What question did this study set out to answer?

The aim is to compare segmentation-based and deconvolution-based methods for analyzing joint space width in hand and wrist joints using CT scans.

March 25, 2026Open Access

Evaluation of the use of segmentation-based and deconvolution-based three-dimensional joint space width analysis of the hand and wrist

Key Points

The aim is to compare segmentation-based and deconvolution-based methods for analyzing joint space width in hand and wrist joints using CT scans.
Computed tomography scans of hand and wrist joints were analyzed using segmentation and deconvolution methods.
Joint space width parameters (mean, max, min, etc.) were calculated from each method.
Agreement between methods was assessed using coefficient of determination (R2) and bias measures.
General estimating equations evaluated the relationship between morphology and radiographic scores.
Agreement between segmentation and deconvolution methods showed low R2 values ranging from 0.42 to 0.87.
Biases were noted up to 0.18 mm ± 0.33 mm between the methods.
Increased joint space narrowing indicated by Sharp/van der Heijde score was associated with lower JSW.Min and higher JSW.SD for both methods.
Deconvolution method revealed a stronger correlation with JSW.SD compared to segmentation when related to SvdH scores.

Abstract

Abstract Three-dimensional morphometric analyses of joint space from computed tomography (CT) scans can provide a quantitative and unbiased evaluation of radiographic damage from arthritis. This potentially offers greater sensitivity for disease assessment and monitoring than traditional ordinal scoring systems. Previously established methods for three-dimensional joint space analysis from CT scans can be categorized into segmentation-based and grayscale deconvolution-based methods. However, segmentation-based methods have not been employed for joints other than metacarpophalangeal joints, nor have they been compared to deconvolution-based methods. This study compared the joint space width (JSW) distribution from various hand and wrist joints in photon-counting CT images, using both measurement approaches. From these distributions, the mean (JSW. Mean), standard deviation (JSW. SD), maximum (JSW. Max), minimum (JSW. Min), 95th percentile (JSW. 95%), and 5th percentile (JSW. 5%) values were calculated. For each morphometry parameter, agreement between the two methods was assessed by calculating the coefficient of determination (R2), and the mean and standard deviation of the individual biases. General estimating equations were used to assess the relationship between morphometry and an equivalent/modified Sharp/van der Heijde (SvdH) ordinal radiographic score. In general, agreement between the two methods was low: R2 ranged between 0. 42 and 0. 87; biases up to 0. 18 mm 0. 33 mm). For both methods, an increased degree of joint space narrowing as assessed with SvdH was associated with reduced JSW. Min and increased JSW. SD; for the deconvolution method, it was also associated with decreased JSW. Mean and increased JSW. Max. Interactions between the methods showed a weaker association to SvdH score for JSW. Min and a stronger association for JSW. SD when measured with the deconvolution method, compared to the segmentation method. We conclude that both methods lead to fundamentally different measurement outcomes, and that the deconvolution method might be more sensitive to radiographic damage.

Read Full Paperexternally

Bookmark

View Full Paper

Cite This Study

Quintiens et al. (Sun,) studied this question.

synapsesocial.com/papers/69c37adcb34aaaeb1a67cca7 https://doi.org/https://doi.org/10.1093/jbmrpl/ziag048

Bookmark

View Full Paper