TL;DR
This study validates 3D Slicer, a free open-source software, as a highly accurate tool for measuring intracerebral hemorrhage (ICH) volume using CTA source images. It significantly outperforms the traditional Tada formula, especially for irregular and larger hematomas, offering a more reliable and accessible option for clinical and research applications.
What Problem Does This Paper Address?
Accurate measurement of intracerebral hemorrhage (ICH) volume is crucial for managing stroke patients, guiding treatment decisions (like surgery), and predicting outcomes. The widely used Tada formula (ABC/2) is simple but relies on a geometric assumption that simplifies complex hematoma shapes into an ellipsoid. This simplification leads to significant and unpredictable errors, particularly for irregular or lobulated hematomas, potentially misclassifying their size and influencing critical treatment choices. Existing validation studies for alternative methods often use non-contrast CT (NCCT) as the "gold standard," which can be problematic as it's the same imaging modality used for the measurement itself, potentially introducing bias. This paper aims to address this by validating 3D Slicer against a more robust reference standard: manual segmentation of high-resolution CTA source images.
Methods
The researchers conducted a retrospective study using imaging data from 118 patients with spontaneous ICH. They employed three methods to measure hematoma volume:
- Gold Standard: Manual segmentation of hematoma regions on thin-slice (0.625 mm) CTA source images using 3D Slicer. This method leverages the high spatial resolution of CTA to accurately delineate hematoma boundaries.
- 3D Slicer (Semi-automatic): Segmentation of hematomas on standard 5mm slice thickness NCCT images using 3D Slicer's semi-automatic tools, with manual refinement.
- Tada Formula (ABC/2): Calculation of hematoma volume on NCCT images using the traditional ABC/2 formula.
To assess accuracy and agreement, they used intraclass correlation coefficients (ICC) for reliability and Bland-Altman analysis for bias and limits of agreement. Linear regression was also used to examine the relationship between measurement error and hematoma characteristics.
Datasets
The study utilized a dataset of 118 patients with spontaneous intracerebral hemorrhage who underwent both non-contrast CT (NCCT) and CTA at Renmin Hospital of Wuhan University. The CTA images were acquired with a thin-slice reconstruction of 0.625 mm. This dataset was used for retrospective analysis and comparison of the three volumetric measurement methods. No new benchmark datasets were introduced.
Key Insights and Ideas
- CTA Source Images as a Superior Reference: Using thin-slice CTA source images for manual segmentation provides a higher-resolution and more anatomically detailed "gold standard" compared to relying solely on NCCT. This mitigates partial volume effects and offers a more reliable basis for validation.
- 3D Slicer's Robustness: The free and open-source 3D Slicer software demonstrates near-perfect agreement with the CTA-based gold standard, highlighting its accuracy and reliability across various hematoma shapes and sizes.
- Tada Formula's Significant Limitations: The study confirms that the Tada formula, despite its simplicity, systematically overestimates hematoma volume and has wide limits of agreement, particularly for irregular and large hematomas. This suggests a potential for misclassification and inappropriate treatment decisions.
- Clinical Implications of Inaccuracy: The Tada formula's tendency to overestimate volume becomes more pronounced with larger hematomas, precisely when accurate measurement is most critical for guiding interventions. This could lead to unnecessary surgeries or misjudgments of disease progression.
Key Results
- 3D Slicer vs. Gold Standard: 3D Slicer achieved near-perfect agreement with the CTA manual segmentation (ICC = 0.997) with a minimal mean bias of +0.77 mL.
- Tada Formula vs. Gold Standard: The Tada formula showed significantly poorer agreement (ICC = 0.887) with a substantial mean bias of +4.65 mL and wide limits of agreement (-10.8 to +20.2 mL).
- Impact of Morphology: For irregular hematomas, 3D Slicer maintained excellent agreement (ICC = 0.996), while the Tada formula's reliability dropped significantly (ICC = 0.835).
- Impact of Volume: For large hematomas (> 60 mL), 3D Slicer remained highly accurate (ICC = 0.950), whereas the Tada formula's reliability drastically deteriorated (ICC = 0.516).
Limitations and Future Work
Limitations: * Retrospective Design: The study's retrospective nature carries an inherent risk of selection bias, although efforts were made to mitigate this. * "Gold Standard" Definition: While CTA manual segmentation is a superior reference, it's not a pathological "gold standard" and might not capture acute expansion perfectly. * Imaging Parameters: The study's findings are specific to the CTA acquisition parameters (submillimeter slices). Different parameters in other centers might affect results.
Future Work: * Multimodal Image Fusion: Exploring the use of multimodal image fusion (e.g., with MRI) could further improve differentiation of hematomas from other structures and better characterize hemorrhage complexity. * AI-Powered Segmentation: Developing and integrating AI-based fully automatic segmentation algorithms could enhance 3D Slicer's efficiency, bridging the gap between research and routine clinical application. * Clinical Trials and Surgical Planning: The authors strongly recommend adopting 3D Slicer for clinical trials, prognostic modeling, and precision surgical planning due to its accuracy. * Awareness of Tada Formula Limitations: Clinicians should remain aware of the Tada formula's limitations, especially when measurements approach critical decision thresholds, and suspect overestimation.
Summary updated