Abstract
Hemorrhage expansion (HE) is a common and serious condition in patients with intracerebral hemorrhage (ICH). While volume changes are well-documented, little is known about the morphological changes that occur during HE. This multicenter retrospective study developed a novel quantitative framework to explore 3D morphological evolution in 1,094 patients. Key findings include a "gravity effect" where hematomas shift downward in supine patients and the identification of "lateral expansion" (left-right growth) as a significant predictor of poor clinical outcomes. A prognostic nomogram incorporating these morphological parameters achieved an AUC of 0.824, suggesting that shape dynamics offer vital prognostic information beyond traditional volume measurements.
Introduction
Spontaneous intracerebral hemorrhage (sICH) is a devastating form of stroke where a ruptured blood vessel causes bleeding into the brain tissue. In approximately one-quarter of cases, this bleed continues to grow—a phenomenon called hemorrhage expansion (HE)—which is a major driver of disability and death.
Current clinical practice relies heavily on measuring the total volume of the hematoma. However, volume is a blunt instrument; it doesn't describe how or where the bleed is growing. Does it expand symmetrically? Does it shift as the patient lies in a hospital bed? This study addresses this gap by developing a quantitative framework to track the morphological "fingerprint" of a hematoma over time, seeking to understand if specific growth patterns correlate with patient prognosis.
Methodology
The researchers developed a novel, semi-automated pipeline to quantify hematoma morphology using paired non-contrast CT scans (an initial scan and a follow-up scan within 72 hours).
Key technical components included:
- Segmentation: Hematoma boundaries were defined using a combination of the ITK-SNAP platform and an in-house AI-based segmentation tool.
- Morphological Metrics: They calculated the Surface Regularity (SR) index (a ratio of surface area to volume) and measured diameters across three axes: Anterior-Posterior (AP), Left-Right (LR), and Superior-Inferior (SI).
- Atlas Registration: To compare changes across different patients, all CT scans were spatially registered to a common coordinate system (the MNI152 brain atlas).
- Center of Mass Tracking: They calculated the geometric center of the hematoma in 3D space to track the distance and direction of its movement.
Study Cohort and Selection
The study utilized the Chinese Intracranial Hemorrhage Image Database (CICHID), a massive repository initiated by Peking Union Medical College Hospital. The researchers applied rigorous exclusion criteria to ensure data quality and analytic rigor.
The final cohort consisted of 1,094 eligible patients from four medical centers who met all inclusion criteria, including successful image registration and available paired CT scans within specific time windows.
Key Insights: The Physics of Bleeding
The analysis revealed several distinct patterns in how hematomas evolve:
- The Gravity Effect: In deep and subtentorial brain regions, the center of the hematoma consistently moved in the direction of gravity (downward, as patients are typically supine). This suggests that the physical weight of the blood and the surrounding tissue's compliance play a role in how bleeds evolve.
- Shape Stability: Despite expansion, the "longitudinal axis" (the primary direction of the bleed's length) remained stable in 95.2% of patients. Most hematomas are longest in the Anterior-Posterior direction (64.7%) and stay that way even as they grow.
- Irregularity Decreases with Growth: Interestingly, initial hematomas tended to be more irregular than the expanded ones. As the bleed grows, it appears to "smooth out" slightly, though it remains far from a perfect sphere.
Clinical Significance: Lateral Expansion
The most striking finding was that growth in the Left-Right (LR) direction, termed "lateral expansion," was a potent predictor of poor outcomes. The risk of a poor prognosis increased by a factor of 1.139 for every 1 mm of lateral expansion.
A prognostic nomogram model including lateral expansion showed good performance in predicting poor clinical outcomes at discharge, achieving an AUC of 0.824 (95% CI: 0.800–0.846). This indicates high accuracy in identifying patients at risk for poor functional recovery.
Discussion and Limitations
The study demonstrates that the way a hematoma grows is as important as how much it grows. However, the authors acknowledge several limitations, including the retrospective design and the geographic specificity of the cohort (entirely from China). Furthermore, patients with very rapid expansion requiring immediate surgery were excluded, potentially biasing the sample toward more stable cases.
Future research could integrate these morphological parameters into real-time AI monitoring tools to alert physicians when a "lateral expansion" pattern is detected, potentially triggering earlier intervention to improve patient outcomes.
Summary updated