Encrypted login | home

Program Information

Automatically Detect Patient Treatment Position and Orientation in KV Portal Images

D Yang

J Qiu1 , D Yang2*, (1) Washington University in St Louis, Taian, SHANDONG, (2) Washington University School of Medicine, St Louis, MO


SU-E-J-15 (Sunday, July 12, 2015) 3:00 PM - 6:00 PM Room: Exhibit Hall

Purpose:In the course of radiation therapy, the complex information processing workflow will result in potential errors, such as incorrect or inaccurate patient setups. With automatic image check and patient identification, such errors could be effectively reduced. For this purpose, we developed a simple and rapid image processing method, to automatically detect the patient position and orientation in 2D portal images, so to allow automatic check of positions and orientations for patient daily RT treatments.

Methods:Based on the principle of portal image formation, a set of whole body DRR images were reconstructed from multiple whole body CT volume datasets, and fused together to be used as the matching template. To identify the patient setup position and orientation shown in a 2D portal image, the 2D portal image was preprocessed (contrast enhancement, down-sampling and couch table detection), then matched to the template image so to identify the laterality (left or right), position, orientation and treatment site.

Results:Five day’s clinical qualified portal images were gathered randomly, then were processed by the automatic detection and matching method without any additional information. The detection results were visually checked by physicists. 182 images were correct detection in a total of 200kV portal images. The correct rate was 91%.

Conclusion:The proposed method can detect patient setup and orientation quickly and automatically. It only requires the image intensity information in KV portal images. This method can be useful in the framework of Electronic Chart Check (ECCK) to reduce the potential errors in workflow of radiation therapy and so to improve patient safety. In addition, the auto-detection results, as the patient treatment site position and patient orientation, could be useful to guide the sequential image processing procedures, e.g. verification of patient daily setup accuracy.

Funding Support, Disclosures, and Conflict of Interest: This work was partially supported by research grant from Varian Medical System.

Contact Email: