Overview

Magnetic resonance imaging (MRI) has become the imaging study of choice for the diagnosis and monitoring of multiple sclerosis. FLAIR sequences display excellent tissue contrast between normal and abnormal regions, thus maximizing the conspicuity of multiple sclerosis lesions on the images. In addition to FLAIR sequences, T1 and T1 post contrast sequences are also helpful for the identification of chronic and acute active multiple sclerosis lesions, respectively. Conventional diagnostic workstations connected to the clinical PACS may already offer tools for measures of MS lesions; however the measurements are mainly manual or semiautomatic. Performing manual measurements of lesion load in terms of lesion volume and number is a tedious and time consuming procedure especially if there is comparison with multiple follow-up studies on a given patient. An automated system to quantify multiple sclerosis lesions and track progression of lesions within a given patient will improve workflow and provide valuable information that would help determine disease severity, disease progression, and disease response to therapeutic treatment

 

Methods

In this project we develop a CAD method for automated analysis of MRI FLAIR axial sequences. The goal is to track changes in the size and quantity of MS lesions and make comparisons between current and follow up studies. This can be achieved by a fuzzy segmentation of brain area in MR images. The segmentation step is followed by preprocessing procedures including background reduction, bias field correction and skull structure removal and location of the brain mask. MS lesions are identified and lesion load is quantified within the brain mask.
Clinical protocol includes the following sequences: T1, T2, FLAIR axial cross-sections acquired by 1.5T Scanner. The slice thickness is 5mm with 1.5 mm spaces between slices.

 

(a) -- (b)



(c)

Fig. 1 Example of automatic segmentation of MS lesions in 35 year old male patient a) original image, b) image with superimposed segmentation results, c) expert outlines.

 

(a) -- (b)

Fig 2. Example of automatic segmentation of MS lesions in 44 year old female patient in (a) original (recent) study, (b) comparison (old) study