Automatic computerized radiographic identification of cephalometric landmarks.

Computerized cephalometric analysis currently requires manual identification of landmark locations. This process is time-consuming and limited in accuracy. The purpose of this study was to develop and test a novel method for automatic computer identification of cephalometric landmarks. Spatial spectroscopy (SS) is a computerized method that identifies image structure on the basis of a convolution of the image with a set of filters followed by a decision method using statistical pattern recognition techniques. By this method, characteristic features are used to recognize anatomic structures. This study compared manual identification on a computer monitor and the SS automatic method for landmark identification on minimum resolution images (0.16 cm2 per pixel). Minimum resolution (defined as the lowest resolution at which a cephalometric structure could be identified) was used to reduce computational time and memory requirements during this development stage of the SS method. Fifteen landmarks were selected on a set of 14 test images. The results showed no statistical difference (p > 0.05) in mean landmark identification errors between manual identification on the computer display and automatic identification using SS. We conclude that SS shows potential for the automatic detection of landmarks, which is an important step in the development of a completely automatic cephalometric analysis.

Investigation of filter sets for supervised pixel classification of cephalometric landmarks by spatial spectroscopy.

The diagnostic process of orthodontics requires the analysis of a cephalometric radiograph. Image landmarks on this two-dimensional lateral projection image of the patient's head are manually identified and spatial relationships are evaluated. This method is very time consuming. A reliable method for automatic computer landmark identification does not exist. Spatial Spectroscopy is a proposed method of automatic landmark identification on cephalometric radiographs, that decomposes an image by convolving it with a set of filters followed by a statistical decision process. The purpose of this paper is to discuss and test appropriate filter sets for the application of Spatial Spectroscopy for automatic identification of cephalometric radiographic landmarks. This study evaluated two different filter sets with 15 landmarks on fourteen images. Spatial Spectroscopy was able to consistently locate landmarks on all 14 cephalometric radiographs tested. The mean landmark identification error of 0.841 +/- 1.253 pixels for a Multiscale Derivative filter set and 0.912 +/- 1.364 pixels for an Offset Gaussian filter set was not significantly different. Furthermore, there were no significant differences between identification of individual landmarks for the Multiscale Derivative and the Offset Gaussian filter set (P > 0.05). These results suggest that Spatial Spectroscopy may be useful in landmark identification tasks.

Height measurement of astigmatic test surfaces by a keratoscope that uses plane geometry surface reconstruction.

PURPOSE: To assess the accuracy with which the Keratron keratoscope (Optikon 2000, Rome, Italy) measured astigmatic test surfaces by a profile reconstruction algorithm within a plane geometry model and to discriminate between error caused by the model and error caused by other factors. METHODS: Height was reported by the Keratron for eight surfaces with central astigmatism ranging from 4 to 16 diopters. A three-dimensional ray tracing simulation produced theoretic reflected ring patterns on which the Keratron's reconstruction algorithm was performed. The Keratron's measurements were compared with the surfaces' formulas and the ray-traced simulations. RESULTS: With a new mathematical filter for smoothing ring data, now part of the Keratron's software, maximum error was 0.47% of the total height and was usually less than 1% of local power for surfaces with 4 diopters of astigmatism. For surfaces with 16 diopters of astigmatism, maximum error was as high as 2.9% of total height and was usually less than 2.5% of local power. The reconstruction algorithm accounted for 40% and 70% of height error, respectively. CONCLUSIONS: The efficacy of keratoscopes cannot be assumed from their design theories but must be tested. Although plane geometry surface reconstruction contributed greatly to total height error, total error was so small that it is unlikely to affect clinical use.

Automatic quantitative measurement of ocular hyperemia.

Evaluation of ocular hyperemia has been an important assessment in research studies of effects of contact lenses, medications, and pollutants on the eye. Hyperemia has been difficult to quantitate objectively. The purpose of this study was to validate a computer based image analysis system to quantitate hyperemia automatically and objectively in pixelated images of the external eye using two measures, the percent of the red color, RR, and the fraction of pixels which are blood vessels, VA. Validation was against an established photographic reference scale of ocular hyperemia and against the clinical pharmacologic effects of 0.5% dapiprazole hydrochloride, known to increase hyperemia, and 2.5% phenylephrine hydrochloride, known to decrease hyperemia. Color transparencies from the reference scale were converted to digital images. Temporal and nasal regions of the external eye were imaged directly to magnetic disk before and after pharmacologic intervention. Custom software automatically excluded unwanted regions, and quantitative image analysis produced RR and VA. RR and VA were each correlated with the reference scale. For each region and for each pharmacologic intervention, the mean RR and the mean VA, respectively, were compared at time zero and at a mean elapsed time of 713 +/- 47 s. RR and VA consistently increased as the hyperemia in the reference scale increased. Pearson correlation coefficients were 0.98 and 0.99, respectively, (p < 0.01). At 713 +/- 47 s after each pharmacologic intervention, RR and VA increased and decreased as expected (p < 0.001). Thus, this study successfully validated the methodology against expert clinical judgment and was able to measure automatically and objectively clinical changes in ocular hyperemia.

Assessment of radial aspheres by the Arc-step algorithm as implemented by the Keratron keratoscope.

PURPOSE: To assess the accuracy with which the Keratron (Optikon 2000, Rome, Italy) measured rotationally symmetric, radially aspheric test surfaces according to an arc-step profile reconstruction algorithm and to discriminate between error caused by the algorithm and error from other sources. METHODS: Height, local power, and axial power calculated from radius of curvature centered on the instrument's axis were reported by the Keratron for four surfaces that had radial profiles similar to normal corneas. The Keratron profile reconstruction algorithm was simulated by using ray tracing. Keratron measurements were compared with the surfaces' formulas and the ray-traced simulations. RESULTS: The heights reported by the Keratron were within 0.25 microns from the four surfaces at less than 3 mm from the keratoscope axis and generally within 1 micron of the height calculated from the surfaces' formulas. The Keratron's axial powers were within +/- 0.1 diopter of the simulation of the axial solution between 1 and 4 mm of the axis but were greater central to 1 mm and peripheral to 4 mm. The Keratron's local powers were within -0.25 diopters at less than 4 mm from the axis and peripherally were between +1.75 diopters and -0.75 diopter of power calculated from the surface's instantaneous radii of curvature. Height error because of the arc-step algorithm was less than -0.2 micron. CONCLUSIONS: The Keratron's arc-step profile reconstruction algorithm contributed to its ability to measure height more accurately than keratoscopes that use spherically biased algorithms and provided measurement of local power.

Assessment of the power and height of radial aspheres reported by a computer-assisted keratoscope.

PURPOSE: The two purposes of this study were (a) to assess the accuracy with which a keratoscope, the Topographic Modeling System (TMS-1), calculated the heights and powers of rotationally symmetric, radially aspheric test surfaces and (b) to determine whether the TMS-1 used an axial solution for radius of curvature to determine the power of a sphere that would produce the same semichord as would the test surface on a keratograph. METHODS: The TMS-1 heights and powers were studied for four test surfaces that had radial profiles similar to those of normal corneas. The powers of the surfaces were calculated from the local radius of curvature derived from the surfaces' manufacturing formulas. The heights and powers that would result from an axial solution were calculated in a TMS-1 simulator. TMS-1 data were compared with data from the surfaces' formulas and with data from the simulation. RESULTS: The TMS-1 data were almost identical to the heights and powers calculated from the simulated axial solution. The TMS-1 data were similar to the heights and powers calculated from the mathematical formulas from the apex to 2 mm from the apex but differed by up to 85 microns of height and 10 diopters of power in the periphery. CONCLUSIONS: The TMS-1 appeared to use the axial solution that does not calculate power from local radius of curvature. Clinicians should use caution when inferring corneal shape from power maps based on an axial solution, especially outside the central 2-mm radius of a normal cornea, because such power does not depict corneal curvature.

Semiautomated nuclear shape analysis of prostatic carcinoma and benign prostatic hyperplasia.

Nuclear shape analysis has predicted outcome better than histologic grading in patients with clinically localized prostatic carcinoma. However, the requirement for manual nuclear contour tracing makes the method tedious and slow. Currently available image analysis systems for nuclear shape analysis using light-absorption microscopy provide nuclear boundaries of insufficient clarity for automatic segmentation. We improved image resolution using confocal laser scanning microscopy, automatically detected nuclear boundaries by a multiscale segmentation algorithm and discriminated artifacts in a semiautomated way. A manual quantitative morphometry system and our semiautomated system distinguished eight cases of prostatic carcinoma from seven cases of benign prostatic hyperplasia by nuclear roundness factor, ellipticity, nuclear area and perimeter. The ease of semiautomated nuclear shape analysis should allow evaluation of large numbers of patients with known outcomes after treatment for clinically localized prostatic carcinoma to determine whether nuclear shape analysis can be extended from research to clinical usage.

Visual information processing of computed topographic electrical activity brain maps.

Effective display of computer-generated biomedical images draws on computer graphics and image processing, display technology and human factors, visual psychophysics and perception, cognitive psychology, and the new field of scientific data visualization. In converting from raw, acquired data to a visual display, developers need to know the limitations of the data and of the display technology. To obtain reliable inferences about the clinical or physiological state of the patient requires that the computer display be matched to the visual information-processing competence and limitations of human observers. The issues that should be considered by both developers and users of computer-based display technologies to enhance clinical performance in observation and diagnosis are surveyed with reference to electrical activity brain maps.

Development and application of a three-dimensional artificial visual system.

A three-dimensional artificial visual system has been developed to aid in the analysis of 3-D fluorescence images of smooth muscle cells. The system consists of three sets of 3-D spatial filters that decompose the image to enable a simple recombination algorithm to locate the discrete bodies of protein concentration in a cell, classify the concentration bodies as globular or oval, and determine the 3-D orientation of the oval bodies. A graphic model of the protein concentration is created from the data provided by the artificial visual system. Patterns of organization in the distribution of the protein bodies are investigated using an interactive graphics system.