Scoliosis is a condition of the human body that arises due to the bending of the human spine. The bending may occur either in the coronal or sagittal plane and may be visible as a ‘C’ or ‘S’ shaped bend in the spine. The extent of affliction of Scoliosis is generally determined by calculating a parameter called the Cobb Angle, which is the angle of bending of the spine. According to the World Health Organization’s Annual Report on Spinal Deformities, about 2-3% of the world’s population was affected by Scoliosis in 2013. The primary age of onset for scoliosis is 10-15 years old. Females are eight times more likely to progress to a curve magnitude that requires treatment. Scoliosis has a rate of 2:1, girls to boys. In most cases, the Cobb Angle is manually calculated from the X-ray or any other medical image. This takes approximately 15-20 minutes per calculation. In some cases where software modules exist, manual specification of region of interest is required. Existing software modules generally employ manual cropping and Hough Transform to calculate the Cobb Angle. The development of a fully automatic module that detects and quantifies Scoliosis would save time and enable speedy diagnosis on a large scale.
To enable easy and efficient diagnosis of Scoliosis, this project discusses the development of a software module that can automatically calculate the Cobb Angle from a medical image. MATLAB is the tool used to develop and implement this module. The module developed calculates the Cobb Angle from MRI images. It employs noise removal filters, morphological edge detection and the concept of tangents to a curve to automatically calculate the Cobb Angle from any given MRI. Alternate parameters that can be measured to detect the degree of Scoliosis, apart from Cobb angle, are also proposed. Their accuracy and efficiency is compared with the accuracy and efficiency obtained using the manual Cobb Angle technique.
The module was used to test MRI images of 9 affected patients and it yielded an accuracy of about 77%. The measurement of the alternate parameters, i.e., distance between two segments and displacement between endplates could be used with 100% accuracy to manually screen the spinal images and classify them as Normal, Refer Specialist and Surgery. The prototype developed can be scaled for use in hospitals and diagnostic centers.