OCT Images and Model Data for Actinic Keratosis Epidermal Segmentation

Gavrielle Rebecca Untracht's repository contains supporting data for a manuscript on a U-Net model for epidermal segmentation in optical coherence tomography images of actinic keratosis lesions. The dataset, 423.8 MB in size and last updated in May 2026, includes files for evaluating model performance across multiple hyperparameter configurations. The optimal model configuration achieved a Dice score of 0.86 and a Jaccard index of 0.76.

Use Cases

• Training and benchmarking deep learning segmentation models based on OCT images of actinic keratosis.
• Analyzing the impact of hyperparameters like image resolution and batch size on model performance based on the described evaluation metrics.
• Developing efficient image analysis pipelines for dermatological diagnostics based on the described U-Net architecture.

Strengths

• Dataset size is 423.8 MB, indicating a medium-scale collection of images and model data.
• Model performance is quantitatively evaluated with reported metrics, including a Dice score of 0.86 and Jaccard index of 0.76 for the optimal configuration.
• Multiple hyperparameter configurations were tested, including image sizes of 256×256, 512×512, 1024×1024, and 464×1356 pixels.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.

Provenance

Source

figshare

Collection Method

Likely contains expert-annotated optical coherence tomography images and associated model training/evaluation data.

Time Range

null

Freshness

Last updated 2026-05-01 07:44:36; freshness should be verified.

Geography

null