Premium
TU‐A‐12A‐03: Monitoring Changes in Tumor Texture Features On Weekly CT and CBCT Scans of NSCLC Patients
Author(s) -
Fave X,
Zhang L,
Yang J,
Fried D,
Balter P,
Court L
Publication year - 2014
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.4889249
Subject(s) - texture (cosmology) , nuclear medicine , feature (linguistics) , medicine , correlation , mathematics , artificial intelligence , filter (signal processing) , radiology , computer science , computer vision , image (mathematics) , geometry , linguistics , philosophy
Purpose: To (1) track changes in CT texture features through time and (2) identify correlations between CT and CBCT texture feature values for NSCLC patients. Methods: Texture features were extracted from weekly CT and CBCT images of 30 NSCLC patients. The physician‐drawn GTV contours on the planning CT, during end of expiration phase (T50), were deformed to weekly CT images. The deformed contours were copied to rigidly registered average‐CT and CBCT images. These contours were imported to our in‐house texture feature analysis program, which performs a Laplacian of Gaussian spatial band‐pass filter at different scales followed by a calculation of the tumor uniformity. Three filter sizes were used (coarse, medium, and fine). (1) The normalized change in uniformity as a function of time was evaluated from the T50 images. (2) The uniformity values from the average‐CT and CBCT images were used to assess correlation. Results: (1) Texture features measured in T50 scans were shown to change over time for individual patients. As filter size increased, the range of measured normalized changes decreased from 103.6% (SD=22.8) to 25.2% (SD=7.9). (2) Texture features extracted from CT were shown to correlate with values in CBCT with R 2 values of 0.79(coarse), 0.84(medium), and 0.84(fine). Correlation values increased for more coarse filters and lower lung tissue thresholds. The correlation was also dependent on removal of patients with metal artifacts, severe truncation of anatomy in CBCT images, or those with tumors smaller than 4.3cc. Seven patients were removed. Conclusions: We showed texture features vary for patients during their course of treatment and these changes can be monitored with CBCT images. Various studies have indicated CT texture features may correlate with patient prognosis. Therefore, our results could be used to observe changes in a patients prognosis and potentially adapt their treatment without any additional imaging studies.