Image Data Processing

Brain CT scans were quantified using 3D-Slicer version 4.10.2 (11). Quantitative measurements were carried out based on the threshold algorithm (20). We coded all the data before the calculations of the analysts to eliminate bias. Brain CT scans were collected in the form of Digital Imaging and Communications in Medicine (DICOM) or, infrequently, as hard copies. Then, the data were transferred to the imaging laboratory. The measurements were performed using the following steps for each CT scan: an original label map was created as the input for the watershed effect. A sphere, namely, the original label map, was painted in the center of the intracranial cavity with one color by the paint effect. The semiautomated pixel thresholding technique was used to label the skull. Meanwhile, the skull was painted with another color. Then, the procedure automatically defaulted the exterior edge of the targeted regions as the internal surface of the skull, which can prevent leakage of the label map. In the center of the intracranial cavity, a sphere was created to execute the initialization of segmentation, which fills in a specific area based on the given outlines created in the label map. With a set of special functions working, the constructure of interesting areas was automatically recognized. The painting should not include a single voxel outside of the region of interest, as the seeding is sensitive and will not respond well to outliers. Next, a gray sphere was extended by iteratively attempting to recognize and mark the partial pixels, which were the same as the pixels of the original label map in the intracranial cavity. Then, the entire endocranial boundary was reconstructed, and the algorithm was stopped. Due to the existence of the skull base foramen, foramen ovale, and foramen magnum, extra images might occur after the calculation. The Scissors tool can be used to remove extra images in the 3D window. After processing, the gray object represents a volumetric representation of the intracranial cavity, which includes brain tissue and non-brain tissue. Because of the distinct signal intensity thresholds of different tissues in the cranial cavity, the semiautomatic pixel threshold method was used to erase all non-brain tissues. The analysts interactively identified the threshold based on the density scale settings of the CT. Ultimately, brain volume was obtained (Figure 3). The CIBV was determined by subtracting the minimum brain volume (BV_max) from the maximum brain volume (BV_min) in the early course of aSAH. As a consequence, the CRBV was determined by CIBV/BV_min × 100% (Figure 4).

Calculation process of brain volume (BV). (A) A gray sphere was drawn in the center of the cranial cavity as the initial value. (B) The inner surface of the skull was defined as the outer boundary of the target segmentation image. (C) The initialization of segmentation started from the initial value and then iteratively filled. (D) The pore structure of the skull base is shown, and a little pixel leakage may be occurred (blue arrow). (E) The volume of the cranial cavity was obtained after the leakage was removed with tools. (F) Pixel threshold technology was used to remove non-brain tissue.

These models present the segmentation result (brain volume) of 3D-Slicer. (A) and (B) showed the brain volume of a patient in brain swelling group at different time points. (A) showed his brain volume within 24 h after bleeding. As shown in (B), the patient developed typically brain swelling with the disease progressed. This patient's CRBV was 11. 35%. By comparison, (C) and (D) showed the brain volume of a patient in non-brain swelling group. The patient's CRBV was 3.26%.