Portal de Periódicos da CAPES

A Software Platform for Rapid Translation of Functional MRI Findings to Radiation Treatment Planning for Prostate Cancer

2011; Elsevier BV; Volume: 81; Issue: 2 Linguagem: Inglês

10.1016/j.ijrobp.2011.06.1334

1879-355X

Radka Stoyanova, R. Garugu, Elizabeth Bossart, Xiaodong Wu, Alan Pollack,

Advanced MRI Techniques and Applications

Purpose/Objective(s)Dynamic Contrast-Enhanced Magnetic Resonance Imaging (DCE-MRI) and Diffusion Weighted Imaging (DWI) have a high sensitivity and specificity for tumor identification in three dimensions. We have developed a comprehensive method of handling the large amounts of imaging data from acquisition to analysis, to putting the images with delineated target areas into the treatment planning station (TPS) for GTV auto-delineation.Materials/MethodsA platform for importing and organizing imaging data was implemented in which multiple datasets can be stored and re-processed. Each dataset is downloaded directly from the RIS/PACS system. Both project and subject organization is maintained by an XML file structure which stores all the respective data and processing information. A software program was created in IDL/Java for analysis and visualization. The program automatically detects the dynamic T1 series and organizes the DCE-MRI data into a 4 dimensional array (3 spatial and one temporal). Principal Component Analysis (PCA) is carried out to identify and remove series with suboptimal quality, most often due to movement. An unsupervised pattern recognition (PR) technique identifies regions with the temporal pattern of fast wash-in followed by gradual wash-out of the contrast. This map is fused to the MRI-T1 series and the resultant DICOM files are pushed to TPS.ResultsWe have imported and analyzed 83 prostate cancer datasets in both the definitive (n = 49) and salvage settings (n = 34). DCE-MRIs typically contain 12 - 13 temporal series. The data are normalized by the pre-contrast MRI. In approximately 10% of the data it was necessary to exclude series due to artifact variations. Using PR we have identified the location and extent of tumor in 70% of the definitive and 20% of the salvage patients. Fitting the average curves from these regions using Tofts pharmoco-kinetic model is ongoing. We have transferred 23 datasets to the TSP (12 with intact prostate, 11 after prostatectomy) with delineated tumor volumes. Since the volumes are visualized on T1-MRI, the fusing with CT is straightforward. We tested the feasibility to utilize these volumes for delivering boost-dose radiation to the tumor lesion.ConclusionsOur software analysis facilitates visualization and GTV boost planning in a semi-automated process. The resultant 3D maps of the GTV should improve tumor targeting and thus realize gains in tumor control and reduced toxicity. Purpose/Objective(s)Dynamic Contrast-Enhanced Magnetic Resonance Imaging (DCE-MRI) and Diffusion Weighted Imaging (DWI) have a high sensitivity and specificity for tumor identification in three dimensions. We have developed a comprehensive method of handling the large amounts of imaging data from acquisition to analysis, to putting the images with delineated target areas into the treatment planning station (TPS) for GTV auto-delineation. Dynamic Contrast-Enhanced Magnetic Resonance Imaging (DCE-MRI) and Diffusion Weighted Imaging (DWI) have a high sensitivity and specificity for tumor identification in three dimensions. We have developed a comprehensive method of handling the large amounts of imaging data from acquisition to analysis, to putting the images with delineated target areas into the treatment planning station (TPS) for GTV auto-delineation. Materials/MethodsA platform for importing and organizing imaging data was implemented in which multiple datasets can be stored and re-processed. Each dataset is downloaded directly from the RIS/PACS system. Both project and subject organization is maintained by an XML file structure which stores all the respective data and processing information. A software program was created in IDL/Java for analysis and visualization. The program automatically detects the dynamic T1 series and organizes the DCE-MRI data into a 4 dimensional array (3 spatial and one temporal). Principal Component Analysis (PCA) is carried out to identify and remove series with suboptimal quality, most often due to movement. An unsupervised pattern recognition (PR) technique identifies regions with the temporal pattern of fast wash-in followed by gradual wash-out of the contrast. This map is fused to the MRI-T1 series and the resultant DICOM files are pushed to TPS. A platform for importing and organizing imaging data was implemented in which multiple datasets can be stored and re-processed. Each dataset is downloaded directly from the RIS/PACS system. Both project and subject organization is maintained by an XML file structure which stores all the respective data and processing information. A software program was created in IDL/Java for analysis and visualization. The program automatically detects the dynamic T1 series and organizes the DCE-MRI data into a 4 dimensional array (3 spatial and one temporal). Principal Component Analysis (PCA) is carried out to identify and remove series with suboptimal quality, most often due to movement. An unsupervised pattern recognition (PR) technique identifies regions with the temporal pattern of fast wash-in followed by gradual wash-out of the contrast. This map is fused to the MRI-T1 series and the resultant DICOM files are pushed to TPS. ResultsWe have imported and analyzed 83 prostate cancer datasets in both the definitive (n = 49) and salvage settings (n = 34). DCE-MRIs typically contain 12 - 13 temporal series. The data are normalized by the pre-contrast MRI. In approximately 10% of the data it was necessary to exclude series due to artifact variations. Using PR we have identified the location and extent of tumor in 70% of the definitive and 20% of the salvage patients. Fitting the average curves from these regions using Tofts pharmoco-kinetic model is ongoing. We have transferred 23 datasets to the TSP (12 with intact prostate, 11 after prostatectomy) with delineated tumor volumes. Since the volumes are visualized on T1-MRI, the fusing with CT is straightforward. We tested the feasibility to utilize these volumes for delivering boost-dose radiation to the tumor lesion. We have imported and analyzed 83 prostate cancer datasets in both the definitive (n = 49) and salvage settings (n = 34). DCE-MRIs typically contain 12 - 13 temporal series. The data are normalized by the pre-contrast MRI. In approximately 10% of the data it was necessary to exclude series due to artifact variations. Using PR we have identified the location and extent of tumor in 70% of the definitive and 20% of the salvage patients. Fitting the average curves from these regions using Tofts pharmoco-kinetic model is ongoing. We have transferred 23 datasets to the TSP (12 with intact prostate, 11 after prostatectomy) with delineated tumor volumes. Since the volumes are visualized on T1-MRI, the fusing with CT is straightforward. We tested the feasibility to utilize these volumes for delivering boost-dose radiation to the tumor lesion. ConclusionsOur software analysis facilitates visualization and GTV boost planning in a semi-automated process. The resultant 3D maps of the GTV should improve tumor targeting and thus realize gains in tumor control and reduced toxicity. Our software analysis facilitates visualization and GTV boost planning in a semi-automated process. The resultant 3D maps of the GTV should improve tumor targeting and thus realize gains in tumor control and reduced toxicity.