Radiomics-based machine learning and deep learning to predict serosal involvement in gallbladder cancer.
Shengnan Zhou1, Shaoqi Han2, Weijie Chen3
1Department of Gastrointestinal Surgery, China-Japan Friendship Hospital, Beijing, China.
Abdominal Radiology (New York)
|October 3, 2023
Summary
Radiomics models using contrast-enhanced computed tomography (CECT) effectively predict serosal involvement in gallbladder cancer (GBC). Logistic regression models achieved high accuracy, demonstrating the potential of radiomics in GBC staging.
Area of Science:
- Radiology
- Oncology
- Medical Imaging Analysis
Background:
- Gallbladder cancer (GBC) staging is crucial for treatment planning.
- Serosal involvement is a key prognostic factor in GBC.
- Accurate prediction of serosal involvement can improve patient outcomes.
Purpose of the Study:
- To evaluate the predictive ability of radiomics models based on contrast-enhanced computed tomography (CECT) for serosal involvement in GBC.
- To compare the performance of various machine learning algorithms and a deep learning model for this prediction task.
Main Methods:
- Retrospective analysis of 152 GBC patients.
- Extraction of 412 radiomic features from CECT images.
- Construction of radiomics models using seven machine learning algorithms and a fully connected neural network (f-CNN) deep learning model.
- Evaluation of model performance using receiver operating characteristic (ROC) curve analysis.
Main Results:
- 75 radiomic features showed significant differences between groups (P < 0.05).
- Logistic regression achieved the highest performance with an Area Under the Curve (AUC) of 0.944 (sensitivity 0.889, specificity 0.8).
- The f-CNN deep learning model demonstrated an AUC of 0.916 (sensitivity 0.733, specificity 0.801).
Conclusions:
- Radiomics models derived from CECT images show significant potential in predicting serosal involvement in GBC.
- These models, particularly logistic regression, offer convincing performance for non-invasive staging of GBC serosal status.
