Model Settings for Lung Cancer Histopathology Classification Techniques

97% balanced accuracy was achieved by a proposed method for classifying lung cancer from histopathological images. The dataset contains model settings for techniques including Deep Cluster, Bootstrap Your Own Latent, and SimCLR, combined with Minimum Redundancy Maximum Relevance feature selection and a Vision Graph Convolutional Neural Network classifier. It was authored by Nepolian Vailankanni and last updated in April 2026.

Use Cases

• Benchmarking self-supervised learning techniques for medical image analysis based on the described combination of Deep Cluster, BYOL, and SimCLR.
• Replicating the feature fusion and selection pipeline for lung cancer classification based on the described Minimum Redundancy Maximum Relevance method.
• Comparing classifier performance between plain Vision Graph Convolutional Networks and the proposed optimized approach based on the reported accuracy scores.

Strengths

• The described method achieved a 97% balanced accuracy on the LungHist700 dataset, outperforming a baseline model by 11 percentage points.
• Model settings are provided for a novel combination of three self-supervised learning techniques, which the description suggests is a previously unaddressed approach.
• The work was validated on three publicly available datasets: LungHist700, LC25000, and TCGA UT.

Limitations

• The dataset is very small at 9.5 KB, likely containing only configuration parameters 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 parameters and settings from the described research methodology.

Time Range

The research used images from the LungHist700, LC25000, and TCGA UT datasets; specific temporal coverage of those source images is not provided.

Freshness

Last updated 2026-04-27 17:29:17; freshness should be verified.

Geography

Geographic coverage of the source histopathology images is not specified.