New 5D model may predict motion of lung tumors

A new 5D mathematical model seems to accurately predict the motion of lung tumors on CT scans obtained while patients breathe, according to a preliminary report presented at the American Society for Therapeutic Radiology and Oncology meeting in October.

The ability to quantitate the motion of internal organs and tumors during respiration would make it possible to compensate for or counteract the effects of breathing, thereby improving the accuracy of radiation treatment planning, said Daniel A. Low, Ph.D., an associate professor of radiation oncology at Washington University in St. Louis.

"A method that modeled the motion as a function of a noninvasive external measurement would be a valuable aid for reducing the effective motion during irradiation. For example, the linear accelerator can be programmed to activate only during a predetermined phase of breathing, based on advanced treatment planning optimization techniques," Low said.

Most groups studying the motion of lung tumors view breathing as a function of time (the fourth dimension in 4D CT), which assumes that breathing is predictably cyclic and that tidal volume remains the same from one breath to the next, he said.

"The reality is absolutely not that. These patients do not breathe, for the most part, in little sine waves. They have severe lung disease, independent of cancer. They have emphysema, and they have serious underlying physiologic damage caused by smoking-they are not normal breathers," Low said.

In addition, most radiation oncology groups use coached breathing, in which patients are instructed when and how to breathe during scanning.

"We specifically avoided coached breathing because we want to understand how well these models work when a patient is not concentrating on breathing," he said.

The investigators first assessed the performance of a 4D CT process in which multislice CT scans were sorted (gated) by tidal volume. Each of 12 patients with lung cancer underwent 15 consecutive CT scans while spirometry measurements were obtained. The scans were sorted and reconstructed to create a 4D data set. Analyses showed near-perfect correlation between tidal volume and the motion of internal objects, which was assessed from internal air content on the CT scan.

They then developed a mathematical breathing motion model based on five dimensions: the spatial location of the object (x, y, and z) during a reference breathing phase, tidal volume, and airflow. To test the model, they compared the actual positions of 40 moving objects in the lung on CT scans obtained during breathing with the predicted positions from the model. This comparison revealed incredibly good fits, according to Low. The objects tracked moved 4 to 11 mm during breathing, and the mean discrepancy between actual and predicted locations was about 0.5 mm.

Although the model had been well tested in only one patient at the time of the ASTRO meeting, its excellent performance was especially noteworthy because that patient had widely ranging tidal volumes, Low said.

Nonetheless, the findings using the new model must be viewed as preliminary, Low said. His team is now rigorously testing the model to see how it performs in a variety of patients and sessions.