segment_lungs

Segment lungs from CT images using threshold and morphological operations

---

This is a simple script to process the COVID-19 CT images.

segment_lungs.py

1. Load a lung 3D CT scan
2. Plot a selected 2D slice
3. Normalise image intensities:
   1. stretch intensity values distribution using percentiles
   2. standardise around mean value
   3. rescale [0-255]
4. Mask the body:
   1. region growing
   2. morphological operations
   3. connected components
5. Mask the lung:
   1. threshold
   2. morphological operations
   3. connected components

---

Results

• The algorithm works fine for most of the cases
• If the right and left lung are connected, they may get segmented as one label (2-3 cases)
• It can miss infected lesions that appears with different intensity values than the lung's
• It won't work on bad converted nifti images from dicom, because of the weird intensity distribution

---

Examples

Control

Pathology

Pathology