Early Treatment Response Assessment Based on Computed Tomography HU Histogram Feature in the Periphery of Lung Tumors International Journal of Radiation Oncology, Biology, Physics POSTER VIEWING ABSTRACT| VOLUME 96, ISSUE 2, SUPPLEMENT , E647-E648, OCTOBER 01, 2016
Date
10/01/2016Abstract
Purpose/Objective(s)
Early detection of treatment response during radiation therapy (RT) delivery is critical for adaptive RT. It has been reported that radiation can induce changes in CT texture features in tumor during RT delivery for lung cancer and such changes can be potentially used to assess RT response. In this work, we investigate if radiation-induced changes in CT Hounsfield unit (HU) histogram feature in the periphery of lung tumor can be correlated to tumor response.
Materials/Methods
Diagnostic-quality CTs acquired with an in-room CT (CT-on-Rails) during daily CT-guided RT for 10 lung cancer patients, including 2 cases from an external source, were analyzed. For each case, all CT sets were acquired with the same protocol. Lesion for each case was contoured using a thresholding algorithm set at a minimum of −100 HUs to exclude lung tissue. This contour was expanded by 1 cm to form the periphery, then cloned and translated to the contralateral lung as control. To obtain the peripheral and contralateral contours, a second threshold region was set to only obtain regions between −1024 and −100 HU within the expanded (peripheral) and translated (contralateral) contours. Various HU histogram characteristics were extracted from both regions and were correlated with tumor response as characterized by good response (e.g., local recurrence-free survival above 3 years) and poor response (e.g., local recurrence within 1 year or high SUV uptake in follow-up PET).
Results
After normalization to account for changes in the volume of the tumor lesion, the HU histograms of different RT fractions in the tumor periphery in patients with good RT response showed a shift toward normal lung histograms as determined using the contralateral lung. These changes could be characterized with 2 general trends: (1) positive shift in peripheral histogram values by at least 20 HU, resulting in overlap with the normal lung; and (2) decrease in the percentage of volume within the shoulder region above the peak. Poorly responding patients showed minimal change in the peripheral HU histograms during RT. Despite decrease in lesion sizes for most patients, overall shape and characteristics of normalized tumor HU histograms remained consistent throughout the course of treatment. No/minimal change was seen in the contralateral lung.
Conclusion
Radiation can induce changes in CT HU histogram features in the tumor periphery during RT delivery for lung cancer patients who have good responses to RT. During RT delivery, HU histograms in the periphery of lung tumors appear to exhibit a treatment-related shift toward histograms more characteristic of normal lungs. Such a shift was observed even for a case with no shift in the HU histograms in the tumor. This shift does not occur in patients who are responding poorly to RT. These radiation-induced changes in HU histogram features in tumor periphery may be used to assess radiation response during RT delivery.
Author List
S.N. LIM X. Chen E.M. Gore A. LiAuthor
Elizabeth M. Gore MD Professor in the Radiation Oncology department at Medical College of WisconsinView Online