Real-time image processing algorithm for marker-less tumor tracking using X-ray fluoroscopic imaging

Purpose: To ensure accuracy in respiratory gating treatment, X-ray fluoroscopic imaging is used to detect tumor position in real-time. Detection accuracy is strongly dependent on image quality, however, particularly positional differences between the patient and treatment couch. We developed a new algorithm to improve image quality in X-ray fluoroscopic imaging and report preliminary results.Methods and Materials: Two oblique X-ray fluoroscopic images were acquired using a dynamic flat panel detector (DFPD) for two lung cancer patients. The weighting factor was applied to the DFPD image in respective columns, because most anatomical structures as well as the treatment couch and port cover edge were aligned in the superior-inferior direction when the patient lay on the treatment couch. The weighting factors at respective columns were varied until the standard deviation of pixel values within the image region was minimized. Once the weighting factors were calculated, DFPD image quality was improved by applying the factors to multi-frame images.Results: Application of the image-processing algorithm produced a substantial improvement in the quality of images, and image contrast was increased. The treatment couch and irradiation port edge, which were not related to patient position, were removed. Average image-processing time for the images was 1.1ms, showing that this fast image processing can be applied to real-time tumor tracking systems.Conclusions: These findings indicate that this image-processing algorithm improves image quality in lung cancer patients, and successfully removes objects not related to the patient.Advances in knowledge: Our image-processing algorithm might be useful in improving gated-treatment accuracy