Frontal radiographs of scoliotic deformity of the spine reveal a characteristic pattern of lateral deviation, lateral tilt and axial rotation of vertebrae. In order to study interrelations between deformation parameters 478 radiographs of idiopathic scolioses, 23 of scolioses after Wilms-tumor treatment and 18 of scolioses following poliomyelitis were digitized. From these the curves of lateral deviation, tilt and rotation are calculated and fitted by Fourier series. By restriction to the first harmonic, analysis reduces to the analysis of a single phase and amplitude for each curve. Justification of this simplification will be discussed. Results provide a general geometric description of scoliotic deformity.

Analysis of three-dimensional back shape enables the calculation of the so-called symmetry line, which is closely related to the line of the spinous processes on the back surface. A comparative photogrammetric measurement proved a mean deviation in the order of the palpation error. Furthermore, the angle of the surface normal on the symmetry line proved to be equal to the angle of vertebral rotation within the limits of radiographic accuracy. These findings enable a three-dimensional reconstruction of the spinal midline. Scoliotic deformity may thus be determined solely from surface data. In frontal projection the mean error of the reconstructed midline is about 4 mm as compared to standard radiographs.

The Integrated Shape Investigation System (ISIS) is a structured light scanner and shape analysis system, developed as a safe alternative to follow-up radiographs for the clinical assessment of deformities of the human back. The system is described and results presented of several clinic studies. These show a significant correlation between ISIS measures and conventional radiographic measures of spinal curvature, such as the Cobb angle. The development of a predictor for deterioration in adolescent idiopathic scoliosis, based on surface shape weasures, is discussed.

Rasterstereography can be improved if photographic recording is replaced by video imaging. Video cameras with a CCD sensor, connected to a video interface supply a digital image in real-time, which subsequently can be processed by a microcomputer within a few minutes. Precautions necessary to obtain photogrammetric accuracy are discussed in detail. An actual system for the measurement of human body is outlined, and some results concerning accuracy and resolution are presented.

Optical techniques avoid mechanical contact with the cartilage and therefore its subsequent deformation. They are well suited for the 3-D vision of articular surfaces shape. Three techniques which have in common the fact to produce contour lines on the surface are reviewed. Holographic interferometry (HI) by refractive index change, by spatial frequency modulation and moire topography are compared with respect to the particular application to anatomic preparations. With HI by spatial frequency modulation the major problem encountered is the dehydration of the specimen which induces its deformation. HI by refractive index change involves level lines with a depth resolution easily adaptable to the curvature of the surface to be studied. The most effective technique is moire topography and we show that the result obtained is within an accuracy acceptable with respect to the wide biological variations encountered. This stays even when the contour lines do not represent strict level lines.

Modern diagnostic imaging techniques are providing three-dimensional images by the combination of analog sensing devices, powerful digital processors and graphic displays. Computer based optical imaging systems are used for detection and tracking of body deformities in Orthopaedics. To establish a morphometric data-base means for comparing and averaging similar shapes have to be prepared. Assuming fast technological advancements use at present and prospective applications are given.

Visual defects of glaucoma have been associated with retinal nerve fiber layer (NFL) defects, which can be detected as color and texture difference compared to normal NFL using photographs of NFL. Frequently these defects are asymmetric. A large defect exists in one eye while the other eye maintains its normal features. These evaluations of NFL defects, however, are subjective and two-dimensional. We have quantitated NFL thickness using photogrammetry and have measured its asymmetry in the eyes of human subjects.

Local and global fit may be used to estimate radius of curvature. Computed radius shows high frequency noise in the local fit case and may be biased by the fitted model in the global case. Al priori knowledge may he used to impose constraints to the parametric model. To further improve radius estimation, splines are fitted to the filtered residue of the previous fit. Radius of curvature measurements by these methods, applied on cornea images, are analyzed in term of repeatability and justness.

The equipping of the Kern Digital Restitution Instrument (DSR) with CCD cameras and image processing hardware allows to use digital techniques to measure automatically the third dimension in a model. This can be done with the so called vertical line locus method. The vertical line locus method was introduced to measure the height in aerial photographs, but it also can be used to determine the height in medical or industrial applications.

Noncontact measurements of the topography of irregular biological surfaces are important needs for biological investigators. Direct measurements with a mechanical probe are not recommended or are not feasible, especially when the biological surface interacts with another physical medium. Existing stereophotogrammetric techniques are usually designed for static measurements and are based on complex optical systems, or require lengthy computations to produce accuracies which are beyond the practical biomedical needs. This work presents a novel noncontact technique for dynamic visualization and measurement of three-dimensional surfaces of moving boundaries for applications to biomedical studies. The system is similar to rasterstereography, but involves advanced methods of image processing and linearizations that simplify the geometry reconstruction procedure. The accuracy of the measured three-dimensional geometry is better than most practical applications in biomedical engineering. Measurements of the geometry of a collapsible tube under static and dynamic conditions are presented.

The abnormity of the quality and quantity for erythrocytes is one of the important changes of blood disease. It shows the abnormal blood-making function of human body. Therefore, the study of the change of shape of erythrocytes is the indispensible and important basis of reference in the clinic, diagnose and research of blood disease. In this paper, a preliminary discussion is made on the acquisition of scanning stereographs for erythrocytes, the application of the theory of photographic measurement on the three dimensional space quantitative analysis of erythrocytes, drawings of isoline map and section map of various erythrocytes for normal persons, paroxysmal nocturanal hemoglobinuria (PNH) patients and aplastic anemia patients, study of the shape characteristics of normal erythrocytes and various abnormal erytnrocytes and the applications in clinic, diagnose and research. This research is a combination of microphotogrammetry and erythrocyte morphology. It is polssible to push fotward the study of erythrocyte morphology from LM, SEM to a higher stage of scanning electron micrographic photogrammetry(SEMP) for stereograpic observationand three diamensional quantitative analysis to explore a new path for the further study of the shape of erthrocytes.

We have assessed the techniques of silhouette imaging, laser scanning triangulation and moire fringe contouring as implemented at the National Engineering Laboratory (UK) for application in 3-D facial imaging. Using model heads under varying clinical conditions with a maximum exposure of a few seconds the suitability of each technique was evaluated.

A method study to improve automatic 3-D data acquisition from stereo images of the human face by using correlation equipment from Kern is presented. Soft tissue morphology can be meas-ured in three dimensions by using close range photogrammetry. The collection of height information in a regular grid pattern was performed automatically utilizing the correlator of Kern DSR-11 analytical plotter. To improve contrast a random point generated pattern was projected onto the patients face. The correlation algorithm used was the "Vertical Line Locus" (VLL). The original Kern program has been modified, and by choosing suitable correlation window size, the number of succesfully correlated points was considerably increased.

With a stereo arrangement of CCD cameras the shape and size of human faces are determined. The 3D-reconstruction algorithm is based on the Multiphoto Geometrically Constrained Matching technique which allows for precise and reliable solutions. The image matching procedure is fully automatic. Results of different object visualization methods are compared to each other: natural facial texture and structured light. The 3D-data is passed over to a finite element surface modelling program, which allows to compute derived information like profiles, single points, grid points, perspective views, etc. The technique and the algorithms may also be applied to other problems in biostereometrics.

In vivo measures of bite force and bone strain obtained in growing African green monkeys (Cercopeithecus aethiops) are being used to study skull biology and geometry. Strain values and distributional patterns seen in association with forceful jaw elevation are inconsistent with conventional explanations linking upper facial morphology with masticatory function and/or using beam models of craniofacial architecture. These results mandate careful use of notions about skeletal geometry based on static analyses that have not been experimentally verified using in vivo procedures.

A method for the detection of differences between two three-dimensional digital models of a single object for medical purposes is discussed. The data used are X, Y, Z co-ordinates in the form of digital "terrain" models presumed to be captured at different epochs. Aspects of the least squares approach are discussed, along with a number of perceived medical applications.

As a rule continuous bony defects of the mandible after tumour resection are temporarily bridged with the AO reconstruction plate. To precontour the plate, CT-scans of the mandible at a scale of 1:1 6 mm above and parallel to the inferior margin are needed. Corresponding layers further cranial show the anatomy of the chin. Using a lateral cephalograph, the mandibular angle and the length of the plate in the ascending ramus are measured. Of the 19 preshaped reconstruction plates implanted, 11 did not need any intraoperative adjustment of shape. The other 8 plates needed only slight modifications of contour. Preoperative bending of the reconstruction plate reduces the time taken for the operation, protects the material and the plate is precisely shaped.

Reliable software for automatic determination of the border between the cancellous bone and the cortical bone of lumbar vertebrae has been developed. An automatic procedure is needed because calculations in a larger series of patient data take too much time due to the inevitable human interaction required by available software packages. Processing in batch mode is essential. An important advantage of automatic outlining is its reproducibility, because only a single technique with objective criteria is used. In a so-called Region Of Interest (ROI) texture analysis can be performed to quantify the condition of the vertebral body in order to diagnose osteoporosis. This technique may be an alternative to a classification based solely on the average X-ray absorption value.

Algorithms for interactive display and measurement of three-dimensional (3D) surface anatomy have been developed and implemented on a PC-based system with a high-speed display architecture. Surfaces are extracted automatically from two-dimensional cross-sectional images and rendered as three-dimensional shaded-surface images at a series of user-specified viewing angles. Surface images are displayed interactively from any of the precomputed views, producing the effect of arbitrary tilt and rotation of the 3D anatomy. Dimensional measurements are obtained in real time through selection of visible surface points in one or more views. The techniques described provide an interactive framework for both visualization and quantitation of 3D surface anatomy.

For preoperative planning of corrective hip osteotomy, quantitative assessment of the coverage and congruency of the femoral head is essential. Based on serial CT-sections, a computer-assisted method has been developed providing for three-dimensional modelling of the in vivo situation and subsequent evaluation of the hip joint. The cartilage structures extracted from each cross-section are triangulated and reconstructed into surfaces. New algorithms were developed to cope with the problems caused by the contour data structures. They include the automated connection of open and closed contours, contour bifurcations and capping of terminating contours. Surgical procedures are simulated using the surface data and a quantitative evaluation of the coverage and centering of the hip joint is achieved. The tiling method presented is not limited to CT-data and can be applied to other structures analyzed with contours.

Numerous interviews with traumatologic and orthopedic surgeons showed us how important and helpful definite information about the spatial topology of the situs would be for the surgical disciplines before the actual operation.

A statistically based relaxation method is applied to segmentation of abdominal X-ray CT-scans. To incorporate prior knowledge about the anatomical structure a stochastic model representing the geometry of the abdomen is used. Properties of each tissue class with respect to X-ray imaging are modelled by mean grey value distributions. The performance of the segmentation procedure has been tested on a set of 18 scan-sequences.

A schemata for 3-D reconstructions from images of serial sections of biological tissues is presented. A classification map is constructed from statistical analysis of populations of n-tupled image pixels. 1-tuples are assigned optical properties such as refractive index, optical density, hue, and saturation. Image sections are transformed into classified sections using the map and then stacked. The data set is processed by a program cognizant of the laws of polychromatic white light propagation through space.

In spite of care taken in endoprosthetic joint replacement, migration of prosthetic components from fractions of a mm to one cm or more can occur. To detect such "micromotion" in a clinical environment, it is proposed that the positions of radiopaque markers implanted in both the prosthesis and the host bone and positions of their images on two or more conventional X-rays be used in a mathematical model. One implementation illustrates that estimates to within 0.5 mm are possible using prototype models.

Microenvironments which can be sampled using clinical imaging devices have distinct physical chemical characteris-tics. Clinical diagnostic procedures have, for some time, exploited these characteristics to highlight relations, con-tinuity of structure, and coloration in tissue samples. As demonstrated by our experience with decision support systems for polysomnography, rather simple inferential engines can recognize the same sorts of characteristics which have been exploited in pathologic procedures to assign properties to voxels in a clinical image.

This paper provides a light introduction to simple graphics techniques for visualizing a large class of biological shapes generated from recursive algorithms. In order to capture some of the structural richness inherent in organisms, the algorithms produce not only extreme variability but also a high level of organization. The material primarily comes from previous published works of the author. For a general background on fractal methods in mathematics and science, see Mandelbrot's famous book. For research on the fractal characterization of other biological structures, such as the lung's bronchial tree and the surfaces of protein molecules.

Although in Magnetic Resonance Imaging (MRI) there is no fundamental restriction to axial sections, specifying an arbitrary orientation and introducing the relevant information in the pulse sequence generator are not trivial. We have developed a solution for cardiac imaging. It consists of a general double oblique pulse sequence and a specialized user interface. The basic problems of double oblique MRI, the specific questions of cardiac imaging and the requirements of the user interface are discussed.

MEDnet, a project sponsored by the National Center for Supercomputing Applications and the University of Illinois, is an advanced computational science and health care telecommunications project. State-of-the-art computational tools and advanced technologies, accessible through MEDnet, can optimize biostereometric data collection, data transfer and the reconstruction of various metrics and images. This national network could also serve as a clearinghouse for biostereometric data, as well as survey hardware and software suitable for academia, industry and health care providers.

Morphometric assessment of brain structures has been shown to be a reliable indicator of pathology. However, prior studies have been inconsistent or even contradictory due to partial volume effects and operator dependence. A preliminary system to automate the measurement process was developed for, and delivered to, Long Island Jewish Hospital. The operator's role is to choose the structure to be measured and to oversee the automatic extraction. A discrete Stokes theorem is used to compute the partial volumes. System accuracy of 2% has been demonstrated.

A photogrammetry technic has been developed to examine the trunk's movement during quiet breathingand under different breathing manoeuvres: a rectangular network projected on the illuminated surface of the trunk and on a screen behind the subject examined . By taking a series of pictures from two directions we can determine the position of each surface points. Construction of horizontal slide surfaces at different phases of the breathing cycle made possible to calculate the volume change , the velocity of displacement of certain surface points or the movement of a representative mamillar line , where several types of breathing can be characterized.

The olfactory system of the rat is a useful model for the study of mammalian sensory systems. However, the anatomy of the nasal epithelium, where the cells responsible for detecting odors are located, is extremely complex. Therefore, we have focused our attention on the development of two- and three-dimensional automated imaging methods. The presentation of pure odorants to the experimental animal together with the injection of [¹⁴M-deoxyglucose has been combined with autoradiography of frozen sectioned material. Several approaches have been used to obtain optimal alignments of the digitized images of the sections so as to be able to generate appropriate 2-D and 3-D reconstructions. Such reconstructions allow visualization of the ethmo-turbinal bones (turbinates) and the associated soft tissue and appear to be useful in analyzing and highlighting differential metabolic activity.

The main steps of the method used were: graphic tablet input and storing of heart contour's data. Construction of 3-D polyhedra. Volume calculations, color imaging, setting any visual angle of obtained perspective hidden graphics. Accuracy was checked on phantom model and a ± 13 % interobserver error was found.

In this paper we present a laser imaging system based on the flying spot scanner concept associated with a computer image processing system of relatively high resolution. The horizontal and vertical deflection of the laser beam is realized by a pair of galvanometers guiding two perpendicular axis oscillating mirrors. A programmable electronic deviation system commands the galvanometers. The reflected laser light from the scene is detected by a high gain photomultiplier improving the signal to noise ratio. Because the total power of the laser beam is used, such a system is able to act with very low light power. Also it gives a large depth of view and offers zoom facilities. Its association with an image frame buffer and processing system offers other advantages such a good resolution and image enhancement. The whole is connected to a host computer for control and processing tasks.

In order to study the biological reaction of patellar tendons from which a graft had been harvested, we have developed a ultrasonographic methode that furnishes three transaxial and one longitudinal scan of the operated and the non operated knee for comparison. Width and thickness can be measured in mm for scaling by a metric device in the scanner. A specially created computer-program using a definite digitizing system allows a very accurate calculation of the patellar tendon-volume for conclusions about recovery of the dimensions before graft harvesting.

35mm photography of denture teeth and resin replicas through a convergent-axes microscope was used in an assessment of wear in dental restoration materials. The difficulty was to isolate and evaluate the significant photogrammetric parameters, but thereafter, the required depths could be calculated to accuracies of 0.01 mm r.m.s. using stereocomparator observations and quite simple formulae. The technique is applicable to biomedical laboratories which have access to an appropriate microscope if photogrammetric observations can be undertaken.

There is a growing demand among electron microscopic users to extract spatial information from photographed micro objects. Methods utilizing the well established photogrammetric theories either using analog or analytical approaches have been successful with electron micrographs. However, these methods are rather time consuming particularly for these applications that require mapping of large number of micrographs. In this paper an interactive display system (Semper V) composed of a camera an IBM/PC microcomputer together with software developed by Image Techniques of Cambridge Ltd is used as an image processing system performing image enhancement and filtering, masking, histograms, as well as gray scaled contouring. Transmission electron microscopes produce highly magnified images that resemble X-ray photography revealing different gray levels proportional to the thickness of the object, thus making them very suited to the prorosed system: Comparison of mapping a microscopic object using Kern DSR 11 analytical plotter and the proposed interactive display system is presented. Results proved to be comparable in accuracy while the time saving is dramatic making this system very attractive to several users.

A new method of calibrating tomodensitometric systems developed at Laval Univer-sity is described. This is based on the basic concept that each image represents parallel projection geometry for the sectional plane it represents. Scale affinity and inherent distortions are mathematically modeled. A series of parallel consecutive images taken of object placed in a specially designed plastic cube gives a cost-effective solution. The obtained results indicate a standard error of around ± 0.2 μm in each of three dimensions.

The advantage of EM over X-ray diffraction when investigating the molecular structure of biological matter is its ability to gather an image (i.e. a phase and amplitude representation) and not just a diffraction pattern (i.e. amplitudes only) of the specimen. Its disadvantage is given by the fact that, despite the near-atomic (i.e. 2-3 Å) resolution performance of state-of-the-art instruments, the practical resolution with most biological specimens is typically limited to 10-30 Å. However, with a few "cooperative" specimens significant structural detail at the 5-10 Å resolution level has been obtained. Hence, in most cases studied to date, the EM has only provided us with relative representations (i.e. the overall size and shape) of protein molecules rather than with their absolute (i.e. near-atomic) structure as this is now commonly achieved by X-ray diffraction analysis. These limitations are primarily specimen-dependent: (1) biological matter has relatively low inherent contrast in an (e.g. 100 kV) electron beam; (2) the alterations (e.g. denaturation and collapse) accompanying the preparation (e.g. dehydration) of biological material for inspection in the high vacuum of an EM cause serious specimen preparation artifacts; (3) biological specimens are extremely radiation sensitive when bombarded with electrons.

Constructive solid geometry has been used to estimate the shape and inertia of the human body. Hanavan I used geometric solids to represent each segment of the body and calculated moments of inertia from the dimensions and estimated mass. Jensen improved on the shape representation by increasing the number of geometric solids per segment, with two cm wide elliptical cylinders of given density and major and minor axes measured from orthogonal 35mm photographs of the supine subject. Changes in segment principal moments and other parameters, with age, between 4 and 20 years have been reported for 88 male subjects. The accuracy of this approach for some segments was questioned by Hatze who proposed diverse geometric solids to accommodate the more complex surfaces. A battery of 242 anthropometric measures was used and inertia parameters reported for four subjects. Computerized tomography has also been used to estimate segment shape and inertia. Ackland et al using six scan sites and the leg segment of a 29 year old male and an embalmed 65 year old cadaver, reported that the error associated with the estimation of volume was greater than the error due to the assumption of uniform density. It appears then, that more exact procedures for estimating segment volume may lead to improved estimates of segment inertias. The purpose of this study was to develop a simple technique for determining the surface coordinates at pre-set elevations on an object in order to investigate further the effect of shape representation on segment inertia estimations.

A simple low-cost microdensitometer is described, based on a mirror stereoscope as it is used in aerial photographic interpretation. The eyepiece head is modified to take two photosensors whose light beams are reflected by a beam splitter while the other parts are passing the eyepieces. Variable apertures in front of the photosensors allow pointwise and surfacewise density measurements. The instrument is used for both near and remote sensing interpretations and measurements and for ground truth purposes of vegetation evaluations primarily for leaf physiological examinations, such as leaf surface changes.

Moving targets are tracked in space by two CCD-cameras, which are oriented using control points. Image coordinates are determined automatically by thresholding and correlation methods. To solve the correspondence problem, nearest neighbor and direction criteria are applied in the image plane, and epipolar contstraints in the stereo setup. Accuracy can be tested by a moving target produced by a plotter. These algorithms run in the window environment of a Sun-workstation equipped with image acquisition boards.

For human motion analysis, the spatio-temporal resolution of cinematographic registrations of body marker positions is still higher than the results of the best opto electronic systems available for this purpose today. So far, the need for manual digitization of several thousand marker positions per tested person has made this method unpractical for regular applications. An interactive and largely automated system for marker recognition and tracking from 16 mm film images based on progress in digital image processing has been developed and tested. Projected pictures are digitized with a high-resolution CCD-camera (1320x1035 pixel), processed, analyzed and serially evaluated with an interactive image analysis system SIGNUM IS200.

The stereophotogrammetric analysis of biological forms in medicine has been usefully employed for many years. Generally the methods adopted have involved the use of stereo pairs of traditional photographs or radiographs. Although very successful, the traditional methods are time consuming. Developments in solid state and videoscan cameras, in computer hardware and in image processing technology have lead to Real-time photogrammetry (RTP) or near Real-time photogrammetry (NRTP). It is anticipated that RTP and NRTP will play a vital role in biostereometric studies. This paper discusses the development of a low cost PC-based NRTP system and the possibility of using NRTP in an ongoing biostereometric study which has been developed for studying regional body surface motion in the evaluation of respiratory muscle function and in which digital cameras will be substituted for analogue cameras and computer digitisation will be substituted for manual observation.

Inclusion of mass distribution information in biomechanical analysis of motion is a requirement for the accurate calculation of external moments and forces acting on the segmental joints during locomotion. Regression equations produced from a variety of photogrammetric, anthropometric and cadaeveric studies have been developed and espoused in literature. Because of limitations in the accuracy of predicted inertial properties based on the application of regression equation developed on one population and then applied on a different study population, the employment of a measurement technique that accurately defines the shape of each individual subject measured is desirable. This individual data acquisition method is especially needed when analyzing the gait of subjects with large differences in their extremity geo-metry from those considered "normal", or who may possess gross asymmetries in shape in their own contralateral limbs. This study presents the photogrammetric acquisition and data analysis methodology used to assess the inertial tensors of two groups of subjects, one with spastic diplegic cerebral palsy and the other considered normal.

Ten fresh cadaveric whole cervical spine specimens (occiput to C7) were studied using well established techniques to document the movements in flexion, extension, left and right lateral bending and left and right axial rotation. Pure moments of maximum 1.5 newton meters were applied incrementally and three dimensional movements of the bones were recorded using stereophotogrammetry. Each moment was applied individually and in a three load/unload cycles. The motion measurements were made on the third load cycle. Parameters of neutral zone, elastic zone and range of motion were computed. Neutral zones for flexion/extension, right/left lateral bending and right/left axial rotation were respectively: 1.1, 1.5 and 1.6 (occiput-C1); and 3.2, 1.2 and 29.6 degrees (C1-C2). Ranges of motion for flexion, extension, lateral bending (one side) and axial rotation (one side) were respectively: 3.5, 21.0, 5.5 and 7.2 degrees (occiput-Cl joint) and 11.5, 10.9, 6.7 and 38.9 degrees (CI-CZ joint). The highest intervertebral motion in the spine was the axial rotation at the Cl-C2 joint, neutral zone constituting 75% of this motion.

For a deeper understanding of the complexity of human walking movement not only a kinematic analysis , but also a comprehensive three-dimensional biomechanical model of the human body is required to detail the kinetic activities. This research combined static stereophotogrammetric determination of body segment mass parameters with three-dimensional gait analysis by cinephotography, direct linear transformation and two force plates. A method of combining the two independent analyses by defining the anatomical axes of each segment is shown. Practical problems arising in dynamic and stereometric analysis are demonstrated. Power spectra of a normal and a matched subject with spastic diplegia were calculated for a proper design of the kinematic analysis.