Anomaly detection in time and spectral domain for the driving motors of multileaf collimator of a medical linear accelerator

Abstract

The multileaf collimator (MLC) is one of the most important devices for beam shaping in a medical linear accelerator. We performed anomaly detection with respect to MLC leaves in this study. The peak voltage associated with MLC leaves of the motors was measured from January 2014 to June 2022. Correlation analysis was performed to detect the fault leaves corresponding to good, normal, and poor conditions. Time-domain analysis was conducted in the threshold range of 15-25 mV. Further, the power spectrum was analyzed in the frequency domain to determine the poor motors among the MLC leaves. For the A24, B01, B02, B03, B24, and B25 leaves, the average power was reported to be 6.55 +/- 1.64 dB. A high mean power difference of 10.26 dB was observed for an average power of - 3.71 +/- 0.39 dB in case of A08 and A19 leaves corresponding to good condition. This indicates that the driving voltage of the poor-condition leaves demonstrated 210.24 +/- 66.77 dB compared to 54.59 +/- 0.79 dB of the good condition leaves at a power spectrum of 10 dB or more. The mean difference in motor load voltage was confirmed to be > 10 dB in case of the poor-condition MLC leaves. The good, normal, and poor conditions of MLC leaves were identified and monitored for maintaining optimal operation of the linear accelerator.