Summary
Background: Spontaneous Spinal Cerebro -spinal Fluid Leaks (SSCFL) is a disease based on tears
on the dura mater. Due to widespread symptoms and low frequency of the disease, diagnosis
is problematic. Diagnostic lumbar puncture is commonly used for diagnosing SSCFL,
though it is invasive and may cause pain, inflammation or new leakages. T2-weighted
MR imaging is also used for diagnosis; however, the literature on T2-weighted MRI
states that findings for diagnosis of SSCFL could be erroneous when differentiating
the diseased and control. One another technique for diagnosis is CT-myelography, but
this has been suggested to be less successful than T2-weighted MRI and it needs an
initial lumbar puncture.
Objectives: This study aimed to develop an objective, computerized numerical analysis method
using noninvasive routine Magnetic Resonance Images that can be used in the evaluation
and diagnosis of SSCFL disease.
Methods: Brain boundaries were automatically detected using methods of mathematical morphology,
and a distance transform was employed. According to normalized distances, average
densities of certain sites were proportioned and a numerical criterion related to
cerebrospinal fluid distribution was calculated.
Results: The developed method was able to differentiate between 14 patients and 14 control
subjects significantly with p = 0.0088 and d = 0.958. Also, the pre and post-treatment
MRI of four patients was obtained and analyzed. The results were differentiated statistically
(p = 0.0320, d = 0.853).
Conclusions: An original, noninvasive and objective diagnostic test based on computerized image
processing has been developed for evaluation of SSCFL. To our knowledge, this is the
first computerized image processing method for evaluation of the disease. Discrimination
between patients and controls shows the validity of the method. Also, post-treatment
changes observed in four patients support this verdict.
Keywords
Spontaneous Spinal Cerebrospinal Fluid Leaks - spontaneous intracranial hypotension
- computer-aided diagnosis - mathematical morphology - biomedical image processing
