Related Experiment Video
Updated: Sep 12, 2025
14:08
Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images
Published on: April 13, 2013
42.8K
HU to RGB transformation with automatic windows selection for intracranial hemorrhage classification using ncCT
Dittapong Songsaeng1, Akara Supratak2, Pantid Chantangphol3
1Department of Radiology, Faculty of Medicine Siriraj Hospital, Mahidol University, Bangkok, Thailand.
Plos One
|August 6, 2025
Summary
This study introduces a novel HU to RGB Transformation (HRT) preprocessing technique to improve Intracranial Hemorrhage (ICH) classification on non-contrast CT scans. HRT enhances hemorrhage visualization, leading to more accurate detection and classification of ICH types.
Area of Science:
- Medical Imaging
- Radiology
- Artificial Intelligence in Medicine
Background:
- Classifying Intracranial Hemorrhage (ICH) on non-contrast computed tomography (ncCT) is crucial for patient care.
- Standard window-width (WW) and window-level (WL) settings can obscure hemorrhage details due to variations in brain components and patient conditions.
- Accurate visualization is essential for effective diagnosis and classification of ICH types.
Purpose of the Study:
- To introduce and evaluate a novel preprocessing technique, HU to RGB Transformation (HRT), for enhancing ICH visualization on ncCT scans.
- To improve the accuracy of classifying five categories of ICH using deep neural network models.
- To assess the impact of HRT on the diagnostic accuracy of medical residents.
Main Methods:
- Developed the HU to RGB Transformation (HRT) technique to dynamically select optimal WW and WL parameters.
- Utilized HRT to accentuate hemorrhage visibility by mapping Hounsfield Units to RGB color components.
- Integrated HRT as a preprocessing step for a deep neural network-based image classification model.
- Leveraged multiple brain components to refine the delineation of hemorrhagic regions.
Main Results:
- The HRT preprocessing method achieved an average sensitivity of 89.35% and specificity of 96.03% in classifying five ICH types and normal slices.
- Direct assessment of HRT preprocessed images by residents improved ICH type classification accuracy, reaching 97.39% sensitivity and 96.19% specificity.
- HRT-enhanced classification accuracy surpassed that obtained from reading standard DICOM files (93.31% sensitivity, 94.81% specificity).
Conclusions:
- The HU to RGB Transformation (HRT) is an effective image preprocessing technique for improving Intracranial Hemorrhage detection and classification on ncCT scans.
- HRT significantly enhances the visibility of hemorrhagic regions, leading to improved diagnostic performance for both automated models and human experts.
- This method offers a valuable tool for radiologists and clinicians, potentially leading to faster and more accurate diagnoses of ICH.
