TOPIC 417: GPU-ACCELERATED 3D MONTE CARLO SPECT RECONSTRUCTION ALGORITHM FOR PERSONALIZED RADIOPHARMACEUTICAL THERAPY

尽管放射性药物治疗（RPT）在淋巴瘤、晚期、转移性前列腺癌和神经内分泌肿瘤患者中效果良好，但众所周知，由于患者药代动力学的变异性，标准剂量导致难以预测且差异巨大的临床结局。然而，通过首先使用定量SPECT成像测量由RPT试剂在体内发射的辐射，然后使用剂量测定软件计算沉积在肿瘤和正常组织中的辐射能量，可以使RPT更安全和更有效。可以得出个性化的RPT处方，使有效性最大化，副作用最小化。然而，由于商业上可用的SPECT重建算法不准确地校正散射，它们不足以确定个性化的RPT处方。因此，对于能够实现真正的定量SPECT成像的更好的重建算法存在临床上未满足的需求。该合同提案的目的是构建一个图形处理单元（GPU）软件平台，该平台可以在5分钟内使用基于蒙特卡罗（MC）的散射校正执行SPECT重建，使其在临床上可行。快速准确的GPU MC SPECT重建（Torch Recon）将进一步提高RPT剂量测定的准确性，从而加速我们目前由SBIR资助的MC剂量引擎（Torch）的临床采用。

Although radiopharmaceutical therapy (RPT) has worked well in patients with lymphoma, late-stage, metastatic prostate cancer, and neuroendocrine tumors, it is well known that because of variability in patient pharmacokinetics, standard dosing leads to clinical outcomes that are difficult to predict and drastically vary. However, it is possible to make RPT safer and more effective by first measuring the radiation emitted by the RPT agent in vivo using quantitative SPECT imaging and then calculating the radiation energy deposited in tumors and normal tissues using dosimetry software. A personalized RPT prescription can be derived that maximizes effectiveness and minimizes side effects. However, because commercially available SPECT reconstruction algorithms do not accurately correct for scatter, they are insufficient for determining personalized RPT prescriptions. Therefore, there is a clinically unmet need for better reconstruction algorithms which enable true quantitative SPECT imaging. The aim of this contract proposal is to build a graphics processing unit (GPU) software platform that can perform SPECT reconstruction with Monte Carlo (MC)-based scatter correction within 5 minutes, making it clinically viable. Fast and accurate GPU MC-based SPECT reconstruction (Torch Recon) will further improve the accuracy of RPT dosimetry and thereby accelerate the clinical adoption of our current SBIR-funded MC dose engine (Torch).