Methods and Techniques

Idealized images were developed from 522 left hand and wrist radiographs of Caucasian boys and girls that were divided into 29 age groups ranging from 8 months to 18 years of age. For each age group, nine images were sorted by two independent radiologists based on the degree of skeletal maturity at different ossification centers. The middle image was then identified as the "average" image; half of the remaining images depicted less skeletal maturity and half depicted more skeletal maturity at the region examined. This process was applied to six different anatomical regions of the hand and wrist: the proximal, middle and distal phalanges, the metacarpals, the carpals, and, lastly, the distal radius and ulna. Frequently, the selected "middle" images for the six anatomical regions belonged to different children from the same sex and age groups. Computer image combinations allowed the merger of the different average images into one single representative idealized image for that age group.

For each age group, prior to creating a composite idealized image from the different selected key images, three image processing steps and enhancements were applied for standardization. First, the background was replaced by a uniform black setting and the image size was adjusted to fit into square images of 800 x 800 pixels. Second, contrast and intensity were optimized using predefined window and level thresholds. Lastly, the image was processed through a special edge enhancement filter based on an unsharp masking algorithm tailored to provide optimum sharpness of bone structure for hand-held devices.

After proper processing and enhancement, the selected images were combined to generate a single "idealized" image for each age group and for each gender. Several images, ranging from two to six, were combined by carefully replacing segments of bones through translation, rotation and warping operations to match the underlying combination image (Fig. 4). The result was a single image representing a combination of parts of hand radiographs from several images. Prior to

Fig. 14. Method of selection of the "average" image from a set of nine normal hand radiographs of a given age group. The images were sorted by skeletal maturity six consecutive times based on six anatomical regions. Each time, the middle image was noted as the "average" image for a given anatomical region for that particular age group. This often resulted in the selection of more than one and up to six different images for a given group. An idealized "average" image was then generated by combination of the selected images. Phalanges 1,2 and 3 indicate proximal, middle and distal, respectively

Fig. 15. Processing and enhancement of images. A) Original image; B) Background replaced by a uniform black setting and image size adjusted; C) Optimization of contrast and intensity

using this computer-generated image as a reference, it was reviewed by two experienced pediatric radiologists to evaluate its congruency with other reference images in the digital atlas and its compliance with existing knowledge of progression of bone maturation.

The final set of images was then resized to an image resolution that is compatible with most display resolutions of existing hand-held devices (240 X 240 pixels). Images were also compressed using a JPEG compression algorithm ensuring a reduced size of the image file down to approximately 15 to 20 Kbytes per file. Two versions of each image were generated: one depicting the entire hand and one with an enlargement of specific anatomical segments to enhance reviewer assessment of subtle details and changes in bone structures for each age group. These images provide a "zoomed" version of each image without the need for implementing complex image enlargement functions on the viewing program.

Special software for hand-held PDA devices was developed for easy navigation through the set of images and visual selection of the closest image to a given patient study. A simple user interface consisting of three buttons and a pull-down menu allows the user to easily navigate through the images either chronologically through the age groups or by selecting a specific age from the pull down list (Fig. 5). Once an image has been identified to match the image of a given patient, the user can determine the degree of bone maturity by comparing the chronological age of the patient to the age of the selected image in the atlas. The program includes a formula to calculate the standard deviation by a simple process entailing data entry of the patient's chronological age.

The basic components of the software were developed in web-compliant HTML format using standard java-scripts for some of the simple functions. This ensures compatibility with most handheld or portable computers that support a web browser capable of displaying standard HTML documents. A special variant of the software was also developed for PALM-operating sys-

Fig. 16. Example of the combination of two hand radiographs from two different individuals of the same age group into a single "idealized" image. The degree of ossification of the carpal bones of the first image were selected as more representative of the "average" skeletal maturation for this age group and were transferred to the second image in replacement of its original carpal bones

Fig. 16. Example of the combination of two hand radiographs from two different individuals of the same age group into a single "idealized" image. The degree of ossification of the carpal bones of the first image were selected as more representative of the "average" skeletal maturation for this age group and were transferred to the second image in replacement of its original carpal bones tems and for PALM-compatible PDAs. The HTML standard allows the same program to run on a laptop or desktop computer with operating systems such as Windows, Unix, Linux or Macintosh OS. In addition, a separate version of the software was developed that supports a larger image size of 512x512 pixels.

0 0

Post a comment