Enhancing the efficacy of radiation therapy by incorporating spatial distribution of heterogeneous biology

Abstract

It is well known that the distribution of clonogen (or functional units), radiosensitivity, and cell proliferation rate in most tumors and sensitive structures are heterogeneous. Recent progress in biological imaging is making the mapping of spatial biology distribution increasingly possible. To optimize efficacy of radiotherapy and in particularly intensity modulated radiation therapy (IMRT), it is desirable to take the biological information into account and to produce customized non-uniform dose distributions on a patient specific basis. The purpose of this work is to establish a theoretical framework to quantitatively incorporate the spatial biology data derived from biological images into IMRT inverse optimization and to show its advantage in enhancing the tumor control probability (TCP) and reducing the normal tissue complication probability (NTCP). A linear quadratic (LQ) model was used as a starting point of the study. According to the LQ model, the status of the tumor at a spatial point was characterized by clonogen density (?), radiosensitivity (?), and proliferation rate (?). For sensitive organs, the functional unit density was employed to describe the functional importance at a point within the given structure. We hypothesize that the biological model parameter distributions can be derived from the biological imaging data (to give a few examples, clonogen density in malignant glioma can be obtained based the choline/creatine ratio by using MR spectroscopic imaging, radiosensitivity depends on the local oxygen level, the tumor proliferation rate is proportional to the voxel activity level in the case of DNA proliferation imaging (e.g., FLT-PET), and functional importance distribution can be obtained by perfusion imaging). The desired spatial variation form of the radiation dose (or prescription dose distribution) depends on (?, ?, ?) distributions and is determined by maximizing the TCP under the constraint of constant integral tumor dose. Using the above definition and the LQ model, we deduced a general formula that allows us to prescribe dose to each tumor voxel in an automatic fashion. An objective function with the voxel-dependent prescription described above was then constructed. A group of voxel-dependent important factors were also introduced to consider the functional unit density distribution in sensitive structures. A conjugate-gradient algorithm was implemented for the optimization of the objective function. A brain tumor case and a prostate case with biology information derived from functional imaging were used to test the new inverse planning system. For comparison, conventional IMRT plans with uniform dose prescription were also generated for these two cases. The biological indices, TCP and NTCP, were used to evaluate the proposed technique and to show its superiority over the conventional planning scheme. An IMRT inverse planning framework has been established for dealing with spatially heterogeneous biology distribution. The influence of the model parameters, ?, ?, and ?, on the final plans were studied and the study seems to indicate that the ? and ? play more important role in comparison to the tumor cell density. Our study also demonstrated that it is technically feasible to produce deliberately non-uniform dose distributions with consideration of biological information. The biological model established in this study was found to be useful for determining the prescription dose distribution. Compared with the conventional inverse planning scheme, the new system was capable of generating IMRT plans with much higher TCPs while keeping the NTCPs at their conventional levels. Incorporating patient-specific biological information into IMRT inverse optimization provides a significant opportunity to truly individualize radiation treatment. It allows us to utilize maximally the technical capacity of IMRT and provides us with much improved means for safe and intelligent dose escalation.