An Anatomical Template for the Normalization of Medical Images of Adult Human Hands

Jay Hegdé, Nicholas J. Tustison, William T. Parker, Fallon Branch, N. E. Yanasak, Lorie A. StumpoDiagnostics 2023 · DOI

TL;DR

Researchers have developed the first anatomical template for adult human hands, crucial for standardizing medical images. This template, built from 27 T1-weighted MRI scans, preserves detailed anatomical structures and exhibits superior image quality compared to individual scans, paving the way for more robust hand image analysis.

What Problem Does This Paper Address?

Medical image analysis often requires comparing images from different individuals or time points. To do this effectively, images need to be aligned to a common standard, a process called normalization, which relies on anatomical templates. While templates exist for many body parts, including the brain and lungs, a standardized anatomical template for human hands has been notably absent. This gap is particularly problematic for hand imaging because hands are highly dynamic and can change shape significantly during various tasks, leading to substantial image variability. This variability hinders accurate comparisons and can impede the diagnosis and treatment of hand-related clinical conditions, which are very common.

Methods

The researchers employed the ANTs (Advanced Normalization Tools) software framework, a widely used, open-source platform for medical image analysis. The core of their method involved a multivariate template construction workflow. This process began with preprocessing 68 T1-weighted MRI scans of adult hands. From these, 27 images deemed anatomically representative and of excellent quality were selected. These 27 images were then registered using a stepwise approach: first rigidly, then affinely, and finally using symmetric normalization (SyN) for deformable registration. This multi-stage registration aims to minimize global and local variations between the input images, leading to a more accurate and consistent template. Post-processing involved creating a binary mask to isolate the hand and setting background voxels to zero, followed by normalization to a specific bit depth.

Datasets

The study utilized 68 T1-weighted MRI scans from 34 healthy adult volunteers. Of these, 27 images from 21 subjects (13 female, 8 male) with diverse demographic backgrounds were selected to construct the template. The paper notes that children's hands and diseased hands were excluded due to anatomical differences and confounding image variations, respectively. The resulting template is made publicly available.

Key Insights and Ideas

  • The "Bootstrap Problem" for Hand Templates: Creating the first template for a new body part is challenging because good templates require similar input images, but achieving similarity computationally often requires an existing template. The authors addressed this by subjectively selecting the initial set of high-quality images.
  • Template Quality Matters: The study demonstrates that the constructed hand template has significantly higher physical image quality (measured by SSIM and MSE) than the individual scans used to create it. This is crucial because a high-quality template is essential for high-quality normalization, which in turn preserves anatomical details in the normalized images.
  • Foundation for Future Work: This initial template serves as a critical stepping stone. It can be used to normalize new hand images, enabling more objective image selection for future, potentially higher-resolution or specialized hand templates (e.g., for pediatric or diseased hands).

Key Results

The constructed hand template is a 3D image with dimensions 336x144x480, preserving essential anatomical features such as bones, muscles, tendons, ligaments, and major nerve and artery branches. Visual inspection of coronal, sagittal, and axial slices confirmed the preservation of detailed anatomy. Quantitative analysis showed that the template's image quality was significantly higher than any of the 27 individual contributing hand images, with statistically significant improvements in both Structural Similarity Index (SSIM) and Mean Squared Error (MSE).

Limitations and Future Work

The authors acknowledge that the initial selection of input images for the template was subjective, a necessary step due to the "bootstrap problem." Future work will focus on developing normalization software that can register hand images from various modalities to this template. They also plan to incorporate greater anatomical detail and enhance the demographic representativeness of the template. Further research will explore the template's utility for normalizing images of diseased hands and pediatric hands. The template, along with a 3D-printable model and an anatomically segmented version, is available for download.

Summary updated

Abstract

During medical image analysis, it is often useful to align (or 'normalize') a given image of a given body part to a representative standard (or 'template') of that body part. The impact that brain templates have had on the analysis of brain images highlights the importance of templates in general. However, templates for human hands do not exist. Image normalization is especially important for hand images because hands, by design, readily change shape during various tasks. Here we report the construction of an anatomical template for healthy adult human hands. To do this, we used 27 anatomically representative T1-weighted magnetic resonance (MR) images of either hand from 21 demographically representative healthy adult subjects (13 females and 8 males). We used the open-source, cross-platform ANTs (Advanced Normalization Tools) medical image analysis software framework, to preprocess the MR images. The template was constructed using the ANTs standard multivariate template construction workflow. The resulting template image preserved all the essential anatomical features of the hand, including all the individual bones, muscles, tendons, ligaments, as well as the main branches of the median nerve and radial, ulnar, and palmar metacarpal arteries. Furthermore, the image quality of the template was significantly higher than that of the underlying individual hand images as measured by two independent canonical metrics of image quality.
BibTeX
@inproceedings{hegde2023anatomical,
  author = {Jay Hegdé and Nicholas J. Tustison and William T. Parker and Fallon Branch and N. E. Yanasak and Lorie A. Stumpo},
  title = {{An Anatomical Template for the Normalization of Medical Images of Adult Human Hands}},
  booktitle = {Diagnostics},
  year = {2023},
  doi = {10.3390/diagnostics13122010},
}
Figure