Performance Comparison of CNN Models for Gastric Cancer Metastasis Detection

An Inception-ResNetV2 model augmented with an SE attention mechanism achieved an ROC-AUC of 0.973 for detecting occult peritoneal metastasis in gastric cancer. The dataset, shared by Chao Zhang on figshare in 2026, likely contains tabular performance metrics for several convolutional neural network architectures evaluated on preoperative CT images. These findings suggest a potential approach for improving clinical predictive modeling by integrating advanced deep learning architectures with attention mechanisms.

Use Cases

• Benchmarking convolutional neural network architectures based on performance metrics like AUC and F1-Score.
• Evaluating the impact of attention mechanism modules on model performance for medical image analysis.
• Developing clinical decision support tools for identifying high-risk gastric cancer patients requiring staging laparoscopy.

Strengths

• The SE-augmented Inception-ResNetV2 model achieved a high ROC-AUC of 0.973 and a Precision-Recall AUC of 0.908.
• Model performance was validated with calibration curves and decision curve analysis, suggesting clinical utility.
• Dataset is openly shared under a CC-BY-4.0 license, facilitating reuse and verification.

Limitations

• Dataset scale is limited at 5.5 KB, suggesting it contains summary results rather than raw image data.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for certain analyses.

Provenance

Source

figshare

Collection Method

Likely contains aggregated performance metrics from a study evaluating deep learning models on preoperative CT images.

Freshness

Last updated 2026-05-14 17:44:46; freshness should be verified.

Geography

Study context mentions gastric cancer prevalence in East Asia.