Related Experiment Video
Updated: Jul 16, 2025
14:08
Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images
Published on: April 13, 2013
42.7K
Fully automated sinogram-based deep learning model for detection and classification of intracranial hemorrhage
Chitimireddy Sindhura1, Mohammad Al Fahim1, Phaneendra K Yalavarthy2
1Department of Electrical Engineering, Indian Institute of Technology, Tirupati, India.
Medical Physics
|September 13, 2023
Summary
This study introduces a novel deep learning method for detecting Intracranial Hemorrhages (ICH) directly from sinograms, bypassing CT reconstruction. The sinogram-based approach improves accuracy and robustness for efficient ICH diagnosis.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Neurology
Background:
- Computed Tomography (CT) reconstruction for Intracranial Hemorrhage (ICH) diagnosis is time-consuming and prone to noise and artifacts.
- Conventional methods require a lengthy image reconstruction process, delaying critical diagnoses.
Purpose of the Study:
- To propose an automated deep learning framework for detecting and classifying ICH directly from sinograms.
- To eliminate the need for CT reconstruction, thereby reducing diagnosis time and minimizing image artifacts.
Main Methods:
- A two-stage automated approach using a deep learning framework.
- Stage 1: Intensity Transformed Sinogram Synthesizer to create sinograms equivalent to intensity-transformed CT images.
- Stage 2: A cascaded Convolutional Neural Network-Recurrent Neural Network (CNN-RNN) model for hemorrhage detection and classification from synthesized sinograms.
Main Results:
- Achieved up to 27% improvement in patient-wise accuracies compared to state-of-the-art methods (ResNext-101, Inception-v3, Vision Transformer).
- Demonstrated increased robustness to noise and offset errors compared to CT image-based approaches.
- Successfully performed multi-label classification for hemorrhage type and provided explainability via activation maps.
Conclusions:
- The sinogram-based approach offers accurate and efficient ICH diagnosis without CT reconstruction.
- This method overcomes limitations of CT image-based approaches and shows promise for clinical application.
- Further research is warranted to explore its potential in clinical settings for hemorrhage detection.
