Development and validation of combined Ki67 status prediction model for intrahepatic cholangiocarcinoma based on clinicoradiological features and MRI radiomics

Abstract

Purpose Incidence and mortality of intrahepatic cholangiocarcinoma (ICC) have been increasing over the past few decades, and Ki67 is an adverse prognostic predictor and an attractive therapeutic target for ICC patients. Thus, we aim to develop and validate a combined Ki67 prediction model for ICC patients. Materials and methods Preoperative contrast-enhanced MR images were collected from 178 patients with postoperative pathologically confirmed ICC, and randomly divided into training and validation cohorts in a ratio of 7:3 (124:54). A time-independent test cohort of 49 ICC patients was used for validation. Independent clinicoradiological features of Ki67 status were determined by multivariate analysis. Optimal radiomics features were selected by least absolute shrinkage and selection operator logistic regression and linear discriminant analysis was used to construct combined models. The prediction efficacy of combined model was assessed by receiver operating characteristics curve, and verified by its calibration, decision and clinical impact curves. Results HBV (p = 0.022), arterial rim enhancement (p = 0.006) and enhancement pattern (p = 0.012) are independent clinicoradiological features. The radiomics model achieves good prediction efficacy in the training cohort (AUC = 0.860) and validation cohort (AUC = 0.843). The combined Ki67 prediction model incorporates clinicoradiological and radiomics features, and it yields desirable predictive efficiency in test cohort (AUC = 0.815). Decision curves and clinical impact curves further validate that the combined Ki67 prediction model can achieve net benefits in clinical work. Conclusion The combined Ki67 model incorporating HBV, arterial rim enhancement, enhancement pattern and radiomics features is a potential biomarker in Ki67 prediction and stratification.

Non-invasive assessment of response to transcatheter arterial chemoembolization for hepatocellular carcinoma with the deep neural networks-based radiomics nomogram

“…This design yields diverse candidate feature subsets, which are instrumental in constructing a powerful predictive model with exceptional performance. Similar methodologies have been applied in previous publications (17,18). Among these three feature subsets, it is noteworthy that the majority of selected features belong to high-order categories rather than first-order features.…”

Association Between MRI Radiomics and Intratumoral Tertiary Lymphoid Structures in Intrahepatic Cholangiocarcinoma and Its Prognostic Significance

“…14,15 MRI is important in the diagnosis and clinical management of ICC patients, including differential diagnoses, preoperative staging evaluation, therapeutic response evaluation, postoperative pathological factors prediction, and prognosis stratification. 9,[16][17][18][19][20][21] This study aimed to develop an MRI-based radiomics signature to preoperatively predict the intratumoral TLSs status of ICC and evaluate its prognostic value in two independent validation cohorts.…”

“…Radiomics can be used to extract, select, and analyze quantitative information that cannot be identified by visual inspection of images to reflect tumoral pathophysiology, intratumoral heterogeneity, and cancer phenotype, leading to better clinical decision‐making 14,15 . MRI is important in the diagnosis and clinical management of ICC patients, including differential diagnoses, preoperative staging evaluation, therapeutic response evaluation, postoperative pathological factors prediction, and prognosis stratification 9,16–21 …”