Quantification of Joint Space Width Difference on Radiography Via Phase-Only Correlation (POC) Analysis: a Phantom Study Comparing with Various Tomographical Modalities Using Conventional Margin-Contouring

Abstract

Several visual scoring methods are currently used to assess progression of rheumatoid arthritis (RA) on radiography. However, they are limited by its subjectivity and insufficient sensitivity. We have developed an original measurement system which uses a technique called phase-only correlation (POC). The purpose of this study is to validate the system by using a phantom simulating the joint of RA patients. A micrometer measurement apparatus that can adjust arbitrary joint space width (JSW) in a phantom joint was developed to define true JSW. The phantom was scanned with radiography, 320 multi detector CT (MDCT), high-resolution peripheral quantitative CT (HR-pQCT), cone beam CT (CBCT), and tomosynthesis. The width was adjusted to the average size of a women's metacarpophalangeal joint, from 1.2 to 2.2 mm with increments of 0.1 mm and 0.01 mm. Radiographical images were analyzed by the POC-based system and manual method, and images from various tomographical modalities were measured via the automatic margin detection method. Correlation coefficients between true JSW difference and measured JSW difference were all strong at 0.1 mm intervals with radiography (POC-based system and manual method), CBCT, 320MDCT, HR-pQCT, and tomosynthesis. At 0.01 mm intervals, radiography (POC-based system), 320MDCT, and HR-pQCT had strong correlations, while radiography (manual method) and CBCT had low correlations, and tomosynthesis had no statistically significant correlation. The smallest detectable changes for radiography (POC-based system), radiography (manual method), 320MDCT, HR-pQCT, CBCT, and tomosynthesis were 0.020 mm, 0.041 mm, 0.076 mm, 0.077 mm, 0.057 mm, and 0.087 mm, respectively. We conclude that radiography analyzed with the POC-based system might sensitively detect minute joint space changes of the finger joint.

A Sub-Pixel Accurate Quantification of Joint Space Narrowing Progression in Rheumatoid Arthritis

“…Previously, we had proposed a JSN progression quantification method [10], [13], which is based on FIPOC [21]. FIPOC is a well-known method for image registration, it can estimate the relative displacement between two images and it is based on the frequency domain analysis.…”

“…As one of the most important indicators for the diagnosis and monitoring of RA, the joint space has always attracted extensive attention of researchers. In recent years, researchers have invested great efforts to study automatic quantification of joint space in RA [6]- [13]. The JSN progression quantification pipline in radiographs is performed in two steps; joint position detection and joint space quantification.…”

“…In recent years, machine learning (ML) based methods have become a very important tool to solve complex medical image processing tasks [14], [15]. For finger joint detection, there are some ML-based studies utilizing key point detection Purpose Requirement [17] [6], [9], [13] [16], [18], [19] Score JSW JSN…”