Scientific Data Article in Press
Dental Odontogenic Lesion CBCT and Histopathology Integrated Dataset for Benchmarking Deep Learning Algorithms
This paper introduces DOLCHID, a first-of-its-kind dataset of 262 paired CBCT scans and H&E-stained histopathology images for four major odontogenic lesion subtypes. Designed to benchmark deep learning algorithms, DOLCHID enables integrative diagnostic modeling by leveraging complementary radiological and histopathological information, crucial for advancing AI in dental diagnostics.
Transforming Odontogenic Lesion Diagnostics with Integrated AI
The DOLCHID dataset addresses a critical gap in medical imaging AI by providing a robust, multimodal foundation for developing sophisticated diagnostic tools for odontogenic lesions.
0
Paired CBCT & H&E Datasets
0
Major Lesion Subtypes
0%
H&E Classification AUC
Key Benefits for Your Enterprise:
Improved Diagnostic Accuracy: Leverage multimodal AI to enhance the precision and consistency of odontogenic lesion diagnoses, reducing inter-observer variability.
Enhanced Surgical Planning: Utilize AI-driven precise lesion segmentation from CBCT scans to optimize pre-operative planning and minimize invasiveness.
Accelerated AI Development: Gain a competitive edge by using DOLCHID as a robust benchmark for developing and validating next-generation dental imaging AI solutions.
Strategic Recommendations:
Integrate Multimodal AI: Prioritize the development and integration of AI systems capable of fusing CBCT and histopathological data for a holistic diagnostic approach.
Invest in Data-Centric AI: Utilize high-quality, comprehensively annotated datasets like DOLCHID to ensure the reliability and generalizability of your AI models.
Drive Personalized Dentistry: Employ advanced AI diagnostics to facilitate more personalized and effective treatment planning for patients with odontogenic lesions.
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
Dataset Overview
Multimodal Advantage
Technical Validation
The Dental Odontogenic Lesion CBCT and Histology Integrated Dataset (DOLCHID) is a unique, publicly available resource comprising 262 paired CBCT scans and H&E-stained histopathology images. It covers four major odontogenic lesion subtypes: Dentigerous Cyst (n=44), Radicular Cyst (n=54), Odontogenic Keratocyst (n=92), and Ameloblastoma (n=72). Each case is de-identified and includes expert-verified CBCT segmentation masks and annotated histopathological regions of interest, specifically curated to support advanced deep learning research in dental imaging.
Odontogenic lesion diagnosis traditionally relies on a combination of pre-operative CBCT imaging and post-operative histopathological examination. CBCT offers a macroscopic view of lesion size, boundaries, and bone involvement, while H&E histopathology reveals microscopic architecture, including epithelial patterns and cellular atypia. The DOLCHID dataset uniquely integrates these complementary modalities, providing a multi-scale representation crucial for understanding disease behavior and biological basis. This multimodal approach has demonstrated clear diagnostic benefits in other oncology fields and is now made possible for odontogenic lesions, enabling AI models to achieve more comprehensive and accurate diagnoses.
To validate DOLCHID's utility, we performed extensive technical evaluations across three tasks: lesion segmentation, single-modal classification, and multimodal classification. Segmentation experiments on both CBCT and H&E images demonstrated that the dataset contains sufficient structural and textural information for diverse deep learning models. Single-modal classification experiments showed robust subtype prediction, with H&E images achieving up to 98.9% AUC. Crucially, multimodal classification integrating CBCT and H&E data yielded superior performance compared to single-modality CBCT, confirming DOLCHID's suitability for advancing integrative diagnostic modeling and consistent benchmarking.
262
Paired CBCT & Histopathology Datasets
The DOLCHID dataset comprises a unique collection of 262 paired Cone-Beam Computed Tomography (CBCT) scans and Hematoxylin & Eosin (H&E) stained histopathology images, offering an unprecedented resource for multimodal deep learning in odontogenic lesion diagnosis.
Enterprise Process Flow: DOLCHID Data Curation
Data Collection
→
Data Filtering
→
Data Annotation
→
Data Processing
| Dataset | Type | Modalities | Labelled | Objective | Our Dataset Advantage |
|---|---|---|---|---|---|
| CTooth+ | Single modality | 168 CBCT | 22 CBCT with segmentation label | Benchmark segmentation methods and establish performance standards for tooth volume segmentation. | Limited modalities, partial labels. |
| ORCHID | Single modality | 14705 H&E | All images with class-wise label | Research in AI-based histology image analytics by encapsulating various oral cancer and precancer categories. | Limited modalities, no paired data. |
| Multimodal Dental Dataset | Partially paired multi-modality | 329 CBCT, 8 PaX, 16203 PeX | None | Dataset with different oral problems and matching data of three modalities to facilitate machine learning algorithm development. | Partially paired, no segmentation, non-CBCT/H&E. |
| STS-Tooth | Non paired multi-modality | 371 CBCT, 4000 PaX | 32 CBCT and 900 PaX with segmentation label | A large tooth segmentation dataset for semi-supervised learning with data from both children and adults. | Non-paired, limited CBCT/no H&E. |
| DOLCHID (Ours) | Fully paired multi-modality | 262 CBCT, 262 H&E | All images with class-wise and segmentation label | Support odontogenic lesion diagnosis and treatment through the paired multimodal dataset with both CBCT and H&E images. | Fully paired CBCT & H&E, comprehensive annotations across both modalities, ideal for multimodal AI. |
98.9%
Peak H&E Classification AUC
The high Area Under the Curve (AUC) achieved in histopathological classification tasks demonstrates the exceptional quality and diagnostic richness of the H&E images within DOLCHID, validating its suitability for advanced computational pathology research.
Multimodal Integration Paves the Way
The DOLCHID dataset enables the development of advanced multimodal classification algorithms, with methods like Grid-based Feature Fusion (GFF) achieving superior performance (60.7% accuracy) compared to single-modality CBCT classification (up to 58.0% accuracy). This underscores the dataset's unique value in integrating complementary radiological and histopathological information for more comprehensive and accurate diagnostic insights into odontogenic lesions.
Calculate Your Potential ROI with Enterprise AI
Estimate the efficiency gains and cost savings your organization could realize by implementing AI-driven solutions.
Estimated Annual Savings
$0
Annual Hours Reclaimed
0
Your AI Implementation Roadmap
A clear, phased approach to integrating advanced AI into your enterprise, ensuring maximum impact and smooth adoption.
Phase 1: Discovery & Strategy
Comprehensive assessment of current dental diagnostic workflows, identification of key pain points, and strategic planning for AI integration based on DOLCHID capabilities. Define clear objectives and success metrics.
Phase 2: Pilot & Proof-of-Concept
Develop and test a pilot AI model using DOLCHID for a specific odontogenic lesion subtype. Validate performance against clinical benchmarks and gather stakeholder feedback for refinement.
Phase 3: Full-Scale Development
Expand the AI solution to cover all four major odontogenic lesion subtypes, integrating multimodal data fusion techniques. Develop robust segmentation and classification modules, ensuring scalability and reliability.
Phase 4: Integration & Deployment
Seamlessly integrate the AI diagnostic tool into existing clinical IT infrastructure. Conduct rigorous user acceptance testing, provide comprehensive training, and prepare for widespread clinical deployment.
Phase 5: Monitoring & Optimization
Continuously monitor AI model performance in real-world clinical settings. Gather ongoing data for iterative improvements, ensuring the system remains cutting-edge and delivers sustained diagnostic value.
Ready to Transform Your Dental Diagnostics?
Schedule a personalized consultation with our AI experts to explore how DOLCHID-powered solutions can revolutionize your practice.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat