AUTOMATED SEGMENTATION OF CORTICAL NECROSIS USING A WAVELET BASED ABNORMALITY DETECTION SYSTEM.

Area of Science:

• Medical Imaging
• Neuroscience
• Radiology

Background:

• Cortical necrosis, dead brain tissue in the cortex, is a consequence of cerebrovascular disease (CVD).
• Accurate segmentation of cortical necrosis is challenging due to similar intensity patterns with cerebrospinal fluid (CSF).

Purpose of the Study:

• To develop and validate an automated method for segmenting cortical necrosis from brain FLAIR-MR images.
• To improve the detection and segmentation of brain lesions in patients with cerebrovascular disease.

Main Methods:

• A model of normal cortical variation was generated using Jacobian-based registration of healthy control MR scans.
• Deviations from the normal variation model in patient scans were identified as abnormalities.
• Abnormalities in the cortex with CSF-like intensity profiles were segmented as cortical necrosis.

Main Results:

• The automated method successfully detected and segmented cortical necrosis in 37 CVD patients.
• The method's performance was evaluated using a model derived from 72 healthy subjects.
• Segmentation overlap with human experts was comparable to inter-expert agreement.

Conclusions:

• The proposed automated method provides accurate segmentation of cortical necrosis in brain MRI.
• This technique shows potential for aiding in the diagnosis and monitoring of cerebrovascular disease.