GDT-SwinKid: Transformer Model Performance Metrics for Kidney CT Lesion Analysis

Performance metrics for the GDT-SwinKid model, a hybrid transformer for renal lesion segmentation and classification, published by Thirupathi Rao N in May 2026. The dataset includes Dice coefficients as high as 0.95 and AUC-ROC values approaching 0.99, comparing the model against conventional U-Net and Swin Transformer baselines. It is stored in an XLS file and shared under a CC-BY-4.0 license.

Use Cases

• Benchmarking segmentation model performance based on reported Dice coefficient and AUC-ROC metrics
• Comparing hybrid transformer architectures against conventional U-Net and Swin Transformer baselines for medical imaging
• Evaluating the integration of statistical modeling, specifically adaptive Gamma distribution, in deep learning models
• Assessing the utility of explainable attention maps and deep supervision for clinical trust in AI diagnostics

Strengths

• Includes specific, high-performance metrics: Dice coefficients up to 0.95 and AUC-ROC values approaching 0.99
• Reports absolute performance improvements of 5–9% in Dice coefficient over specified baselines
• Model was validated on a large set of clinical datasets, as mentioned in the description

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment
• The dataset is very small at 5.5 KB, indicating limited scope

Provenance

Source

figshare

Collection Method

Likely contains performance metrics from model validation experiments.

Freshness

Last updated 2026-05-21 07:33:35; freshness should be verified