Radiomics Based Treatment Selection for Liver Resection and TACE in Hepatocellular Carcinoma: A Cohort Study by Inverse Probability of Treatment Weighting

Please click here to download the model.

Instruction

      The models constructed in this study was aimed to provide noninvasive methods assisting treatment decision between liver resection (LR) and transarterial chemoembolization (TACE) for hepatocellular carcinoma (HCC).

      This study included 520 HCC patients, and model consisted two parts: firstly, a nomogram used to predict PFS after resection and TACE (Fig 1 and Fig 2); secondly, survival plots to stratify patients (Fig 3), ΔModel^CRR = Model^CRR (liver resection score)- Model^CRR (TACE score), for patients with ΔModel^CRR > -5.00, both LR and TACE yielded similar PFS and OS; for patients with a ΔModel^CRR < -5.00, LR yielded better PFS and OS.

Radiomic feature description
      FOS_Kurtosis: Kurtosis of first order statistics (0), which represents the first-order feature on the original image

      CO_IV: inverse variance of co-occurrence (2,1), which represents the co-occurrence textural feature inverse variance on the x-direction high-pass and the y-direction low-pass filtered image

      POF_entropy: peer-off feature of entropy (9), which represents the peel-off feature entropy on the most inside layer.