Related Experiment Video
Updated: Sep 9, 2025
Tracking the Mammary Architectural Features and Detecting Breast Cancer with Magnetic Resonance Diffusion Tensor Imaging
Published on: December 15, 2014
Multi-parametric MRI diffusion models combined with clinical information for predicting Ki-67 expression in
Ke Li1,2, Jing Li1, Xiaoling Liu1
1Department of Radiology, Southwest Hospital, Army Medical University (Third Military Medical University), Chongqing, China.
Purpose:
To evaluate the diagnostic value of intravoxel incoherent motion (IVIM) and diffusion kurtosis imaging (DKI) parameters combined with clinical information for predicting Ki-67 expression in pancreatic ductal adenocarcinoma (PDAC).
Methods:
This prospective cohort study enrolled 65 patients with histopathologically confirmed PDAC between January 2024 and May 2025. All patients underwent 3.0T MRI including conventional sequences and advanced diffusion-weighted imaging sequences. Clinical data and laboratory parameters were collected within one week before surgery or biopsy. Ki-67 expression was assessed using immunohistochemical staining with 50% as the cutoff value. Two radiologists independently performed quantitative measurements with excellent inter-observer reliability (ICC > 0.85). Univariate and multivariate logistic regression analyses identified independent predictors. ROC curve analysis and DeLong test evaluated diagnostic performance.
Results:
Based on Ki-67 expression threshold of 50%, 48 patients (73.8%) were classified as low expression and 17 patients (26.2%) as high expression. Compared to the low Ki-67 group, the high expression group demonstrated significantly lower monocyte count (0.35 ± 0.09 vs. 0.49 ± 0.16 × 10⁹/L, P = 0.001), higher IVIM f-value (14.08 ± 3.41% vs. 10.90 ± 3.83%, P = 0.004), and lower DKI-MD (1.26 ± 0.17 vs. 1.65 ± 0.17 × 10⁻³ mm²/s, P < 0.001). Individual prediction models achieved AUCs of 0.763 (monocyte count), 0.732 (IVIM f-value), and 0.800 (DKI-MD). The combined prediction model integrating these three parameters demonstrated excellent diagnostic performance with AUC of 0.913 (95% CI: 0.841-0.985), sensitivity of 82.4%, and specificity of 83.3%, significantly outperforming all individual models (P < 0.001).
Conclusion:
This multi-parametric combined prediction model achieves excellent diagnostic performance for preoperative non-invasive assessment of Ki-67 expression status in PDAC, providing a reliable tool for precision medicine practice and personalized treatment strategies.
