Contour-guided deep learning based deformable image registration for dose monitoring during CBCT-guided radiotherapy of prostate cancer

Abstract

Purpose: To evaluate deep learning (DL)-based deformable image registration (DIR) for dose accumulation during radiotherapy of prostate cancer patients.

Methods and materials: Data including 341 CBCTs (209 daily, 132 weekly) and 23 planning CTs from 23 patients was retrospectively analyzed. Anatomical deformation during treatment was estimated using free-form deformation (FFD) method from Elastix and DL-based VoxelMorph approaches. The VoxelMorph method was investigated using anatomical scans (VMorph_Sc) or label images (VMorph_Msk), or the combination of both (VMorph_Sc_Msk). Accumulated doses were compared with the planning dose.

Results: The DSC ranges, averaged for prostate, rectum and bladder, were 0.60-0.71, 0.67-0.79, 0.93-0.98, and 0.89-0.96 for the FFD, VMorph_Sc, VMorph_Msk, and VMorph_Sc_Msk methods, respectively. When including both anatomical and label images, VoxelMorph estimated more complex deformations resulting in heterogeneous determinant of Jacobian and higher percentage of deformation vector field (DVF) folding (up to a mean value of 1.90% in the prostate). Large differences were observed between DL-based methods regarding estimation of the accumulated dose, showing systematic overdosage and underdosage of the bladder and rectum, respectively. The difference between planned mean dose and accumulated mean dose with VMorph_Sc_Msk reached a median value of +6.3 Gy for the bladder and -5.1 Gy for the rectum.

Conclusion: The estimation of the deformations using DL-based approach is feasible for male pelvic anatomy but requires the inclusion of anatomical contours to improve organ correspondence. High variability in the estimation of the accumulated dose depending on the deformable strategy suggests further investigation of DL-based techniques before clinical deployment.

Keywords: CBCT; deformable registration; dose monitoring; prostate.

FIGURE 1

Three‐fold cross‐validation of the VoxelMorph models. Patients were randomly distributed in training, validation and test datasets.

FIGURE 2

Geometric evaluation of the DIR methods for (a) DSC, (b) ASD, (c) HD, and (d) 95th HD metrics for the prostate, bladder, and rectum of the test dataset. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001, adjusted analysis of variance with Tukey HSD test. Absence of * means difference was not statistically significant. DSC, dice similarity coefficient; HD, Hausdorff distance; MAD, mean absolute distance.

FIGURE 3

Qualitative evaluation of deformations per method for a median case in terms of DSC. The first row contains the planning and CBCT images. The second to fourth rows contain the Jacobian determinant and deformed CBCT images. The green, yellow, and light blue colors represent the prostate, bladder, and rectum, respectively. The image intensity parameters were 2000 and 0 for the window and level, respectively. The determinant of Jacobian intensity parameters were 2 and 0 for the window and level, respectively.

FIGURE 4

Dosimetric evaluation of accumulated doses per DIR method compared to the planning dose for the prostate, bladder, and rectum of the test dataset. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001, adjusted analysis of variance with Tukey HSD test. Absence of * means difference was not statistically significant.

FIGURE 5

Dosimetric evaluation of accumulated dose differences with the planning dose per DIR method for the prostate, bladder, and rectum of the test dataset.

FIGURE 6

Qualitative evaluation of accumulated doses per DIR method for the overall median case regarding DSC. The first row contains the planning CT and dose. The second and third rows contain the accumulated dose per method with the planning contours. The green, yellow, and light blue colors represent the prostate, bladder, and rectum, respectively. The dose intensity parameters were 80 and 40 in Gy for the window and level, respectively.