Author: Jeremy T. Booth, Martin Andrew Ebert, Robert Finnegan, Annette Haworth, George Hruby, Burhan Javed, Kazi Ridita Mahtaba, Leyla Moghaddasi, Yutong Zhao 👨🔬
Affiliation: Northern Sydney Cancer Centre, Royal North Shore Hospital, The University of Sydney, The University of Western Australia, Genesis Care, Rockhampton Hospital 🌍
Purpose:
To evaluate the efficacy of an atlas-based tumor control probability (TCP) model in predicting prostate cancer (PCa) recurrence by retrospectively integrating patient-specific primary radiation therapy (RT) and histopathology data. The study explored segment-wise TCP model parameter adjustments to generate realistic TCP values to enhance recurrence prediction, based on histopathological findings in a patient cohort with biopsy-confirmed local recurrence.
Methods:
Nine patients with available histopathology reports and dose-fractionation schedules from primary cancer RT were selected from an ethics-approved study (NCT03073278) on re-irradiation of locally recurrent PCa. Two previously reported population-based biological atlases, comprising cell density data and tumor probability data respectively, were deformably registered to each patient’s prostate contour and segmented anatomically based on individual histopathology reports. Some PCa grade dependent and independent radiosensitivity parameters, including a single independent alpha/beta ratio, four Gleason Pattern (GP)-dependent alpha parameters, and nine Gleason Score (GS)-dependent alpha/beta ratios, were derived from a separate cohort's histology data using numerical optimization methods. The TCP model calculated overall TCP and generated voxel-wise TCP distribution maps for each patient. Three segment-wise histopathology-based parameter adjustment approaches (cell density alone, cell density with GP-dependent alpha, and cell density with GS-dependent alpha/beta) were compared to a baseline model without adjustments. Recurrent tumor contours were overlaid on TCP maps to evaluate alignment with lower TCP regions.
Results:
Adjustments combining cell density and GS-dependent alpha/beta ratios showed superior predictive capabilities, with significant reductions in overall TCP for all nine (100%) patients and alignment of lower TCP regions with relapsed tumor sites in seven (78%) patients. In contrast, GP-dependent alpha adjustments failed to predict recurrence, and cell density adjustments alone yielded moderate results.
Conclusion:
The atlas-based TCP model enhanced with patient-specific histopathology data shows potential for personalized treatment planning by identifying high-risk regions and optimizing dose distributions to reduce recurrence risk.