Check Application Using An Integration of Statistical Process Control and Machine Learning

T Fuangrod1*, W Puyati2, A Khawne2, M Barnes3, P Greer3, (1) Faculty of Medicine and Public Health, HRH Princess Chulabhorn College of Medical Science, Chulabhron Royal Academy, Bangkok,TH (2) Department of Computer Engineering, Faculty of Engineering, King Mongkuts Institute of Technology Ladkrabang, Bangkok, TH (3) Department of Radiation Oncology, Calvary Mater Hospital Newcastle, Newcastle, NSW, AUS

Presentations

(Sunday, 7/12/2020)   [Eastern Time (GMT-4)]

Room: AAPM ePoster Library

Purpose: The aim of this study is to demonstrate the feasibility of predicting linac failure risks based on machine performance check (MPC) tests using statistical process control (SPC) and machine learning. This will allow proactive preventative maintenance procedures to be carried out to better optimal linac performance and minimize downtime.


Methods: MPC data was acquired for 490 days or measurements. The data was divided into two set; 85% initial data was used to develop the upper and lower control limits in SPC and prediction model using an autoregressive integrated moving average (ARIMA) method. The remaining 15% of data were used for testing the prediction model accuracy. The system performed two types of predicted data; one-step-ahead values (the next day QA results) and trend with six-step-ahead values (the next week QA trend result). The control chart was constructed to indicate a normal stage of machine performance, while the clinical tolerance level was determined from AAPM TG-142. The gap between normal level to tolerance level was defined as warning stage that provides a window opportunity for rectifying linac performance risks before they become clinically significant. Root-mean-square error (RMSE), absolute error and average accuracy rate were applied to evaluate the system accuracy.

Results: The accuracy of the predictive model is considered high (average root-mean-square error and absolute error for all parameters of less than 0.05). The average accuracy rate for indicating the normal/warning stages was higher than 85.00%.


Conclusion: Linac failure risks from daily MPC tests with an integration of SPC and machine learning will allow preventative maintenance that should reduce the unscheduled linac downtime. Such remedial action can be scheduled out of standard treatment hours to avert disruption to the clinical workflow. The predicting of linac failure risks provides a measure of how long this window of opportunity is.

Keywords

Quality Assurance, Risk, Statistical Analysis

Taxonomy

IM/TH- Formal Quality Management Tools: Machine Learning

Contact Email

Additional Links