571

Abstract No. 199

Abstract No. 200

--.....51 R.!..~.~E~I!.~

1), and uncharacterized (n =3). Average tumor diameter was3.0 cm (1.8-4.1cm). The procedure was technically successfulin all patients. Eight (62%) patients developedpneumothoraces with three (23%) requiring chest tubeplacement. Three patients (23%) required admission post­procedure: two patients were admitted overnight, while onepatient required a 4 day admission for respiratorydecompensation. Currently, we have a mean follow-up intervalof 6 months (range 1-14 months) with either CT and/or PETscans. There were no treatment-related deaths. Follow-upscans have not shown any evidence of residual unablatedtumor at the treatment site in any of the thirteen patients.

CONCLUSION: Percutaneous radiofrequency ablation inconjunction with brachytherapy is a promising minimallyinvasive treatment modality providing local tumor eradicationfor medically inoperable patients with lung cancer.

Oncology Ablation

CONCLUSION: Three-dimensional ultrasound was integratedwith CT using a stereotactic mechanical frame, taking advantageof temporal resolution of US and spatial resolution of CT.

Registration and Fusion of Multi-Slice CT and Real­Time 3-D Ultrasound for RF Ablation of Liver Tumors.B.J. Wood, NIH, Bethesda, MD, USA 'J. Locklin 'J. Kruecker·A.Viswanathan .z. Neeman '. Yano/, et at.

PURPOSE: A preliminary system for registering and fusing3D-US with retrospective CT data sets is described.

MATERIALS AND METHODS: This allows pre-procedureplanning and intra-procedural instrument tracking ofpercutaneous interventions such as RFA. For registration, theUS probe is mechanically attached to a 3-D tracking devicewhich is seamlessly pre-registered into CT coordinates. Asthe US probe is positioned on the surface of an ex vivo liverand phantom, two perpendicular CT MPR images arereconstructed to be coplanar with 3D-US bisectional images.

RESULTS: Needle positioning is demonstrated with singlem'odalities and with the registered display during RFA.Methods to measure registration accuracy and reasons formisregistration, such as organ shift, will be presented. Colormaps and alpha blending methods for image fusion will bereviewed. The complementary aspects of 3D-US and CTduring needle positioning, tissue movement, critical structureavoidance, and gas bubble formation will be illustrated.

Radiofrequency Ablation of Renal Masses: Methods ofBowel Protection for Tumors within 1 em of Bowel.D.A. Gervais, Massachusetts General Hospital, Boston, MA,USA ·R.S. Arellano·F.J. McGoverll'WS. McDougal·P.R.Mueller

PURPOSE: The need to protect bowel during RFA hasreceived attention recently, but absolute safe distances fromtumor to bowel have not been empirically established. Thepurpose of this study was to review our series of RCCstreated with RFA focusing on strategies for bowel protectionfor tumors within 1 cm of bowel.

MATERIALS AND METHODS: From 1998 to 2004, 100RCCs in 82 patients underwent RFA. Medical records andimaging were reviewed to determine bowel site in nearest

Abstract No. 197

Abstract No. 198

Oncology

"Multimodality Roadmapping" with ElectromagneticTracking for IR Procedures: From Bench to Bedside.B.J. Wood, National Institutes of Health / National CancerInstitute, Bethesda, MD, USA·N. Glossop·A. Viswanathan 'J.Kruecker '. Banovac'K. Cleary, et al.

PURPOSE: To assess the feasibility of electromagnetictracking for navigation of IR devices in phantoms, animals,and patients.

MATERIALS AND METHODS: Needles, catheters,guidewires, and ablation devices were constructed with smallposition sensors in the tips. A tetrahedral weak magnetic fieldgene~ator was placed in proximity to phantom, pig, or patient.MR, PET, or CT scans of phantom or pig were preloaded incustomized tracking, registration, and display software.Needles were manipulated within the phantom, pig, or patientwith guidance from the previously-acquired pre-proceduralimaging.

RESULTS: The CT scan position of the devices wasdetermined within phantoms, pigs and humans for variousprocedures and simulated procedures (liver ablation, biopsy,selective angiography, selective catheterization). Pre­procedural imaging in the region of the devices was displayedwith real-time updating of the device position. Accuracy inphantom models was - 1-3 mm, although less accuracy wasseen in breathing pigs and patients.

CONCLUSION: CT, MR or PET data can be displayed duringbiopsy, ablation, or selective angiographic procedures withreconstructions displaying the position of the devices withinthe pre-procedural imaging. This system enables"multimodality road mapping", and holds promise formulttffiodality interventions by allowing the real-time use ofpreviously acquired imaging during interventional procedures.

Percutaneous Radiofrequency Ablation Followed byBrachytherapy in Inoperable Stage 1 Non Small CellLung Cancer.C.J. Simon, Rhode Island Hospital/Brown Medical School,Providence, RI, USA-T.A. Dipetrillo·S. Gandhi·T. Ng'WWMayo-Smith 'DE Dupuy

PURPOSE: To determine the safety and efficacy ofradiofrequency ablation (RFA) followed by high dose ratebrachytherapy for the treatment of medically inoperable stage1 non small cell lung cancer (NSCLC).

MATERIALS AND METHODS: Medically inoperable patientswith biopsy proven NSCLC underwent CT-guided RFAfollowed by brachytherapy. Patients were deemed medicallyinoperable due to poor cardiopulmonary reserve or coexistentmorbidities. Patients were staged pre-therapy with CT andFDG-PET scans. Percutaneous RFA was performed with asingle (n = 1) or cluster (n = 12) cool-tip RF electrode and aCC-l generator (Radionics Inc., Burlington, MA). Immediatelyfollowing RFA, a brachytherapy catheter was inserted directlyinto the tumor via a vascular sheath (once the central needlewas removed) and robot after-loaded iridium-192 (I92Ir) seedsdelivered a single fraction dose of 16-18Gy. Patients werefollowed-up with CT and/or PET scans at regular intervals.

RESULTS: We treated a total of 13 patients, seven men andsix women, with a mean age of74.5 years (60-86 years). Thehistologic tumor subtypes were: squamous cell carcinoma (n=5), adenocarcinoma (n =4), bronchoalveolar carcinoma (n =---

S72

proximity to tumor, tumor to bowel distance, and methods toprotect bowel such as increasing bowel distance in positioningfor RFA, change in percutaneous puncture site, strategicelectrode positioning, and, more recently, hydrodissection.Changes in distance from diagnostic CT to CT at RFA wererecorded as were results and complications.

RESULTS: Mean distance to bowel was 3.5 cm (Range 0.1 ­12 cm). 18 tumors were within 5 mm and 27 within 1 cm atdiagnostic CT. Of those tumors within 1 cm or less, nearestbowel was colon (15), small bowel (nonduodenum) (8),duodenum (3), and stomach (1). Potentially important changesin tumor proximity to bowel occurred in 21 of the 100 RCCsbetween diagnostic CT and the localizing CT at RFA.Decreases (12) in bowel distance were slightly more commonthan increases (9). Trends with positioning were observedbut did not reach significance (p =0.07): for prone patients,decreases (n =10) were more common than increases (n =3);for decubitus patients, increases (n =5) were more commonthan decreases (n = 3). Changes in bowel location at RF causedchanges in percutaneous puncture site for 3 RCCs. 21 caseswere performed with strategic electrode positioning to avoidmaximal heating at the RCC edge near bowel, and 4 were doneusing hydrodissection to displace bowel. 24 of these 25 RCCswere completely ablated. There were no bowel relatedcomplications. Post RFA CT showed the nearest bowel to benormal in all cases.

CONCLUSION: RFA of RCC within 1 cm of bowel can beperformed safely and effectively using positional maneuvers,changes in puncture site, strategic electrode positioning orhydrodissection. Changes in bowel distance from supine toprone or decubitus CT for RFA are common and requirevigilance.

Abstract No. 201

Semi-Automated Treatment Planning Software for RFAblation.B.J. Wood, National Institutes ofHealth, Bethesda, MD, USA 'A.Viswanathan 'E. McCreedy 'J. Yanof'M. McAuliffe 'A. Sofer, etat.

PURPOSE: Successful RFA oflarger tumors requires preciseoverlapping of multiple ellipses to completely cover the tumor.Multiple optimization tools are presented to automaticallydetermine the number of needle electrodes, their entry pointand positioning, organ and vessel location, and robotic needleplacement.

MATERIALS AND METHODS: The model imposes a meshgrid over the VOL A binary variable is defined for each possibleentry point and each possible tip point of the electrode withinthe grid, that determines whether it is treated or not.Mathematical constraints are imposed to ensure that eachtumor pixel is completely ablated by at least one electrode.Inaccuracies in placement are accounted for. The least numberof entry points that guarantees full ablation is determined.The resulting integer program optimization model hasthousands of variables that are 0 or I. The efficient optimalsolution is computed from amongst the billions of possiblesolutions, using software CPLEX 9.0. Java-based imageprocessing software (MIPAV) was modified to semi­automatically segment, threshold, and display pre-proceduralimages, as well as fuse pre and post-RFA images. Overlappingsphere-packing was planned on CT software (Philips MedicalSystems) with a CT-integrated robot for needle placement inphantoms to test the software.

RESULTS: RFA optimization was simulated for a variety oftumor shapes and sizes ranging up to 8 cm. Solution variationscorrespond to various electrode size and shape, and accountsfor vascularity and convective heat loss. Solution times rangefrom just seconds to a few thousand seconds. Semi-automaticimage processing was performed on RFA patients to showfeasibility (fusion, segmentation, propagation, display).Robotic sphere packing was planned and performed onsimulated phantoms in the CT platform.

CONCLUSION: Current planning models are limited topacking spherical lesions onto a spherical tumor. Ourformulation allow for treatment of a tumor of any shape bymore realistic ellipsoid lesions. The resulting optimizationproblems are solved within times that are viable for pre- andintra-operative planning. Automation for image processingand robotic needle placement may facilitate RFA.

Oncology Embolization

Abstract No. 202

Chemoembolization for Hepatocellular Carcinoma UsingInterventional Radiology (IVR)·CT.M. Honda, Showa University Fujigaoka Hospital, Yokohama,Kanagawa Pref, Japan'T Shimizu'T Kitanosono 'H.Takase 'H. Kubota'M. Obuchi

PURPOSE: To introduce our method of chemoembolizationfor hepatocelluar carcinoma (HCC) using IVR-CT that consistsof DSA and two-channel multi-detector CT scanner.

MATERIALS AND METHODS: From July 2003 to August2004, 126 patients with HCC were treated withchemoembolization using IVR-CT. First, CTAP wasperformed with injection of a total of 60 mL of nonioniccontrast material (150mgIlmJ) through a S-F catheter with atip in the superior mesenteric artery, and CT scanning throughthe liver was performed 25 seconds after the injection with arate of 3 mUsec. Second, CTHA was performed with injectionof a total of 40 mL of contrast material( 150mgIlml) through a5-F or 3-F catheter with a tip in the common hepatic artery orthe proper hepatic artery, and CT scanning thJ:ough the liverwas performed 5 seconds after the injection with a rate of2mL/sec.The location and the number of nodules showed lowdensity area on CTAP and stained on CTHA were recognizedas HCC pattern nodules. Third, the feeder of HCC noduleswere selected by microcatheter guided by 3D-CT images.Forth, mixture of anti-cancer agent and lipiodol was injectedand embolized by gelatin sponge particles. Fifth, immediatelyafter chemoembolization, lipiodol-CT was performed toconfirm accumulation of lipiodol in HCC nodules.Compalisonbetween intravenous dynamic CT using four-channel multi­detector CT within one month prior to chemoembolizationand CTAP, CTHA with IVR-CT was made in 45 patientswith HCC. Hepatic arterial phase and portal venous phasescanning began 40 seconds and 60 seconds, respectively, afterthe injection of 150 mL of nonionic iodinated contrast material(300mgIlml) at a rate of 3-5mL/sec. The number of HCCpattern nodules on dynamic CT and combination of CTAPand CTHA were compared statistically.

TEACHING POINTS: l)Using IVR-CT, HCCs could belocalized and their feeding arteries identified during thetreatment session. 2) The extra-hepatic arteries (ex. inferiorphrenic artery, falciform ligament artery and accessory leftgastric artery) are easily recognized on CT images. 3) Thenumber of HCC pattern nodules obtained by combination ofCTAP and CTHA was significantly higher than that ondynamic CT (p<O.OOl; paired t-test). [MP]