Radiomic Feature Extraction Based on Computed Tomography for Stroke Patients
Introduction: Stroke continues to be a major cause of death and disability worldwide, with a substantial effect on healthcare systems and people’s quality of life. Effective therapy depends on a timely and precise diagnosis. To improve diagnosis, classification, and prediction for the best possible...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wolters Kluwer Medknow Publications
2025-04-01
|
| Series: | Journal of Medical Physics |
| Subjects: | |
| Online Access: | https://journals.lww.com/10.4103/jmp.jmp_199_24 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Introduction:
Stroke continues to be a major cause of death and disability worldwide, with a substantial effect on healthcare systems and people’s quality of life. Effective therapy depends on a timely and precise diagnosis. To improve diagnosis, classification, and prediction for the best possible patient outcomes, this study investigates how radiomics might improve computed tomography (CT)-based stroke evaluation.
Methodology:
The research included 65 stroke patients from our two institutions from 2023 to 2024. We performed noncontrast CT imaging via a Siemens Somatom Perspective 128-slice multidetector CT scanner and a Supria True64 CT scanner (Fujifilm, USA). Radiomic information, including shape, first-order statistics, and texture properties, was extracted via three-dimensional Slicer software. Statistical examination via Student’s t-test revealed significant differences (P < 0.05). This study successfully applied various image processing techniques, including erosion, dilation, and open-closed images, to analyze CT images of stroke patients in axial, coronal, and sagittal views.
Results:
Feature extraction from these images revealed significant features, with receiver operating characteristic (ROC) analysis revealing promising diagnostic performance, with an accuracy of 87.2% for K-fold (k =10) cross-validation. The sensitivity and specificity values and area under the curve metrics of 0.868, 0.259, and 0.879 for @10Percentile features and 0.842, 0.185, and 0.911 for @90Percentile features, respectively.
Conclusion:
In summary, stroke imaging analysis benefits greatly from the deployment of complex image processing techniques, reliable feature extraction methods, and cutting-edge segmentation algorithms. Accurate stroke lesion identification and stratification are made possible by the thorough assessment of diagnostic parameters such as ROC curves and the application of visualization tools such as heatmaps. |
|---|---|
| ISSN: | 0971-6203 1998-3913 |