A fiducial-less tracking method for radiation therapy of liver tumors by diaphragm disparity analysis part 2: validation study by using clinical data

Abstract

Objective

Motion management must be considered in treating liver tumors. One effective approach is real-time tumor tracking, which can be performed by the CyberKnife® Robotic Radiosurgery System through the Synchrony® Respiratory Tracking System. It uses a combination of kV images, LED markers, an infrared camera, and surgically implanted fiducial markers to track tumors under the influence of respiration. However, the use of fiducial markers through an invasive procedure can lead to complications. In our previous simulation study, we were able to demonstrate the feasibility of our proposed fiducial-less tracking technique using a digital phantom under regular respiratory motion. The aim of this study is to further validate this innovative method by using a digital phantom data under the irregular respiratory cycles as well as clinical data from patients under the Cyberknife environment.

Methods

As performed in our previous simulation study, abdominal 4DCT datasets of one breathing cycle, from the digital phantom and from four patients, were previously generated or acquired. Associated with the breathing cycles in the 4DCT datasets, one set of DRR images (+ 45° or − 45°) was produced for each breathing phase. On each DRR, an outline of the lung-diaphragm border was detected using an edge detection algorithm. The tracked target volume's gravity center was identified for each phase of the breathing cycle by a MATLAB program, serving as the ground truth for the validation. Using artificial neural networks (ANN), four models for the phantom and six models for the patient data, correlating the diaphragm's location with the corresponding 3D location of the tracked target volume, were compared. Assessment was performed by using the root-mean-squared error (RMSE) values through the leave-one-out (LOO) validation criterion.

Results

The averaged RMSE for the phantom data was 1.05 ± 1.14 mm. When using the patient data from the + 45° projection, the averaged RMSE was 2.13 ± 1.79 mm, while from the − 45° projection, the averaged RMSE was 2.26 ± 2.40 mm. Using the proposed method in both phantom validation and patient data validation, the RMSE is closely related to the 4DCT reconstruction error and to the distance from the lung-diaphragm border to the tracked tumor.

Conclusion

We proposed and investigated the fiducial-less tracking method to follow tumor motion in the real-time under the influence of respiration. The study shows the feasibility of accurately predicting the tumor's position with the use of lung-diaphragm border's information through available kV images without gold fiducial markers. This developed diaphragm disparity-analysis-based approach, verified with clinically accepted errors, has the potential to replace fiducial markers in clinical applications.

https://ift.tt/2AgG5pN