Abstract No. 196 Prospective Multi-Center Pivotal Trial of a Novel Software Application for Real-Time Assessment of Tissue Damage as Depicted Sonographically During Liver Tissue Ablation Procedures

An increase of liver microwave ablation (MWA) volume within 24 hours post-procedure due to delayed tissue injury was previously reported. The current study aims to demonstrate the effectiveness of an artificial intelligence-based image analysis software (SW) for estimation of the ablation zone from intra-procedural ultrasound as compared to 24-hour post-procedure contrast-enhanced computed tomography (CECT). A total of 50 patients with 51 liver tumors undergoing real-time ultrasound monitored MWA with single antenna were included. Median age was 67.5 years (range 40-81) with 74% male patients (n=37). 65% were naïve HCC tumors (n=33) and the remaining (n=18) were various metastases. Median tumor diameter was 1.7 cm (range 1.0-3.0). Ultrasound cine clips were recorded during the procedure and analyzed by the SW to create an ablation zone map. Two CECT scans with similar acquisition parameters were acquired immediately (T=0) and 24hours post-procedure (T=24). A team of three interventional radiologists approved the ablation zone segmentation on each CECT and the location of the CECT oblique plane corresponding to the intra-procedure ultrasound recording. The T=0 ablation zone and the SW ablation zone map were compared to the T=24 ablation zone on the 2D oblique plane using Dice Coefficient, Sensitivity and Precision (PPV). The average calculated Dice between T=0 vs. T=24 ablation zones were 76.8% (range 10-88%). The average Dice between the SW ablation zone map and the T=24 ablation zone was 85.5% (range 56-94%). The sensitivity of the SW map was found to be higher than that of the T=0 CECT, with 81.6% versus 63.7%, compared to the T=24 ablation zone. The average PPV value of the SW Map was 91.2% when compared to the T=24 ablation zone. Statistical analysis showed that the mean Dice for the SW vs. T=24 CECT was significantly higher than T=0 CECT vs. T=24 CECT (p-value < < 0.05). We demonstrate that the proposed SW can accurately estimate the ablation zone from intra-procedural ultrasound as would be visualized on T=24 CECT. This may provide the physician with real-time information regarding the final ablation zone extent.