Encrypted login | home

Program Information

Validation of 4D CT Image Reconstruction Using a 3D Lung Phantom

no image available
Y Zheng

Y Zheng*, M Girbino , R Oshinsky , L Vroege , F Jesseph , S Lee , J Yuan , M Machtay , J Sohn , Case Western University, Cleveland, OH


SU-I-GPD-J-75 (Sunday, July 30, 2017) 3:00 PM - 6:00 PM Room: Exhibit Hall

Purpose: To validate the accuracy of 4D CT image data and optimize CT parameters per motion speed and range, a 3D lung-tumor motion simulator was designed and constructed using a 3D printer, a single-board microcontroller, stepper motor, and firmware

Methods: Our lung motion simulator consists of three different size spheres representing tumors, a stepper motor, circuitry (including an Arduino Due microcontroller), and a platform. We printed out three different sized (1.98, 2.48, and 4.01cm) spheres resembling lung tumors. A sphere is attached to a shaft connected to the stepper motor. The microcontroller implements a stepping algorithm utilizing pulse frequency modulation to control the motion of the phantom. Tumor motion was designed to be an arc in x, y, and z axes. We produced two breathing waveforms (5.4 second period + 2.5 cm motion, same period + 1.2 cm motion). The spheres were scanned by using a Siemens Sensation 4D CT scanner. The 4D scanning of breathing cycles is phase-based instead of amplitude-based, and limited to a maximum of eight phases. A total of 6 sets (two breathing patterns of three different sized tumors) were exported to MiM Vista. The tumor volumes were analyzed for possible shape deformation. We auto-contoured the tumor with a consistent threshold intensity value thru the tumor center along axial and sagittal axes. They were fused and overlaid. Two-dimensional analyses were performed using Matlab image toolbox software, and congruence histograms were generated.

Results: We generated 2-D congruence images and histograms from 8 phase studies. Axial contours showed better congruence between phases compared to the sagittal images.

Conclusion: A tumor moving in the axial direction is most likely captured by multi-slice detectors. However, in sagittal and coronal planes, we hypothesize it may be overlapped within a course phase due to a longer time frame compared to tumor speed.

Contact Email: