cv-center-z-slice-selection
Select a fixed number of Z-slices centered around the volume midpoint for memory-efficient 2.5D input from 3D CT/MRI stacks
SKILL.mdskills/cv/center-z-slice-selection/SKILL.md
GitHubCenter Z-Slice Selection
Overview
3D medical/scientific volumes often have 30-65+ slices, but loading all of them as input channels is memory-prohibitive. Selecting a fixed number of slices centered around the volume midpoint captures the most informative region (where the surface of interest typically lies) while reducing input channels from 65 to e.g. 6-12. This enables using standard 2D pretrained backbones with multi-channel input.
Quick Start
import cv2
import numpy as np
def load_center_slices(slice_dir, total_slices=65, n_channels=6):
mid = total_slices // 2
start = mid - n_channels // 2
end = mid + n_channels // 2
images = []
for i in range(start, end):
img = cv2.imread(f"{slice_dir}/{i:02d}.tif", cv2.IMREAD_GRAYSCALE)
images.append(img)
return np.stack(images, axis=-1) # (H, W, n_channels)
volume = load_center_slices("fragment_01/surface_volume", n_channels=8)
# Feed to 2D model with in_channels=8
Workflow
- Determine total number of Z-slices in the volume
- Compute center index:
mid = total // 2 - Select symmetric range:
[mid - n//2, mid + n//2) - Load only those slices as grayscale images
- Stack along channel dimension for 2D model input
Key Decisions
- n_channels: 6-12 is typical; match to model's
in_channelsparameter - Center assumption: works when the signal is near the volume center; adjust offset if not
- Normalization: 16-bit TIFFs → divide by 65535.0 for float32 [0, 1] range
- vs all slices: reduces memory 5-10x with minimal information loss for centered signals