Quantitative evaluation of 4D Cone beam CT scans with reduced scan time in lung cancer patients

Abstract

Purpose: Image guided radiotherapy (IGRT) based on respiration correlated cone-beam CT (4D-CBCT) provides accurate tumour localisation in lung cancer patients by taking into account respiratory motion when deriving setup correction. However, 4D-CBCT scan times are typically longer than for acquisition of 3D-CBCT scans, e.g. 4 min. This work aims to quantitatively evaluate the effect of reduced scan times on 4D-CBCT image quality and registration accuracy in lung cancer patients.

Methods and materials: Scan times down to 1 min were simulated by retaining only projection images corresponding to every second, third or fourth respiratory cycle in forty-four 4D-CBCTs from 15 lung cancer patients. In addition twenty 2-minute scans were acquired for 12 lung cancer patients. Image quality was quantified by assessing registration accuracy in the shorter scan times, comparing to the 4-minute scan registration result where available as reference.

Results: Use of 2-minute scans had little impact on registration accuracy or ability to detect tumour motion: automatic registration accuracy was within 2 mm in 6/8 scans analysed with 2-minute acquisitions, and 96.6% of registration discrepancies were within 2 mm for the simulated scans. When the scan time simulated was below 2 min, automatic registration results still agreed within 2 mm for 84.7% of scans, however visual image quality was considerably degraded.

Conclusion: A 4D-CBCT acquisition time of 2 min produces scans of sufficient image quality for IGRT in most lung cancer patients, as demonstrated quantitatively by assessing the impact on automatic registration accuracy in simulated and real acquisitions.

Keywords: Image guided.

Fig. 1

Comparison of visual image quality for Patient 02 showing axial (left) and coronal (right) slices at peak exhale (phase 0) and mid exhale (Phase 7) to illustrate phases with best and worst image quality for each simulated scan time.

Fig. 2

Tumour motion profiles for Patient 09 comparing the acquired 2-minute scan (a) and simulated 2-minute (b), 1 min 20 s (c) and 1 min (d) scans against the 4-minute scan. The acquired 2-minute scan data are offset by the applied couch shifts between scans such that the profiles can be compared.

Fig. 3

Visual comparison of the acquired 2-minute and simulated 2-minute scans with the original 4-minute scan for phases 0 and 7 in axial (left) and coronal (right) view for Patient 11.