Related Experiment Video
Updated: May 11, 2025
14:08
Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images
Published on: April 13, 2013
42.4K
Automated Classification of Intravenous Contrast Enhancement Phase of CT Scans Using Residual Networks.
Akshaya Anand1, Jianfei Liu1, Thomas C Shen1
1National Institutes of Health, Imaging Biomarkers and Computer-Aided Diagnosis Laboratory Radiology and Imaging Sciences, Clinical Center, Bethesda, MD, USA.
Summary
Accurate classification of computed tomography (CT) scan phases is crucial for AI diagnostics. A ResNet34 model achieved 99% accuracy in identifying five common CT contrast enhancement phases, improving data curation for deep learning.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Radiology
Background:
- CT scan appearance varies significantly across contrast enhancement phases.
- Accurate phase information is vital for computer-aided diagnosis and deep learning model training.
- Current Picture Archiving and Communication Systems (PACS) often lack reliable phase data.
Purpose of the Study:
- To develop an automated method for classifying multiphase CT scans.
- To address the challenge of missing or unreliable contrast enhancement phase information in clinical settings.
Main Methods:
- A residual network (ResNet34) was employed for classification.
- The model was trained on a weakly-labeled dataset of 395 multiphase CT scans.
- Performance was compared against VGG19 and DenseNet121 models.
Main Results:
- ResNet34 achieved a 99% accuracy in classifying five common CT phases.
- VGG19 and DenseNet121 achieved accuracies of 97% and 95%, respectively.
- ResNet34 demonstrated fewer parameters and reduced inference time.
Conclusions:
- Automated multiphase CT classification is feasible with high accuracy.
- This method can significantly improve data curation for deep learning models in medical imaging.
- Enhanced dataset curation can lead to more robust and generalizable AI diagnostic tools.
