Prostate Magnetic Resonance Imaging Delta-Radiomics during Image-Guided Conventional Fractionation and Stereotactic Ablative Radiotherapy

Abstract

<h3>Purpose/Objective(s)</h3> The ability to detect changes during radiotherapy to the prostate with radiomics techniques may allow prediction of clinical outcomes and influence subsequent treatment choices. In this feasibility study, we aimed to assess changes in features (delta-radiomics) across a course of treatment in men undergoing MRI-guided radiotherapy for prostate cancer. We compared changes between conventional fractionation and ultrahypofractionated SABR treatment. <h3>Materials/Methods</h3> MRI images from men receiving radiotherapy only to the prostate on an MR Linac developed by a precision radiation medicine company in a large tertiary cancer center were analyzed. Treatment localization scans were transferred from one treatment planning system, (from a precision radiation medicine company, Crawley UK) to a different treatment planning system. The whole prostate was re-contoured from T2 sequences obtained at fraction 1 and at 4 evenly spread time points. For SABR this was at fractions 2 to 5 and for conventional fractionation at fraction 5, 10, 15 and 20 respectively. Radiomics data was extracted with an open-source software platform. Data for energy, entropy, kurtosis and shape parameters were analyzed using a linear mixed-effects model to explore changes over time within a patient (within subject variability) as well as between subject variability between men treated with conventional fractionation v SABR. <h3>Results</h3> 40 men were analyzed, 20 receiving conventional fractionation (60Gy in 20 fractions over four weeks) and 20 receiving SABR (36.25Gy in 5 fractions over two weeks) respectively. Mean age was 69.3years (range 53-82) and mean PSA at diagnosis 9.37 (range 2.18 – 32). Gleason score was 6 in 1 patient, 7 in 38 patients and 8 in 1 patient. Five men did not receive any androgen deprivation therapy prior to radiotherapy (3 patients receiving conventional fractionation, 2 receiving SABR). The only radiomics variable that showed a time-trend was first order entropy in the conventional fractionation group, not in the SABR group (p<0.001). Treatment type was not a significant covariate for any of the other metrics. I.e., in long fractionation a reduction in entropy is observed during treatment, showing the prostate gray values become more homogeneous over time. <h3>Conclusion</h3> Our analyses showed a reduction in entropy for conventional fractionation rather than ultrahypofractionated SABR suggesting that image changes only become apparent over longer treatment courses. During SABR, no radiomics features showed a significant trend. These may still develop on later imaging as changes in cell appearance from DNA damage may only be apparent over time. Further analyses are planned to explore the relationship between delta radiomics and clinical outcomes and whether entropy can be used as an imaging biomarker.