Purpose or Case Report: Percutaneous CT-guided biopsy and ablation are established techniques to diagnose and treat solid tumors. However, some tumors are not adequately visualized with CT, even after IV contrast administration, because of transient enhancement. Many of these “CT-occult” tumors are optimally seen on preprocedural MRI or PET scans. To enable targeting of such tumors, we developed software that fuses MRI/PET images to intraprocedural CT in real time. The software corrects for soft-tissue deformations which occur due to differences in patient positioning or motion. The fusion imaging is generated after each intraprocedural CT scan and takes advantage of all available imaging data. We report retrospective evaluation of this fusion method for pediatric interventional radiology procedures.
Methods & Materials: The software was tested on four patients with CT-occult tumors undergoing biopsy. Two patients had bone lesions (Salmonella osteomyelitis and Ewing Sarcoma) that were optimally visualized on preprocedural MRI and PET. Two patients with neurofibromatosis had soft-tissue lesions with regions that were suspicious for malignant transformation best seen on preprocedural MRI and PET. During biopsy, CT images were pushed from the scanner to the workstation and fused with the corresponding MRI or PET. The time and accuracy of image fusion were measured. Accuracy was measured as target registration error (TRE) at anatomic landmarks identified by an expert in individual images while blinded to the fusion images.
Results: Clinical experts reviewed the fusion images after the procedures and judged them to be visually adequate and accurate. The mean TRE of 4.1 mm supported this finding. The TRE data further showed that the algorithm achieved subvoxel registration accuracy (registration error < voxel size) of 0.85 voxels, averaged over all four cases. The mean time of image data transfer from CT scanner to the fusion workstation was 2.0 s for volumetric and 0.4 s for fluoroscopic scans. The mean time to perform image fusion was 35.7 s for volumetric scans and 7.0 s for fluoroscopic scans.
Conclusions: The image fusion software was fast and allowed accurate visualization of MRI/PET visible tumors on the intraprocedural CT images. When fully integrated in the IR workflow, this tool has the potential to provide improve targeting of CT-occult tumors and improve efficiency and efficacy for CT-guided biopsy and ablation. A prospective clinical trial is planned to quantify the anticipated clinical benefits.