Return To Abstract Listing

Gastrointestinal Imaging

E1013. Role of Radiomic Texture Analysis Mapping in Prediction of Microvascular Invasion in HCC

Khalaf A1,2,  Morshid A1,  Fuentes D1,  Hazle J1,  Elsayes K1,  Qayyum A.1 1. The University of Texas, MD Anderson Cancer Center, Houston, TX; 2. The University of Alabama at Birmingham (UAB), Birmingham, AL

Address correspondence to A. Khalaf (amkhalaf@uab.edu)

Objective: Clinical Relevance: HCC tumors that present with vascular invasion are aggressive tumors that invade nearby anatomy, generally do not have a well-defined boundary, and are associated with poor treatment response. Imaging features that can prognosticate treatment response for hepatocellular carcinoma (HCC) patients promise to be a powerful selection methodology that will enable clinicians to identify patients that are likely to respond well to therapy. Purpose: The presented efforts focus on establishing quantitative imaging correlations with aggressive HCC as characterized by vascular invasion.

Materials and Methods: Methods. This IRB approved, retrospective single-institution study included 32 patients diagnosed with HCC, including 16 with pathologically confirmed microvascular invasion (MVI) and 16 without vascular invasion. Inclusion criteria were unresectable HCC, transarterial chemoembolization (TACE) as the first line of therapy, Child-Pugh score A or B, and multiphasic contrast-enhanced CT at baseline. Intensity, shape, and co-occurrence image features were computed within the viable, necrotic, and background liver parenchyma. Independent t-test was conducted to compare quantitative imaging features between tumors with confirmed MVI and tumors without vascular invasion.

Results: Results. Statistical analysis shows that the gradient and neighborhood variance image features are able to differentiate between the presence or absence of microvascular invasion (p < .05). These features intuitively represent local variations in the tissue properties.

Conclusion: Conclusions. This study demonstrates the feasibility of quantifying vascular invasion using quantitative image features. Ongoing efforts to extend and validate these methods in larger patient populations and correlate these features with survival are likely to provide useful information to assist in the current patient staging methods (TNM, BCLC, CLIP).