Radiomics Features for Discriminating Sellar Region Brain Tumors, 93 Patient Cases

A dataset of 93 patients with sellar region brain tumors, including 40 Langerhans cell histiocytosis (LCH) and 53 germ cell tumor (GCT) cases, collected between April 2012 and April 2024. Radiomics features were extracted from multiparametric MRI scans (T1WI and T2WI) with manually segmented regions of interest. The data was used to develop and validate machine learning models for tumor classification.

Use Cases

• Training a classifier to distinguish between Langerhans cell histiocytosis and germ cell tumors based on radiomics features.
• Comparing the performance of different machine learning models (SVM, RF, LR) for medical image analysis.
• Building a combined diagnostic model integrating radiomics, clinical, and imaging semantic features.
• Evaluating feature selection methods like LASSO regression for high-dimensional medical data.

Strengths

• Includes 93 patient cases with pathological or therapeutic diagnosis confirmation.
• Features were extracted from multiparametric MRI, including T1-weighted and T2-weighted imaging.
• Model validation used a five-fold cross-validation approach and reported performance metrics like AUC (0.81 for the best model).

Limitations

• Row count and column-level documentation are unknown, requiring manual inspection after download.
• Data is from a single institution, which may limit generalizability.
• The dataset is small in scale (722.7 KB), which may limit model complexity and statistical power.

Provenance

Source

figshare, author Hongting Jiang.

Collection Method

Retrospective study with features extracted from manually segmented MRI scans.

Time Range

April 2012 to April 2024.

Freshness

Last updated 2026-04-24 05:40:17; freshness should be verified.