Radiomic Feature Selection Using Gradient Loss of Deep Neural Network for Lung Cancer Stage Detection

2026-06-03 · Source: Takara TLDR - Daily AI Papers · Field: Science & Research — Artificial Intelligence & Machine Learning, Health & Medical Research · Depth: Expert, medium

Summary

A Gradient-Loss Recursive Feature Elimination (GL-RFE) framework is proposed for identifying influential radiomic features in lung cancer stage detection. This method extracts 106 radiomic features from chest Computed Tomography (CT) scans using the PyRadiomics extension of the 3D Slicer platform. GL-RFE integrates gradient sensitivity analysis from a deep neural network, evaluating feature importance by computing gradients of the network loss with respect to input features and recursively eliminating those with minimal contribution. The top-15 selected features then train a deep neural network classifier to distinguish early-stage from advanced-stage lung cancer. The framework achieved a 90.22% accuracy, 90.10% precision, 90.24% recall, and 90.16% F1-score on the test dataset. This approach effectively captures nonlinear feature interactions, enhances model generalization, and is particularly suitable for high-dimensional, small-sample biomedical datasets.

Key takeaway

For AI Scientists or Research Scientists developing predictive models from high-dimensional, small-sample biomedical datasets, you should consider implementing the Gradient-Loss Recursive Feature Elimination (GL-RFE) framework. This method, which employs deep neural network gradients for feature importance, can significantly enhance model generalization and reduce feature redundancy, as demonstrated by its 90.22% accuracy in lung cancer stage detection. Integrating GL-RFE could improve the reliability and interpretability of your radiomics, genomics, or multimodal clinical analysis.

Key insights

Gradient-Loss Recursive Feature Elimination (GL-RFE) utilizes deep neural network gradients for robust feature selection in high-dimensional medical data.

Principles

• Gradient sensitivity quantifies feature influence on network loss.
• Recursive elimination refines feature sets for improved model generalization.
• Nonlinear feature interactions are effectively captured by DNN gradients.

Method

The GL-RFE method computes gradients of a deep neural network's loss concerning input features, recursively eliminating those with minimal contribution. The resulting top features then train a classifier.

In practice

• Implement GL-RFE for radiomics-based cancer stage detection.
• Apply this framework to genomics datasets.
• Utilize for multimodal clinical analysis.

Topics

• Radiomics
• Feature Selection
• Deep Neural Networks
• Lung Cancer Detection
• Gradient Loss
• Medical Imaging

Best for: AI Scientist, Research Scientist

Related on AIssential

• Diffusion Attention Expert Model for Predicting and Semi-automatic Localizing STAS in Lung Cancer Histopathological Images — DAEM accurately detects STAS in lung cancer histopathology, offering strong generalizability and TME-based localization.
• ResNet-34 with Lightweight Decoder for Accurate and Efficient Segmentation of Fetal Brain MRI — A ResNet-34 encoder with a lightweight MLP decoder achieves accurate and efficient fetal brain MRI segmentation, outperforming baselines.
• GLeVE: Graph-Guided Lesion Grounding with Proposal Verification in 3D CT — GLeVE uses graph reasoning and anatomical verification to precisely ground radiology report lesions in 3D CT scans.
• Hybrid deep learning model for multimodal vocal and lung signal analysis in health monitoring — A hybrid deep learning model effectively integrates vocal and lung signal analysis for accurate disease detection.
• Margin-Consistent Deep Subtyping of Invasive Lung Adenocarcinoma via Perturbation Fidelity in Whole-Slide Image Analysis — Margin consistency with Perturbation Fidelity enhances deep learning robustness and accuracy for lung adenocarcinoma subtyping.
• JASMINE — Early skin cancer detection in rural areas can be significantly improved with low-cost, offline, edge AI devices.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Takara TLDR - Daily AI Papers.