Top > Research > Human Centered Design > Link Structure from 3D Medical Images (MRI/CT)


Link Structure from 3D Medical Images (MRI/CT)


To model human hand motion, it is important to measure and reconstruct human hand postures accurately. For this purpose, it is necessary to know the structure of the human hand's internal linkage. For example, the joint axes of two joints in each finger are not parallel to each other. This fact is well known qualitatively in anatomy, however, the quantitative observation of living human subjects has not been carried out adequately so far.

In this research, we have developed the methods to investigate human hand linkages using representative bone polygons extracted from 3-D medical images. After explaining how to derive link structures using bone polygons, we show how to reliably acquire bone polygons from multiple sets of MR images in a practical period of time. While CT is more commonly used to study bones, we use MRI because it is preferable to avoid the risk of unnecessary radiation exposure.

Internal Link Structure using Bone Polygons from 3-D Medical Images

How to Derive Joint Axes

To understand hand structure as a linkage in motion, we take 3-D medical images of different static poses. The hand's linkage structure is derived from relative movement of bones using the bone surface polygon data extracted from each volume. After aligning the root link (proximal bone) of PoseN to the link in the base pose (Pose1), we calculate the helical axis that describes the relative movement of the child link (distal bone) (Figure 1, 2, 3.)

Figure 1: Relative movement of the distal bone (child link) against the proximal bone (root link) is investigated by aligning the root link of PoseN to the link in the base pose (Pose1)
Figure 2: Helical Axis
Figure 3: Relative movement of the distal bone against the proximal bone(Index Finger)

Joint Axes Calculation Using Bone Polygons Extracted from MR Images

Figure 4 Captured Postures of MR Images
Figure 5: Bone Polygons and the Derived Joint Axes
Figure 6: Joint Axes(Index Finger)

Top ∆

Identification of the Position and Orientation of Hand Bones from MR Images by Bone Model Registration


To reliably identify bone configurations from MR images of multiple, different postures in a practical period of time

Challenging Points and Our Approach

A direct way to identify bone configuration is to segment bone regions and construct surface polygons. Automatic segmentation of hand bones is a challenging task because of the complicated intensity distribution according to the material in MR images (Figure 7). Manual segmentation ([1]) is, in contrast, robust but also time-consuming and laborious because a set of volume data for one pose includes 100 slices of image data to be processed. In addition, a single bone from any given subject cannot be extracted as exactly the same shape polygon from different sets of posture data, which causes problem when using the polygon for joint axes investigation.

Therefore, we propose the following, "registration-based," approach to identify configuration of the bone polygons.

This approach enables us to reduce the time required to acquire polygons when compared to completely manual processing.

Figure 7: Characteristics of the Hand MR Images
Figure 8: Schematic View of our Registration-Based Approach


Top ∆