Radiomic Features and Carotid Stenosis in Periodontitis: A Two-Stage Bootstrap and Multimodal Machine Learning Study for Oral-Vascular Research

Mengqiang Zhang, Yun Yang, Qian Cao, Jing Cai, Lei Zheng, Zhipeng Chen, Subinuer Maimaitiaili, Shuai Yuan, Zhen Li, Zhen Zhang, Tao Yang, Tong QiaoResearch Square 2025 · DOI

Abstract

Objectives: This study aims to develop and validate a machine learning model based on jaws Cone Beam Computed Tomography (CBCT) radiomic features for the early detection of carotid atherosclerosis in periodontitis patients. Materials and Methods: Data from 279 observations (206 features each) were analyzed. SMOTE oversampling addressed class imbalance, and a two-stage bootstrap method (1,000 iterations each) was used for robust feature selection via Spearman correlation, LASSO regression, and Logistic Regression with AIC. Results: A final set of 20 features was used to train Logistic Regression, SVM, and Random Forest models. The Random Forest model performed best, achieving an AUC of 0.892, sensitivity of 0.957, and accuracy of 0.859. Conclusion: The developed model demonstrates high predictive performance and clinical utility as a non-invasive screening tool for carotid atherosclerosis in dental settings.

Introduction

There is a well-documented but complex link between periodontitis—chronic gum inflammation—and cardiovascular diseases such as carotid artery stenosis (CAS), a major risk factor for ischemic stroke. While periodontitis is routinely managed in dental clinics using Cone Beam Computed Tomography (CBCT), screening for CAS typically requires specialized vascular imaging. This study addresses the challenge of identifying subtle "radiomic" signatures—quantitative patterns in image intensity and texture—that link jawbone health to systemic vascular risk. By leveraging machine learning to extract these hidden biomarkers from routine dental scans, the research aims to provide a path for early detection of high-risk patients during standard dental care.

Methods and Datasets

The study utilized a retrospective dataset of 279 patients from Nanjing Drum Tower Hospital, including 168 patients with both CAS and periodontitis and 111 with periodontitis alone. The baseline characteristics of the study population are summarized in the table below.

Baseline clinical characteristics of the study population, comparing the CAS with periodontitis group and the periodontitis-only group.

The researchers implemented a multimodal machine learning pipeline focused on statistical stability:

  1. Feature Extraction: 206 morphological, intensity, and texture features were extracted from segmented maxilla and mandible regions using the Pyradiomics library.
  2. Data Balancing: SMOTE (Synthetic Minority Over-sampling Technique) was applied to balance the dataset, resulting in 390 observations.
  3. Two-Stage Bootstrap Selection: A rigorous selection process involved 1,000 bootstrap iterations using Spearman correlation and LASSO regression, followed by another 1,000 iterations using Logistic Regression with the Akaike Information Criterion (AIC) to identify the most stable predictive features.
  4. Model Benchmarking: Three architectures—Logistic Regression (LR), Support Vector Machines (SVM), and Random Forest (RF)—were evaluated using 5-fold cross-validation.

Results

The Random Forest (RF) model demonstrated superior performance for clinical screening purposes:

  • AUC: 0.892
  • Sensitivity: 0.957 (minimizing missed cases of stenosis)
  • Specificity: 0.710
  • Accuracy: 0.859

Decision Curve Analysis (DCA) further confirmed that the RF model provided a higher net benefit across a range of risk thresholds compared to traditional logistic regression, indicating its robustness for clinical decision-making.

Discussion

Analysis of the selected radiomic features revealed that predictive markers appeared symmetrically in both the upper and lower jaws. This suggests that the bone changes being detected are likely reflections of systemic pathophysiological influences, such as chronic inflammation, rather than localized dental issues. Furthermore, first-order intensity statistics like Kurtosis and Skewness were highly frequent in the final models, suggesting that the distribution of bone density is a more potent biomarker for vascular risk than jaw geometry.

While the RF model's ability to handle non-linear interactions proved advantageous, the study's retrospective nature and single-center data source remain limitations. Future research should focus on prospective clinical trials and external validation across different CBCT hardware and patient demographics to ensure broader applicability.

Conclusion

This study successfully developed a Random Forest model using jaw CBCT radiomics to detect carotid atherosclerosis in periodontitis patients. With a high sensitivity of 95.7%, this machine learning framework offers a promising, non-invasive screening tool that could integrate cardiovascular risk assessment into routine dental practice.

Summary updated

Abstract

<title>Abstract</title> Objectives This study aims to develop and validate a deep learning model based on jaws Cone Beam Computed Tomography (CBCT) radiomic features to achieve early detection of potential carotid atherosclerosis in periodontitis patients. Materials and methods The study utilised data from 279 observations, each with 206 features, to distinguish between periodontitis patients with and without concomitant carotid atherosclerosis. To address class imbalance, SMOTE oversampling was applied (dup_size=1), increasing the sample size to 390 observations. A bootstrap method (n_bootstrap=1000) was employed for feature selection. In each iteration, a dataset was created by resampling with replacement. Features were first filtered using Spearman’s rank correlation to remove redundant variables (correlation coefficient &gt;0.8), followed by Lasso regression with ten-fold cross-validation to select predictive variables based on non-zero coefficients. High-frequency features identified through 1000 iterations underwent a second round of bootstrap analysis, where Logistic Regression combined with the Akaike Information Criterion (AIC) was used to determine the final variable set. This rigorous process ensured optimal feature selection for developing an effective early detection model for carotid atherosclerosis in periodontitis patients. Results The study analyzed data from 279 observations, with each observation characterized by 206 features, to differentiate between periodontitis patients with concurrent carotid atherosclerosis and those without. After SMOTE oversampling, the dataset was increased to 390 observations. Feature selection through bootstrap methods identified 26 high-frequency features (&gt;500 times), which were further refined to a final set of 20 features using Logistic Regression combined with AIC. Three machine learning models—Logistic Regression (LR), Support Vector Machine (SVM), and Random Forest (RF)—were developed and evaluated using five-fold cross-validation. The best-performing model was the RF model, achieving an AUC of 0.892, sensitivity of 0.957, specificity of 0.710, and accuracy of 0.859. ROC curves and calibration plots demonstrated good predictive performance and model calibration across all three models. Decision curve analysis showed that the RF model provided the highest net benefit across a range of risk thresholds, indicating its potential for clinical utility in early detection of carotid atherosclerosis in periodontitis patients. Conclusion This study developed a random forest model using jaws CBCT radiomics to detect carotid atherosclerosis in periodontitis patients early. After rigorous feature selection and five-fold cross-validation, it achieved an AUC of 0.892, with sensitivity of 0.957 and specificity of 0.710. The model shows high predictive performance and clinical utility, offering an effective tool for early detection.
BibTeX
@inproceedings{zhang2025radiomic,
  author = {Mengqiang Zhang and Yun Yang and Qian Cao and Jing Cai and Lei Zheng and Zhipeng Chen and Subinuer Maimaitiaili and Shuai Yuan and Zhen Li and Zhen Zhang and Tao Yang and Tong Qiao},
  title = {{Radiomic Features and Carotid Stenosis in Periodontitis: A Two-Stage Bootstrap and Multimodal Machine Learning Study for Oral-Vascular Research}},
  booktitle = {Research Square},
  year = {2025},
  doi = {10.21203/rs.3.rs-8134935/v1},
}
Figure