Potential Improvement in Apparent Diffusion Coefficient (ADC) Measurement by Respiratory Correlated Four Dimensional Diffusion-Weighted MRI (4D-DWI): Initial Investigation on Digital Phantoms and Human Subjects
Yilin Liu1, Fang-Fang Yin2, Brian Gary Czito2, Mustafa R. Bashir 3, Manisha Palta 2, Xiaodong Zhong 4, Brian M. Dale 5, and Jing Cai2

1Medical Physics Graduate Program, Duke University Medical Center, Durham, NC, United States, 2Radiation Oncology, Duke University Medical Center, Durham, NC, United States, 3Department of Radiology, Center for Advanced Magnetic Resonance Development, Duke University Medical Center, Durham, NC, United States, 4MR R&D Collaborations, Siemens Healthcare, Atlanta, GA, United States, 5MR R&D Collaborations, Siemens Healthcare, Cary, NC, United States

Synopsis

Diffusion-weighted imaging (DWI) has been shown to have superior tumor-to-tissue contrast for cancer detection in abdominal region. However, the respiratory motion may induce severe imaging errors or artifacts for DWI images. This study aims at developing and evaluating a respiratory correlated 4D-DWI technique using a retrospective sorting method for imaging respiratory motion on human subjects. Comparing to free breathing DWI, 4D-DWI can lead to more accurate measurement of ADC. This has a great potential to improve the visualization and delineation of cancer tumors for radiotherapy.

Purpose

Diffusion-weighted imaging (DWI) has been shown to have superior tumor-to-tissue contrast for cancer detection compared to other MRI sequences and CT, especially in the abdominal region. However, respiratory motion may induce severe imaging errors or artifacts in DWI images. This study aims at developing and evaluating a respiratory correlated 4D-DWI technique using a retrospective sorting method for imaging respiratory motion of human subjects. We also evaluated its effect on Apparent Diffusion Coefficient (ADC) measurement using feature analysis.

Methods

Image acquisition was performed by repeatedly imaging a volume of interest using an interleaved multi-slice single-shot echo-planar-imaging (EPI) 2D-DWI sequence in the axial plane. Cine MRI using steady state free precession was also acquired as a reference showing respiratory motion. Each 2D-DWI image with an intermediately high b-value (b=500 s/mm2) was acquired in x, y and z diffusion directions. Respiratory motion was simultaneously recorded using a respiratory bellows, which provided a synchronized respiratory signal.

The respiratory signal was used in the retrospective phase sorting algorithm to re-sort DWI images acquired in x, y and z diffusion directions, respectively. Then the sorted DWI images in three directions were combined to reconstruct 4D-DWI, and ADC was calculated for each phase. As a comparison, DWI with no motion correction (free breathing DWI) was also reconstructed using the same datasets, as well as the ADC. The process is illustrated in Fig.1.

As a preliminary feasibility study, this technique was implemented on a computer simulated 4D digital human phantom (XCAT) [reference 1, reference 2] with a heterogeneous liver tumor. The respiratory motion of the phantom was generated using 10 liver cancer patients’ breathing profiles measured previously. Image acquisition process was simulated. 4D-DWI, free breathing DWI and the corresponding ADC maps were reconstructed. Motion trajectories of the tumor were extracted from 4D-DWI and compared with average breathing curves calculated from the input profiles. The mean motion trajectory amplitude differences (D), mean ADC value and entropy of the tumor were calculated.

The technique was then evaluated on two healthy volunteers and one lung cancer patient (under a HIPAA-compliant IRB-approved study protocol with informed consent). Motion trajectories of defined regions of interest (ROI), right kidney of the healthy volunteers and tumor of the patient, respectively, were extracted from 4D-DWI and compared with those obtained from the cine MRI acquisition as a reference. D values were calculated. Mean ADC value and entropy of Volume of Interest (VOI: liver for the healthy volunteer and tumor of the patient) was calculated for one healthy volunteer and one patient.

Results

Tumor trajectories extracted from simulated XCAT 4D-DWI were consistent with the input signal: average D values for the tumor were 1.9 mm in the superior-inferior (SI) direction, and 0.4mm in the anterior-posterior (AP) direction. Fig.2(a) shows example ADC map set for the XCAT heterogeneous liver tumor. On average for the 10 patients’ breathing profiles, the mean tumor region ADC value was 2.7×10-3 mm2/s with 4D-DWI and 4.3×10-3 mm2/s with free breathing DWI, respectively. The ground-truth was 2.3×10-3 mm2/s, as shown in Fig.2(b). The mean tumor region entropy was 0.29 with 4D-DWI and 0.87 with free breathing DWI, respectively. The ground-truth was 0.24, as shown in Fig.2(c). The Wilcoxon Signed Rank test shows that ADC measurements were significantly more accurate with the 4D-MRI technique.

Reconstructed 4D-DWI of the human subjects also revealed the respiratory motion clearly. Figure 3 shows example healthy volunteer 4D-DWI images, and the free-breathing DWI images in comparison. The corresponding ADC maps for this healthy volunteer are shown in Fig.4. In addition, ADC maps for the patient are shown in Fig.5. The mean values of D were 2.6 mm (SI) and 1.7 mm (AP) for the two healthy volunteers; and 1.6 mm (SI) and 1.4 mm (AP) for the patient. Mean ADC values of VOI calculated from 4D-MRI (the healthy volunteer:1.5×10-3 mm2/s; the patient:1.7×10-3 mm2/s) were smaller than that calculated from free breathing DWI (the healthy volunteer:1.7×10-3 mm2/s; the patient:2.2×10-3 mm2/s). Entropy measurements of VOI calculated from 4D-MRI (the healthy volunteer:1.08; the patient:1.22) were also smaller than that calculated from free breathing DWI (the healthy volunteer:1.11; the patient:1.35). The ADC feature analysis results have tallied in general with the XCAT simulation results.

Conclusion

A respiratory correlated 4D-DWI technique has been developed and evaluated using digital phantom and human subjects. Comparing to free breathing DWI, 4D-DWI can lead to more accurate measurement of ADC. This has a great potential to improve the visualization and delineation of cancer tumors for radiotherapy.

Acknowledgements

This work is partly supported by funding from NIH (1R21CA165384) and a research grant from the Golfers Against Cancer (GAC) Foundation.

References

Reference 1. W. P. Segars, G. Sturgeon, S. Mendonca, J. Grimes and B. M. W. Tsui, “4D XCAT phantom for multimodality imaging research,” Med. Phys. 37, 4902-4915 (2010).

Reference 2. J. Cai, Y. Zhang, I. Vergalasova, F. Zhang, WP. Segars, F. Yin, “Developing a 4D Radiation Therapy Simulation System Based On a Realistic 4D Digital Human Phantom: Simulation of Imaging and Dose Delivery,” J. Cancer. Ther. 5,749-758 (2014).

Figures

Fig. 1. Illustration of 4D-DWI image acquisition scheme, reconstruction process and ADC measurements.

Fig. 2. Summary of XCAT phantom mean ADC value and entropy analysis for tumor region, using 10 liver patients breathing profiles to control the respiratory motion of XCAT.

Fig. 3. 4D-DWI images and motion trajectories of the healthy volunteer.

Fig. 4. ADC maps of the healthy volunteer.

Fig. 5. ADC maps of a liver cancer patient. The white arrows points out the lung tumor region.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
0624