DOLCHID: Paired CBCT and Histopathology Scans for 262 Dental Lesion Cases

262 patient cases of odontogenic lesions, each pairing a pre-operative cone-beam computed tomography scan with expert-verified segmentation masks and a corresponding post-operative H&E-stained histopathology slide with annotated regions of interest. The dataset covers four major lesion subtypes: 44 dentigerous cysts, 54 radicular cysts, 92 odontogenic keratocysts, and 72 ameloblastomas. This resource was created by Jinman Kim and published on figshare in May 2026 to support AI development in dental imaging.

Use Cases

• Benchmarking AI algorithms for lesion classification based on paired radiological and pathological data.
• Developing models for automated segmentation of lesions in CBCT scans based on expert-verified masks.
• Training multimodal networks that fuse features from CBCT and H&E-stained slide images for improved diagnosis.
• Studying the correlation between radiological appearance and histological ground truth across four lesion subtypes.

Strengths

• 39.7 GB of paired, high-dimensional medical imaging data.
• 262 patient cases with expert-verified segmentation masks and annotated histopathology slides.
• Covers four clinically relevant lesion subtypes with specified counts (44, 54, 92, 72).
• Designed as the first publicly available paired resource for this research domain.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for specific modeling tasks.
• Data may reflect clinical bias inherent to the source collection platform.

Provenance

Source

figshare

Collection Method

Clinical collection of pre-operative CBCT scans and post-operative histopathology slides.

Freshness

Last updated 2026-05-20 07:04:40.