Tensor Regression-based Model to Investigate Heterogeneous Spatial Radiosensitivity After I-125 Seed Implantation for Prostate Cancer.

BACKGROUND/AIM: We established a data-driven method for extracting spatial patterns of dose distribution associated with radiation injuries, based on patients with prostate cancer who underwent iodine-125 (I-125) seed implantation.

PATIENTS AND

METHODS

Seventy-five patients underwent I-125 seed implantation for prostate cancer. We modeled the severity of lower urinary tract symptoms (LUTS) to be estimated using a linear model, which is formulated as an inner product between the dose distribution D and voxel-wise radiosensitivity B inside the prostate. For the estimation, tensor regression based on a low-rank decomposition with generalized fused lasso penalty was applied.

RESULTS

The spatial distribution of B was visually assessed. Positive parameters appeared dominantly in the region close to the urethra and the prostate base.

CONCLUSION

Our tensor regression-based model can predict intra-organ radiosensitivity in a data-driven manner, providing a compelling parameter distribution associated with the development of LUTS after I-125 seed implantation for prostate cancer.