Related Experiment Video
Updated: Jun 17, 2025
06:46
Competing-Risk Nomogram for Predicting Cancer-Specific Survival in Multiple Primary Colorectal Cancer Patients after Surgery
Published on: September 27, 2024
239
Comparison between Three Radiomics Models and Clinical Nomograms for Prediction of Lymph Node Involvement in PCa
Domiziana Santucci1, Raffaele Ragone1, Elva Vergantino1
1Department of Diagnostic Imaging, Campus Bio-Medico University of Rome, 00128 Rome, Italy.
Cancers
|August 10, 2024
Summary
The random forest (RF) radiomics model demonstrated superior accuracy in predicting lymph node involvement in prostate cancer (PCa) compared to traditional clinical nomograms. This advancement offers improved diagnostic capabilities for PCa management.
Area of Science:
- Urology
- Medical Imaging
- Computational Pathology
Background:
- Accurate prediction of lymph node involvement (LNI) is crucial for effective prostate cancer (PCa) management and treatment planning.
- Current clinical nomograms have limitations in precisely predicting LNI, necessitating the exploration of advanced predictive models.
Purpose of the Study:
- To compare the performance of radiomics models (logistic regression, random forest, support vector machine) against established clinical nomograms (Briganti, MSKCC, Yale, Roach) for predicting LNI in PCa patients.
- To evaluate the diagnostic accuracy of different radiomics features and models derived from multi-parametric MRI (mp-MRI).
Main Methods:
- A retrospective study of 95 PCa patients who underwent radical prostatectomy and pelvic lymphadenectomy.
- Extraction of radiomic features from mp-MRI (T2, DWI, ADC) using 3D SLICER and Pyradiomics, alongside clinical and histological data.
- Development and independent testing of logistic regression, random forest, and support vector machine radiomics models, with performance assessed by accuracy and AUC.
Main Results:
- The random forest (RF) model achieved the highest predictive performance, particularly with DWI (accuracy 86%, AUC 0.89) and ADC (accuracy 89%, AUC 0.67) sequences.
- Radiomics models, especially RF, showed higher predictive performance compared to clinical nomograms in predicting LNI.
- Significant radiomic features included T2_nodulofirstordervariance and T2_nodulofirstorderkurtosis.
Conclusions:
- The random forest (RF) radiomics model offers superior diagnostic accuracy for predicting lymph node involvement in prostate cancer compared to traditional clinical nomograms.
- Integrated radiomics and semantic data analysis holds significant promise for enhancing the prediction of LNI in PCa.
- Further validation of radiomics models is recommended for clinical implementation in PCa management.
