Patient‐specific voxel‐level dose prescription for prostate cancer radiotherapy considering tumor cell density and grade distribution

Abstract

Background: In prostate radiation therapy, recent studies have indicated a benefit in increasing the dose to intraprostatic lesions (IPL) compared with standard whole gland radiation therapy. Such approaches typically aim to deliver a target dose to the IPL(s) with no deliberate effort to modulate the dose within the IPL. Prostate cancers demonstrate intra-tumor heterogeneity and hence it is hypothesized that further gains in the optimal delivery of radiation therapy can be achieved through modulation of the dose distribution within the tumor. To account for tumor heterogeneity, biologically targeted radiation therapy (BiRT) aims to utilize a voxel-wise approach to IPL dose prescription by incorporating knowledge of the spatial distribution of tumor characteristics. Purpose:The aim of this study was to develop a workflow for generating voxelwise optimal dose prescriptions that maximize patient tumor control probability (TCP), and evaluate the feasibility and benefits of applying this workflow on a cohort of 62 prostate cancer patients. Method: The source data for this proof -of -concept study included high resolution histology images annotated with tumor location and grade. Image processing techniques were used to compute voxel-level cell density distribution maps. An absolute tumor cell distribution was calculated via linearly scaling according to published estimated tumor cell numbers. For the IPLs of each patient, optimal dose prescriptions were obtained via three alternative methods for redistribution of IPL boost doses according to maximization of TCP. The radiosensitivity uncertainties were considered using a truncated log-normally distributed linear radiosensitivity parameter (𝛼 k ) and compared with Gleason pattern (GP) dependent radiosensitivity parameters that were derived based on previously published methods. An ensemble machine learning method was implemented to identify patient-specific features that predict the TCP improvement resulting from dose redistribution relative to a uniform dose distribution. Results:The Gleason pattern-dependent radiosensitivity parameters were calculated for 20 published prostate cancer 𝛼∕𝛽 ratios. Optimal voxel-level dose prescriptions were generated for all 62 PCa patients. For all dose redistribution scenarios, the optimal dose distribution always shows a higher (or equivalent) TCP level than the uniform dose distribution. The applied random forest regressor could predict patient-specific TCP improvement with low root mean square error (≤1.5%) by using total tumor number, volume of IPLs and the standard deviation of tumor cell number among all voxels.

Towards optimal heterogeneous prostate radiotherapy dose prescriptions based on patient‐specific or population‐based biological features

“…The methodology for data extraction and processing has already been described in detail in a previous publication. 15 In summary, by processing digitized whole mount prostate histology slides, 18 tumor annotation masks of 9 Gleason Scores (GS) (2 + 2, 3 + 2, 3 + 3, 3 + 4, 4 + 3, 4 + 4, 4 + 5, 5 + 4, 5 + 5) 18 and cell density maps (CD-maps) indicating area cell density distributions 19 were prepared for all patients in the cohort. To generate a 3D volume, the histology slides, which were obtained 5.0 mm apart, were co-registered with corresponding ex vivo T2 weighted (T2w) MRI of the prostate specimens which had 2.5 mm axial slice thickness, and linear interpolation was applied to generate the missing histology data from the CD-maps and tumor annotation masks.…”

“…14 This work was later extended for a 63 patient cohort, and a cell density distribution, tumor location probability and grading of intra-prostatic lesions derived from prostate histology data was used to yield the optimal voxel-level dose distributions to guide the generation of treatment plans. 15 In a clinical situation, spatial maps of tumor characteristics derived from histology data, as used by Her et al 14 and Zhao et al, 15 cannot be obtained prior to optimal dose distribution generation. Machine learning estimates based on multi-parametric MRI (mpMRI) data can be used to estimate the data in vivo.…”

“…By incorporating the work conducted by Finnegan et al, 17 this research aimed to further expand the studies of Her et al 14 and Zhao et al 15 to generate optimal dose prescriptions without, or partially with, the distribution of population histology data using four fractionation schedules for whole prostate gland radiotherapy. Heterogeneous dose prescriptions were generated and compared with spatially homogeneous prescriptions as well as heterogeneous prescriptions generated via precision (i.e., patient-specific) histology data.…”

“…In the previous research of Her et al., 14 biologically targeted radiation therapy (BiRT) was introduced whereby the optimal redistribution of dose according to patient‐specific cell density distributions for five patients was determined for intensity‐modulated radiotherapy (IMRT) 14 . This work was later extended for a 63 patient cohort, and a cell density distribution, tumor location probability and grading of intra‐prostatic lesions derived from prostate histology data was used to yield the optimal voxel‐level dose distributions to guide the generation of treatment plans 15 …”

“…In a clinical situation, spatial maps of tumor characteristics derived from histology data, as used by Her et al 14 . and Zhao et al., 15 cannot be obtained prior to optimal dose distribution generation. Machine learning estimates based on multi‐parametric MRI (mpMRI) data can be used to estimate the data in vivo 16 .…”