Purpose

Internal target volume (ITV) is defined to accommodate tumor motion in radiotherapy treatment planning (RTP), and is commonly determined based on 4D-CT images. But, the reliability of the 4D-CT determined ITV for cancer in the abdomen is much limited by the intrinsic low soft-tissue image contrast of CT and the short scan time (to minimize radiation burden), particularly for irregular respirations. By taking the advantages of the non-ionizing radiation nature of MRI as well as its superior soft-tissue image contrasts, 4D-MRI is anticipated to be a valuable tool for radiotherapy for ITV determination in treatment planning and inter-fractional ITV verification [1]. In this study, we propose a novel ITV determination technique based on the volume of the positional probability distribution (PPV) estimated using 4D-MRI.

Image acquisition

4 abdominal cancer patients (3 males, 1 females, age = 68±16 years, 5 lesions) were recruited. 4D-CT images were acquired on a CT simulator with a camera system to monitor the abdominal AP shift as the external surrogate of respiratory motion (120kV, 50 mAs, 2mm slice thickness, scan time = 60~70s). The 4D-CT images were retrospectively sorted into 10 respiratory phases by the external respiratory surrogate system. The ITV was delineated by an oncologist to cover the gross target volume (GTV) on 4D-CT in all ten phases with respect to the mid-inhalation position. Free-breathing abdominal volumetric 4D-MRI were axially acquired on a 1.5T MR-simulator with same positioning and immobilization as 4D-CT (CAIPIRINHA-TWIST-VIBE, TE/TR=0.35/1.57ms, flip angle=5o, FOV=350-382mm, acquisition matrix size= 128x128; slices=52-64; slice-thickness=4.5-5mm; reconstruction voxel size=2.7x2.7x4.5 - 3.0x3.0x5.0mm3, phase FOV= 71.9%; partial Fourier factor=75% (phase encoding)/62% (slice encoding); slice over-sampling factor=20%; central and 33% peripheral K-space sampling percent by TWIST; frame-rate = 2.5-3.0 frame-per-second (fps); 144 time-frames, acceleration factor=4).

PPV generation

GTV was masked on the first time frame of the 4D-MRI to generate the baseline anatomical segmentation (Seg₁). Images at each frame j (j>1) were non-linearly registered to the first time frame to obtain the transformation matrix, which was applied to Seg1 for the generation of Seg_j. The positional probability (PP) of a voxel in the imaged volume was calculated by the number of time frames that a voxel had been occupied by the organ-of-interest divided by the total elapsed frames, as presented in Eq.1, where j denotes the time frame index, and M is the binary mask value of 1 (the voxel is occupied by Seg_j) or 0 (the voxel is not occupied by Seg_j). rPPj=(M₁+...+M_j)x100%/j if Mj∈Seg_j, then M_j=1; if M_j∉Seg_j, then M_j=0 [Eq.1]
Positional probability volume (PPV) was estimated based on the relative frequency of the voxels in the imaged volume being occupied by the organ-of-interest during the 4D-MRI scan, where rPPV_j,threshold corresponds to the number of voxels with the PP_j value larger than the pre-defined threshold.

Statistical analysis

Subject-specific PPV_144,0% and PPV_144,5% of the GTV were compared with the ITV using signed-rank test. The similarity between PPVs and ITV was also evaluated based on the dice similarity coefficient (DSC), the volumetric difference (Diff_ITV,i%) and Hausdorff distance (HD).

Results and discussion

As illustrated in Table 1, the ITV determined by 4D-CT was generally larger than the PPV_144,0% (mean Diff_ITV,0% = 23%) and rPPV_144,5% (mean Diff_ITV,5% = 36%) derived from 4D-MRI, indicating the overestimation of respiratory motion by 4D-CT. Despite the small patient sample size, the volume difference between ITV and PPV_144,0% and PPV_144,5% were just marginally insignificant(p=0.0625). ITV and PPV were similar in larger tumors with relatively large DSC (>0.80), suggesting the efficacy of 4D-MRI for respiratory motion characterization. In contrast, small DSCs (<0.70) were found between the ITV and PPV in particular for small tumors, implying the possible inaccuracy of 4D-CT in respiratory motion estimation resulted from poor tumor visibility.
Besides the small sample size, the study limited in the PPVs and ITV comparison based on the possibly different respiratory profiles, a common problem for all in vivo 4D-MRI studies. Further study is warranted to increase the sample size to strengthen statistical power, and to compare the PPV and the 4D-CT ITV with the same respiratory profile based on an MR-CT compatible motion phantom.

Conclusion

A novel volumetric 4D-MRI derived rPPV approach was proposed for ITV determination and preliminarily validated in a small cohort of real patients. This rPPV approach should be potentially useful in patient-specific ITV determination in abdominal precision radiotherapy.

Acknowledgements

No acknowledgement found.

References

[1] Paulson ES, Ahunbay E, Chen X, Mickevicius NJ, Chen GP, Schultz C, Erickson B, Straza M, Hall WA, Li XA. 4D-MRI driven MR-guided online adaptive radiotherapy for abdominal stereotactic body radiation therapy on a high field MR-Linac: Implementation and initial clinical experience. Clin Transl Radiat Oncol. 2020 May 15;23:72-79.