Radiation and Cancer Physics
SS 27 - Physics 7 - Special Session: Outcome Analysis and Modeling
197 - A New Paradigm for Radiation-Induced Toxicity Analysis: Space Based Normal Tissue Complication Probability Modeling
Tuesday, October 23
3:55 PM - 4:05 PM
Location: Room 214 C/D
Giuseppe Palma, PhD
National Council of Research (Institute of Biostructures and Bioimaging)
No relationships to disclose.
A New Paradigm for Radiation-Induced Toxicity Analysis: Space Based Normal Tissue Complication Probability Modeling
G. Palma1, A. Buonanno2, S. Monti3, R. Pacelli4, and L. Cella1; 1National Research Council, Institute of Biostructures and Bioimaging, Napoli, Italy, 2Università degli Studi della Campania "Luigi Vanvitelli", Dipartimento di Ingegneria Industriale e dell'Informazione, Aversa, Italy, 3IRCCS SDN, Napoli, Italy, 4Federico II University School of Medicine, Department of Advanced Biomedical Sciences, Napoli, Italy
Purpose/Objective(s): An increasing awareness have been triggered in the RO community regarding the value of including the spatial information of dose distributions within the analysis of radiation-induced toxicity (RIT). Up to now, statistical inference on spatial signature of RIT has been addressed [Acosta 2013, Dean 2016], but no insight on actual NTCP models including full dose spatial info has been achieved. The purpose of this study is to propose a new formalism to fill this gap of knowledge and to develop a space based NTCP (SpNTCP).
Materials/Methods: A SpNTCP approach was devised assuming the availability of a training set of N patients’ dose distributions (sampled on m voxels each) classified according to a binary RIT. A logistic regression on the dose values is performed voxelwise, and, accordingly, a collection of m RIT probabilities is computed on each voxel dose for a test dose distribution. Similarly, a collection of m weights is obtained as the inverse of each logistic Confidence Interval (CI) length. The SpNTCP is then computed as the weighted mean of the m RIT probabilities. The SpNTCP was first demonstrated on two sets of synthetic dose maps (with different and equal DVHs, respectively), classified by thresholding the generalized Equivalent Uniform Dose (gEUD) simulated for given Radio-Sensitivity Map (RSM) and n volume-effect parameter. Moreover, an estimate of the RSM was reconstructed by computing the previous SpNTCP on a new set of hot spots evenly spanning the field of view. The performance was evaluated on a cohort of 98 RT Hodgkin lymphoma survivors classified for lung fibrosis (18 events). For this purpose, each dose map was normalized to a common anatomical reference via a log-diffeomorphic demons registration tool [Palma 2016]. Lyman-Kutcher-Burman (LKB) models were trained for comparison.
Results: Asymptotic values of learning curves show that the SpNTCP model outperformed LKB one, with an increasing model bias at lower n values (Table). On the real patients, the area under the ROC curve of the SpNTCP was 0.72 (CI: [0.62-0.81]), significantly higher than 0.61 (CI: [0.51-0.70]) obtained by the LKB (p<0.002).
|Accuracy of models |
| ||Different DVHs ||Equal DVHs |
|n ||SpNTCP ||LKB ||SpNTCP ||LKB |
|.1 ||.77 ||.81 ||.77 ||.5 |
|.25 ||.85 ||.77 ||.81 ||.5 |
|.5 ||.90 ||.76 ||.85 ||.5 |
|1 ||.93 ||.78 ||.88 ||.5 |
Conclusion: The proposed SpNTCP model, differently from LKB, proved able to distinguish the toxicity effects of given dose distributions in presence of inhomogeneous RSM, even for equal DVHs. Interestingly, our approach also proved able to map a posteriori the organ RSM, by calculating the SpNTCPs on a set of dose distributions with a single hot spot, whose position spans the FOV. As expected, the performance of the SpNTCP model, at least in this preliminary format, decreases as the volume effect decreases. Of note, the SpNTCP modeling of real lung RIT was more accurate than LKB, seemingly in agreement with the evidences of inhomogeneous RSM of the lungs. In conclusion, we believe that the proposed germinal idea could pave the way toward a new NTCP modeling philosophy.
Author Disclosure: G. Palma: None. A. Buonanno: None. R. Pacelli: None.