State-based prognostics with similarity enhancement for CT X-ray tubes using operational logs

The tube is a core component of the CT system and is characterized by high cost and limited lifespan. This study aims to develop a method for estimating the remaining useful life (RUL) of CT X-ray tubes without requiring a preset failure threshold or prior knowledge of tube life.

To reduce the impact of varying scan conditions, different scan types are first analyzed, and the one with the most stable tube current distribution is selected. Then, a composite metric is applied to degradation parameters to obtain a candidate health indicator (HI). Next, the candidate HI is normalized using the training set and transformed into a percentage health indicator (PHI) to represent tube degradation. Finally, a state-based prognostics with similarity enhancement is applied to the PHI to estimate RUL.

The proposed method is evaluated using real tube log data under leave-one-out cross-validation. The results demonstrate that it achieves optimal prediction performance and enables cross-tube RUL estimation. Moreover, the prediction results can reflect the actual degradation progress of the tube without requiring a preset failure threshold.

The proposed approach provides an effective and practical framework for CT X-ray tube RUL estimation, demonstrating good potential for practical medical equipment maintenance.