Value of 18F-FDG PET/CT radiomics combined with clinical features for predicting Ki-67 expression in colorectal cancer

Abstract Objective To evaluate the value of 18F-FDG PET/CT radiomics combined with clinical features for the preoperative prediction of Ki-67 expression in colorectal cancer (CRC). Methods A total of 117 patients with pathologically confirmed CRC who underwent pretreatment 18F-FDG PET/CT between June 2016 and October 2024 were retrospectively included. Patients were classified into high- and low-expression groups using a 50% Ki-67 cutoff and were randomly divided into a training cohort and an internal validation cohort at a 7:3 ratio. Radiomic features were extracted from CT and PET images and reduced by standardization, univariate screening, correlation analysis, and least absolute shrinkage and selection operator (LASSO) regression. Logistic regression (LR), support vector machine (SVM), k-nearest neighbors (KNN), ExtraTrees, LightGBM, and multilayer perceptron (MLP) classifiers were compared. Independent clinical predictors were identified to construct a clinical model and a combined clinical-radiomics nomogram. Model performance was assessed using receiver operating characteristic (ROC) analysis, the DeLong test, and decision curve analysis (DCA). Results Six CT features, four PET features, and ten combined PET/CT features were retained. TNM stage, SUVmax, and SUVmean were independent predictors of high Ki-67 expression. The LightGBM-based PET/CT radiomics model achieved AUCs of 0.943 and 0.911 in the training and internal validation cohorts, respectively, with an accuracy of 0.861 in the validation cohort. The combined nomogram achieved AUCs of 0.951 and 0.913, with validation-cohort accuracy, sensitivity, specificity, and F1 score of 0.917, 0.920, 0.909, and 0.939, respectively. The AUC difference between the combined model and the PET/CT radiomics model was not statistically significant, whereas DCA showed a higher net benefit for the combined model within part of the clinically relevant threshold range. Conclusion Multimodal 18F-FDG PET/CT radiomics characterizes CRC from complementary morphologic and metabolic perspectives. Combining radiomics with TNM stage and conventional metabolic parameters may improve preoperative identification of high Ki-67 expression. External validation is required before clinical implementation.

Authors

Publication Details

Journal
Research Square
Published
2026-11-13
DOI
https://doi.org/10.21203/rs.3.rs-11577691/v1
Primary Topic
Radiomics and Machine Learning in Medical Imaging
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Value of 18F-FDG PET/CT radiomics combined with clinical features for predicting Ki-67 expression in colorectal cancer

yu zhang, Hu Wang, ShuangronD Tian
2 links
Research Square
Radiomics and Machine Learning in Medical Imaging
preprint

Value of 18F-FDG PET/CT radiomics combined with clinical features for predicting Ki-67 expression in colorectal cancer

yu zhang, Hu Wang, ShuangronD Tian
preprint en

Abstract

Abstract Objective To evaluate the value of 18F-FDG PET/CT radiomics combined with clinical features for the preoperative prediction of Ki-67 expression in colorectal cancer (CRC). Methods A total of 117 patients with pathologically confirmed CRC who underwent pretreatment 18F-FDG PET/CT between June 2016 and October 2024 were retrospectively included. Patients were classified into high- and low-expression groups using a 50% Ki-67 cutoff and were randomly divided into a training cohort and an internal validation cohort at a 7:3 ratio. Radiomic features were extracted from CT and PET images and reduced by standardization, univariate screening, correlation analysis, and least absolute shrinkage and selection operator (LASSO) regression. Logistic regression (LR), support vector machine (SVM), k-nearest neighbors (KNN), ExtraTrees, LightGBM, and multilayer perceptron (MLP) classifiers were compared. Independent clinical predictors were identified to construct a clinical model and a combined clinical-radiomics nomogram. Model performance was assessed using receiver operating characteristic (ROC) analysis, the DeLong test, and decision curve analysis (DCA). Results Six CT features, four PET features, and ten combined PET/CT features were retained. TNM stage, SUVmax, and SUVmean were independent predictors of high Ki-67 expression. The LightGBM-based PET/CT radiomics model achieved AUCs of 0.943 and 0.911 in the training and internal validation cohorts, respectively, with an accuracy of 0.861 in the validation cohort. The combined nomogram achieved AUCs of 0.951 and 0.913, with validation-cohort accuracy, sensitivity, specificity, and F1 score of 0.917, 0.920, 0.909, and 0.939, respectively. The AUC difference between the combined model and the PET/CT radiomics model was not statistically significant, whereas DCA showed a higher net benefit for the combined model within part of the clinically relevant threshold range. Conclusion Multimodal 18F-FDG PET/CT radiomics characterizes CRC from complementary morphologic and metabolic perspectives. Combining radiomics with TNM stage and conventional metabolic parameters may improve preoperative identification of high Ki-67 expression. External validation is required before clinical implementation.

Research Square
Good health and well-being
Radiomics and Machine Learning in Medical Imaging
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon