Improving cervical spinal cord lesion detection in multiple sclerosis using filtered fused proton density-T2 weighted images

Khalid O Alharbi; Abdullah H Abujamea; Othman I Alomair; Hussein M Alsakkaf; Abdulaziz A Alharbi; Sami A Alghamdi; Abdullah G Alharbi

doi:10.1177/20584601221105228

Improving cervical spinal cord lesion detection in multiple sclerosis using filtered fused proton density-T2 weighted images

Acta Radiol Open. 2022 Jun 2;11(6):20584601221105228. doi: 10.1177/20584601221105228. eCollection 2022 Jun.

Authors

Khalid O Alharbi^{1

2}, Abdullah H Abujamea³, Othman I Alomair², Hussein M Alsakkaf³, Abdulaziz A Alharbi¹, Sami A Alghamdi², Abdullah G Alharbi⁴

Affiliations

¹ Department of Radiology and Medical Imaging, Ministry of Health, King Fahad Specialist Hospital Buriydah - Al-Qassim, Buraidah, Kingdom of Saudi Arabia.
² Radiological Science Department, College of Applied Medical Science, King Saud University, Riyadh, Kingdom of Saudi Arabia.
³ Department of radiology and Medical Imaging, King Saud University Medical City & College of Medicine, King Saud University, Riyadh, Saudi Arabia.
⁴ Department of Radiology and Medical Imaging, Ministry of Health, Oyoun Aljawa General Hospital - Al-Qassim, Oyoun Aljawa, Kingdom of Saudi Arabia.

Abstract

Background: Magnetic Resonance Imaging (MRI) is considered a vital in depicting multiple sclerosis (MS) lesions. Current studies demonstrate that proton density (PD) weighted images (WI) are superior to T2 WI in detecting MS lesions (plaques) in the spinal cord.

Purpose: To evaluate the diagnostic value of filtered fused PD/T2 weighted images in detecting cervical spinal cord MS lesions.

Material and methods: In this retrospective study, we selected a sample size of 50 MS patients. Using contrast limited adaptive histogram equalization (CLAHE), a digital image processing filter was used on the (PD/T2) fused images. The produced images were inspected and compared to the original PD images by two experienced neuroradiologists using interobserver and intraobserver. An ROI analysis was also performed on the processed and original PD images.

Results: The repeatability measurement of the match between the two examinations was highly consistent for both neuroradiologists. The repeatability for both neuroradiologists was 96.05%, and the error measurement was 3.95%. The reproducibility measurement of the neuroradiologist's evaluation shows that the processed images could help to identify lesions better [excellent (84.87%)] than PD images [good (61.19%)]. ROIs analysis was performed on 113 MS lesions and normal areas in different images within the sample size. It revealed an enhanced ratio of 2.2 between MS lesions and normal spinal cord tissue in processed fused images compared to 1.34 in PD images.

Conclusion: The processed images of the fused images (PD/T2) have superior diagnostic sensitivity for MS lesions in the cervical spine than PD images alone.

Keywords: contrast limited adaptive histogram equalization; digital image processing; image contrast enhancement; magnetic resonance imaging; multiple sclerosis; spinal cord.