Overview

ApX™ is a suite of imaging development tools from Amerinex Applied Imaging, Inc. (AAI). Especially designed for original equipment manufacturers and value added resellers, ApX lets developers build and deliver medical imaging applications in Visual Basic, Visual C ++, Delphi, or Java. In addition to extensive image processing and analysis libraries, ApX provides controls for creating graphical user interfaces and example VB and VC + + applications, which can serve as jumping-off points for medical imaging application development. ApX's dual interface capability makes this imaging functionality available as a set of ActiveX controls callable from any software environment supporting the Microsoft COM technology, or as a set of dynamic-link libraries (DLLs) callable from Visual C + +.

ApX provides functionality for medical imaging applications that require analysis, automatic segmentation, or recognition. Operator libraries allow users to create totally automated systems that mimic visual processing. ApX users can draw on a collection of adaptive line and region segmentation functions to automatically extract even low-contrast and ambiguous objects of interest from images. In addition, ApX's library of morphology operators is useful in separating touching or overlapping objects, filling in holes, and in extracting specific shapes of interest from image data. Other ApX segmentation operators, such as the elastic contour algorithm, solve the problem of accurately and automatically finding missing or low-contrast object contours and are thus useful in segmenting MRI and CT datasets.

A unique feature of ApX is the capability to separate and differentiate objects within the data and extract features that are essential for tasks such as quantitative analysis, treatment planning, and disease monitoring. To support feature-based processing, pixel-based image data must be transformed into explicit symbolic representations of the objects. ApX provides adaptive and robust functionality for the extraction, quantification, storage, retrieval, and processing of both pixel-based and symbolic information contained in the imagery.

ApX supplies this support via a proprietary data structure-oriented processing level called the Intermediate Symbolic Representation (ISR). In addition to supporting object processing, the ISR can also be used to support multimodal imaging applications that need to represent and process information from different kinds of images such as CT, MRI, thermal, visible, and X-ray. The ISR is an advanced and flexible system for processing data structures such as points, lines, regions, chains, and other objects, giving users an extension to pixel-only based image processing solutions (Fig. 2). When performing tasks such as pattern recognition or matching, the use of symbolic representations reduces computation time because the imaging algorithms operate with small sets of objects rather than large arrays of pixels. These representations also allow application developers to construct solutions that match image data to real-world objects in a straightforward and natural manner.

Recognition, the ability to train software to recognize specific kinds of objects or data in images, can be based on statistics derived from large sets of training data or discrimination rules derived from a few prototype examples. Most of the recognition tools available for imaging have been designed for industrial applications where parts are well defined and have little variance. Medical imaging systems, however, need to recognize structures that have great variability in form and contrast, such as cells and organs. ApX's statistical pattern and fuzzy logic capabilities are well suited for such applications. This functionality can use shape, size, color, texture, and other characteristics to discriminate between objects of interest and other nonessential or spurious information. For example, ApX recognition operators can be trained to distinguish among different types of cells, or to distinguish normal from deformed cells (Fig. 3).

At press time, AAI had the following application-specific controls available:

Masking

This tool provides point-and-click masking of areas of interest, eliminating the need for manually tracing around an object. The ApX mask grower automatically creates a mask (outline) for an object of interest from user-selected background and foreground points.

Elastic Contours

This application component is based on active contours (snake) used to fit pliable contours to object boundaries. This ApX control supports the development of applications that require outlining areas of interest, such as organ or tumor boundaries. With ApX snake technology, users need to supply a seed region, which could be provided automatically or by having the user roughly circle the item of interest in an image. From there, the ApX algorithms perform detailed fitting. This tool can also be used with 3D datasets to automatically grow contours in consecutive slices in CT and MRI data.

Gauging

This ApX tool supports instantaneous width measurements in gray-scale images with subpixel accuracy. Application uses include measuring blood vessel diameters, bone width, and spacing between vertebrae. Because fitting a gauge tool is subject to user error, the gauge control provides precise, repeatable measurements through automatic edge extraction and fitting algorithms. Users need only click on an item of interest and the tool conducts all computations.

Deconvolution

The deconvolution module allows users to take a set of images that has been blurred by the imaging process and invert the blurring to obtain a clearer image, such as sharpening images from three-dimensional microscopy. The ApX deconvolution control employs an AAI proprietary optimized algorithm for fast performance on large datasets.

FIGURE 2 The Intermediate Symbolic Representation (ISR) concept of ApX.
FIGURE 3 Cell recognition with ApX. See also Plate 139.

Lighting Correction

Lighting irregularities often make it difficult to accurately process medical images. The ApX lighting correction control corrects for nonuniform light fields such as those found in endoscopic images. The module can also perform flat field correction.

0 0

Post a comment