A unified workflow for classifying patterns of locoregional failure using radiotherapy treatment planning dose distributions

Abstract

Objectives: This work describes a unified workflow for classifying patterns of locoregional recurrence (LRR) using radiotherapy planning dose distributions. This approach aims to incorporate dose parameters into LRR classifications and facilitate application across different treatment sites and dose prescriptions to standardise classification terminology.

Methods: The relapse diagnostic CT (rCT) and manually delineated relapse gross tumour volume (rGTV) were co-registered with the radiotherapy planning CT (pCT) using deformable image registration (DIR). The DIR accuracy was quantified using the target registration error (TRE) using the absolute centroid distance between cancer site-specific regions of interest (ROIs). Dosimetric structures were delineated for planning regions receiving 95% of the dose prescribed to high-risk, intermediate-risk, and low-risk CTVs, relative to the cancer site or trial. The mapped rGTV was compared relative to each dose structure and classified into one of five categories: central and peripheral high-dose (Type A, Type B), central and peripheral elective-dose (Type C, Type D), and extraneous dose (Type E) failures.

Results: The unified workflow was successfully implemented on two different patient use cases, one from the IMPORT HIGH breast cancer trial, one from the VoxTox head-and-neck study, classifying LRR as Type A and Type E failures, respectively.

Conclusion: This workflow for classifying LRR is applicable across different cancer sites, despite differences in treatment protocol, target dose, and dose delivery. This provides a basis for utilising radiotherapy dose distributions to analyse patterns of failure irrespective of trial design or cancer-site.

Advances in knowledge: Standardised classifications of LRR that are correlated with the planning dose distribution could provide insight into the underlying causes of LRR burden post-radiotherapy and allow for critical evaluation of the current concepts of defined clinical tumour volumes and optimal PTV dose regions.

Keywords: image analysis; image registration; local recurrence; methodology; radiotherapy.

Figure 1.

The classification scheme for classifying LRR in head and neck patients based on geometric (centroid) and dosimetric (volume) parameters into categories A-E from high to extraneous dose failures.

Figure 2.

Registration results for Patient X. (A) The fixed image (pCT) and the final result of the transformed moving image (rCT) using B-spline deformation. (B) Overlay of the pCT sternum ROI and spatially transformed rCT sternum ROI. Absolute centroid distance used to calculate individual patient TRE. (C) Perform registration and calculate cohort TRE for all patients.

Figure 3.

Recurrence structure sets and dose structures for Patient X. (A) Clinical tumour bed volume, mapped rGTV volume as outlined by the radiation oncologist, and the rGTV sphere generated around the mapped rGTV centroid. (B) 95% dose structures generated from the planned treatment dose distribution for 95% dose to high-risk (TB), intermediate-risk (PB) and low-risk (WB) regions to capture dose spread outside clinical CTVs. (C) The planning CT with overlaid 95% dose structures and spatially mapped recurrence sphere and centroid. This demonstrates the difference in classification between a CTV-based comparison (A) and dose structure comparison (C).

Figure 4.

Registration results for Patient Y. (A) The fixed image (pCT) and the final result of the transformed moving image (rCT) using B-spline deformation. (B) Overlay of the pCT thyroid cartilage ROI and spatially transformed rCT thyroid cartilage ROI. Absolute centroid distance used to calculate individual patient TRE. (C) Perform registration and calculate cohort TRE for all patients.

Figure 5.

Structure sets for Patient Y. (A) Clinical tumour bed volume, mapped rGTV volume as outlined by the radiation oncologist, and the relapse sphere generated around the mapped rGTV centroid. (B) 95% dose structures generated from the planned treatment dose distribution for 95% dose to high-risk region (CTV1) to capture dose spread outside the clinical CTV. (C) The planning CT with overlaid 95% dose structure and spatially mapped recurrence sphere and centroid.