Virtual Biopsy for Diagnosis of Chemotherapy-Associated Liver Injuries and Steatohepatitis: A Combined Radiomic and Clinical Model in Patients with Colorectal Liver Metastases

Abstract

Non-invasive diagnosis of chemotherapy-associated liver injuries (CALI) is still an unmet need. The present study aims to elucidate the contribution of radiomics to the diagnosis of sinusoidal dilatation (SinDil), nodular regenerative hyperplasia (NRH), and non-alcoholic steatohepatitis (NASH). Patients undergoing hepatectomy for colorectal metastases after chemotherapy (January 2018-February 2020) were retrospectively analyzed. Radiomic features were extracted from a standardized volume of non-tumoral liver parenchyma outlined in the portal phase of preoperative post-chemotherapy computed tomography. Seventy-eight patients were analyzed: 25 had grade 2-3 SinDil, 27 NRH, and 14 NASH. Three radiomic fingerprints independently predicted SinDil: GLRLM_f3 (OR = 12.25), NGLDM_f1 (OR = 7.77), and GLZLM_f2 (OR = 0.53). Combining clinical, laboratory, and radiomic data, the predictive model had accuracy = 82%, sensitivity = 64%, and specificity = 91% (AUC = 0.87 vs. AUC = 0.77 of the model without radiomics). Three radiomic parameters predicted NRH: conventional_HUQ2 (OR = 0.76), GLZLM_f2 (OR = 0.05), and GLZLM_f3 (OR = 7.97). The combined clinical/laboratory/radiomic model had accuracy = 85%, sensitivity = 81%, and specificity = 86% (AUC = 0.91 vs. AUC = 0.85 without radiomics). NASH was predicted by conventional_HUQ2 (OR = 0.79) with accuracy = 91%, sensitivity = 86%, and specificity = 92% (AUC = 0.93 vs. AUC = 0.83 without radiomics). In the validation set, accuracy was 72%, 71%, and 91% for SinDil, NRH, and NASH. Radiomic analysis of liver parenchyma may provide a signature that, in combination with clinical and laboratory data, improves the diagnosis of CALI.

Keywords: chemotherapy-associated liver injuries; colorectal liver metastases; diagnostic imaging; liver surgery; nodular regenerative hyperplasia; radiomics; sinusoidal dilatation; steatohepatitis; textural features; virtual liver biopsy.

Figure 1

ROC curve analysis referring to the prediction of sinusoidal dilatation (grade 0–1 vs. grade 2–3), considering clinical, laboratory, and radiomic variables for model training.

Figure 2

Decision tree for the prediction of grade 0–1 vs. grade 2–3 sinusoidal dilatation (based on model in Table 3). Nodes correspond to the decision steps. Each colored square reports: (a) the response mode class in the node, i.e., the predicted outcome of that node (presence of grade 2–3 sinusoidal dilatation = 1; absence of grade 2–3 sinusoidal dilatation = 0, the top number in the square); (b) the percentages of observations in the node belonging to the first response class (absence of grade 2–3 sinusoidal dilatation) and the second response class (presence of grade 2–3 sinusoidal dilatation) (the two central numbers in the square, summing up to 1); (c) the percentage of the total population falling into the node (the bottom number in the square). Decision rules are specified on each node.

Figure 3

ROC curve analysis referring to the prediction of NRH (no vs. yes), considering clinical, laboratory, and radiomic variables for model training.

Figure 4

Decision tree for the prediction of NRH (based on model in Table 4). Nodes correspond to the decision steps. Each colored square reports: (a) the response mode class in the node, i.e., the predicted outcome of that node (presence of NRH =1; absence of NRH =0, the top number in the square); (b) the percentages of observations in the node belonging to the first response class (absence of NRH) and the second response class (presence of NRH) (the two central numbers in the square, summing up to 1); (c) the percentage of the total population falling into the node (the bottom number in the square). Decision rules are specified on each node.

Figure 5

ROC curve analysis referring to the prediction of NASH (yes vs. no), considering clinical, laboratory, and radiomic variables for model training.

Figure 6

Decision tree for the prediction of NASH (based on model in Table 5). Nodes correspond to the decision steps. Each colored square reports: (a) the response mode class in the node, i.e., the predicted outcome of that node (presence of NASH =1; absence of NASH =0, the top number in the square); (b) the percentages of observations in the node belonging to the first response class (absence of NASH) and the second response class (presence of NASH) (the two central numbers in the square, summing up to 1); (c) the percentage of the total population falling into the node (the bottom number in the square). Decision rules are specified on each node.