Quantitative Automated Segmentation of Lipiodol Deposits on Cone-Beam CT Imaging Acquired during Transarterial Chemoembolization for Liver Tumors: A Deep Learning Approach.

PURPOSE: To show that a deep learning (DL)-based, automated model for Lipiodol (Guerbet Pharmaceuticals, Paris, France) segmentation on cone-beam computed tomography (CT) after conventional transarterial chemoembolization performs closer to the "ground truth segmentation" than a conventional thresholding-based model. MATERIALS AND METHODS: This post hoc analysis included 36 patients with a diagnosis of hepatocellular carcinoma or other solid liver tumors who underwent conventional transarterial chemoembolization with an intraprocedural cone-beam CT. Semiautomatic segmentation of Lipiodol was obtained. Subsequently, a convolutional U-net model was used to output a binary mask that predicted Lipiodol deposition. A threshold value of signal intensity on cone-beam CT was used to obtain a Lipiodol mask for comparison. The dice similarity coefficient (DSC), mean squared error (MSE), center of mass (CM), and fractional volume ratios for both masks were obtained by comparing them to the ground truth (radiologist-segmented Lipiodol deposits) to obtain accuracy metrics for the 2 masks. These results were used to compare the model versus the threshold technique. RESULTS: For all metrics, the U-net outperformed the threshold technique: DSC (0.65 ± 0.17 vs 0.45 ± 0.22, P < .001) and MSE (125.53 ± 107.36 vs 185.98 ± 93.82, P = .005). The difference between the CM predicted and the actual CM was 15.31 mm ± 14.63 versus 31.34 mm ± 30.24 (P < .001), with lesser distance indicating higher accuracy. The fraction of volume present ([predicted Lipiodol volume]/[ground truth Lipiodol volume]) was 1.22 ± 0.84 versus 2.58 ± 3.52 (P = .048) for the current model's prediction and threshold technique, respectively. CONCLUSIONS: This study showed that a DL framework could detect Lipiodol in cone-beam CT imaging and was capable of outperforming the conventionally used thresholding technique over several metrics. Further optimization will allow for more accurate, quantitative predictions of Lipiodol depositions intraprocedurally.

Lipiodol as an intra-procedural imaging biomarker for liver tumor response to transarterial chemoembolization: Post-hoc analysis of a prospective clinical trial.

BACKGROUND: The use of the ethiodized oil- Lipiodol in conventional trans-arterial chemoembolization (cTACE) ensures radiopacity to visualize drug delivery in the process of providing selective drug targeting to hepatic cancers and arterial embolization. Lipiodol functions as a carrier of chemo drugs for targeted therapy, as an embolic agent, augmenting the drug effect by efflux into the portal veins as well as a predictor for the tumor response and survival. PURPOSE: To prospectively evaluate the role of 3D quantitative assessment of intra-procedural Lipiodol deposition in liver tumors on CBCT immediately after cTACE as a predictive biomarker for the outcome of cTACE. MATERIALS & METHODS: This was a post-hoc analysis of data from an IRB-approved prospective clinical trial. Thirty-two patients with hepatocellular carcinoma or liver metastases underwent contrast enhanced CBCT obtained immediately after cTACE, unenhanced MDCT at 24 h after cTACE, and follow-up imaging 30-, 90- and 180-days post-procedure. Lipiodol deposition was quantified on CBCT after cTACE and was characterized by 4 ordinal levels: ≤25%, >25-50%, >50-75%, >75%. Tumor response was assessed on follow-up MRI. Lipiodol deposition on imaging, correlation between Lipiodol deposition and tumor response criteria, and correlation between Lipiodol coverage and median overall survival (MOS) were evaluated. RESULTS: Image analysis demonstrated a high degree of agreement between the Lipiodol deposition on CBCT and the 24 h post-TACE CT, with a Bland-Altman plot of Lipiodol deposition on imaging demonstrated a bias of 2.75, with 95%-limits-of-agreement: -16.6 to 22.1%. An inverse relationship between Lipiodol deposition in responders versus non-responders for two-dimensional EASL reached statistical significance at 30 days (p = 0.02) and 90 days (p = 0.05). Comparing the Lipiodol deposition in Modified Response Evaluation Criteria in Solid Tumors (mRECIST) responders versus non-responders showed a statistically significant higher volumetric deposition in responders for European Association for the Study of the Liver (EASL)-30d, EASL-90d, and quantitative EASL-180d. The correlation between the relative Lipiodol deposition and the change in enhancing tumor volume showed a negative association post-cTACE (30-day: p < 0.001; rho = -0.63). A Kaplan-Meier analysis for patients with high vs. low Lipiodol deposition showed a MOS of 46 vs. 33 months (p = 0.05). CONCLUSION: 3D quantification of Lipiodol deposition on intra-procedural CBCT is a predictive biomarker of outcome in patients with primary or metastatic liver cancer undergoing cTACE. There are spatial and volumetric agreements between 3D quantification of Lipiodol deposition on intra-procedural CBCT and 24 h post-cTACE MDCT. The spatial and volumetric agreement between Lipiodol deposition on intra-procedural CBCT and 24 h post-cTACE MDCT could suggest that acquiring MDCT 24 h after cTACE is redundant. Importantly, the demonstrated relationship between levels of tumor coverage with Lipiodol and degree and timeline of tumor response after cTACE underline the role of Lipiodol as an intra-procedural surrogate for tumor response, with potential implications for the prediction of survival.

Use of Artificial Intelligence in Non-Oncologic Interventional Radiology: Current State and Future Directions.

The future of radiology is disproportionately linked to the applications of artificial intelligence (AI). Recent exponential advancements in AI are already beginning to augment the clinical practice of radiology. Driven by a paucity of review articles in the area, this article aims to discuss applications of AI in non-oncologic IR across procedural planning, execution, and follow-up along with a discussion on the future directions of the field. Applications in vascular imaging, radiomics, touchless software interactions, robotics, natural language processing, post-procedural outcome prediction, device navigation, and image acquisition are included. Familiarity with AI study analysis will help open the current 'black box' of AI research and help bridge the gap between the research laboratory and clinical practice.