We attempt to measure image quality to determine the performance of an optical system so that its range of applicability and the information content of the data collected can be specified. Why is there any problem? Mainly because we always attempt to use as simple a measure as possible to describe what is in fact a fairly complicated process. Furthermore as various measures have been used (and abused!) the technology of system design has always been more sophisticated than the measure was designed for. A simple example of this is the two-point resolution criterion first propounded by Rayleigh. The criterion which only truly applies to the resolution of two incoherent individually unresolved objects is directly applicable to the specification of performance of a telescope in viewing star fields; it is also applicable (under carefully specified and controlled conditions) to a microscope viewing fields of individual cells. However, it is not of much value in specifying performance of a telescope viewing the moon, mars or other planets; or to a microscope viewing a general biological specimen.
When our eyes follow an object that moves in space, a number of very complex control systems operate in order to maintain a sharp image at a specific retinal location while keeping the overall retinal illumination more or less constant. In the eye movement control system, it is the distance between retinal image (or area of interest in this image) and the fovea that is minimized. Therefore, the object we fixate with our eyes is always imaged onto that part of the retina with the highest visual acuity. In the pupillary control system, it is the average light level that is measured and kept within comfortable limits by changing the pupil diameter . Finally, in the accommodation control system, certain image errors are evaluated (presently, we do not know exactly which ones) and reduce(' by changing the lens power. In all these control systems, the retina plays a crucial role as the first stage of the error detector. However, the retina is a far more complex structure than the familiar photo-sensitive surfaces in physical light detectors. There are relay stations (bipolar cells, ganglion cells) in each signal channel between a photoreceptor and optic nerve fiber, and abundant anatomical and physiological evidence exists for interactions between signal channels at various levels in the retina (horizontal cells, amacrine cells). These facts have to be kept in mind whenever detector properties of the eye are discussed. Another complication arises from the fact that we can never define the output of the system exactly. In a photocell it would be a single variable, for example, the cathode photocurrent. In the human eye, it is an already coded signal that is transmitted over approximately 10 optic nerve fibers to higher centers of the central nervous system. It is impossible at the pre-sent time to monitor simultapeously signals transmitted over large numbers of optic nerve fibers.
The photometric fidelity of the optical system of microdensitometers is compared to that of flying - spot scanning devices. Microdensitometers, which were originally developed to study fine structure of silver halide photographic images, move the object plane through a fixed, Koehler-illuminated dual projection optical system that controls scattered light and ensures uniform photo-metric response everywhere on the scanned area. Flying-spot and video-based scanners now being applied to photographic photometry are optical reciprocals of one another. They operate on a critically-illuminated image of the film similar to that produced by a camera, and therefore require precautionary measures to limit errors from flare light and their inherent lack of spatial stationarity. The low inertia of such systems permits high data-sampling speeds with the possibility of preprogrammed or adaptive scans under computer control. Advances in stage design, however, now allow microdensitometer optical systems also to sample at photon noise limited rates. Conditions under which image-plane scanners can be used for quantitative microphotometry of photographically-recorded images are outlined.
The volume of the heart as a function of time over a period of several full cardiac cycles can be used in conjunction with concurrent pressure measurements to obtain important indications of myocardial function: total work output, efficiency, stroke volume and so forth. Of the four heart chambers, the left ventricle (LV) is of primary importance medically since it is the systematic pumping chamber of the heart and consequently provides most of the work output. For this reason, the LV is the object of this study.
Last year at your convention in Anaheim we reported our first experience with a new optical system, the so-called "rod-lens" system, designed by Professor Hopkins (England) for medical purposes and is produced by Karl Storz Company (W. Germany). In every disease of hollow deeply situated organs which have some sort of outlet to the surface, for instance the gullet, windpipe, the urinary bladder or back passage, etc. a rigid tube can be introduced. It was already discovered in the last century that there are many diseases which cannot be assessed by other means than with direct visual inspection and this is called "endoscopy", which means looking inside. For the examination of the lung, for example, a tube is introduced through the mouth into the trachea to keep it apart. A small little light carrier is placed on the working end. The Wattage of this globe is approximately 3. The examination of the inside of the organ is performed by naked eye. The length of these instruments are 400 mm. (16") which is not optimal, by any means, for the human vision. In addition, the dim light interferes with the accurate vision of our receptor system.
I want to thank the Society for inviting me to present this material to you. I hope that it will give you some idea of what we can do now, what some of the technical problems are, and perhaps, stimulate your interest and solicit your help.
Reliable studies show that under routine conditions radiologists fail to detect approximately 30% of abnormalities present on chest x-ray films. These failures of detection are perceptual in nature and not due to limitations in such physical parameters as resolution or contrast. There are many image processing techniques that are being proposed to help the radiologist. Most of these techniques have been implemented by us in TV process-ing of x-ray films. Each processing technique was evaluated experimentally using radiologists as observers in a test based on concepts originating from signal detection theory. A numerical value for d', an index of merit, was assigned to each technique according to how well it improved the accuracy of the radiologists' decision making.
Studies of the mechanics of motion of the human body have developed to the point where accurate determination of body segment parameters is essential. However, little data has been available on body parameters such as mass, center of mass, and moment of inertia of the major body segments.
This presentation describes an experimental method of image contrast en-hancement adaptable to a variety of photo interpretative disciplines including radiology. Because of the relatively low order of spatial frequency resolution required for interpretation of radiographs, it is in this area that the principle can first be brought to a state of practical application. In the course of the work reported here, a laboratory bread-board, satisfactorily demonstrating the usefulness and practicability of the contrast enhancement technique, was fabricated and evaluated.
The application of neutron radiography to imaging biomedical objects is in the initial stages of development. A possible application of fast neutron radiography is the imaging of air-filled cavities embedded in tissue, particularly in the vicinity of bones. A fast neutron radiography facility has been developed using an accelerator to produce 14.7 Mev neutrons by the T(d,n) reaction and a divergent collimator (six-inch-square radiograph area) in a large water tank to reduce interference fast neutrons at the image point. Test objects were constructed of plexiglass (nearly tissue equivalent) in the form of stepped slabs with holes in each step ranging from 1/32 to 1/4 inches in diameter. The objects with holes were radiographed at several distances from the film to determine the resolution capabilities of the facility. The slabs with holes were also placed behind several thicknesses of plexiglass and radiographed to demonstrate the ability to detect air-filled cavities buried in tissue. The radiographs were produced by direct exposure using a scintillator-film detector. The detail was improved by computer aided image analysis which increases the contrast, an apparent contribution in minimizing the dose levels required. The results show that cavities can be observed in configurations approximating pratical specimens.
Xeroradiography is the technology for recording radiographic images without the need for any chemical processing whatsoever. To make an image a selenium photoreceptor, or plate, is charged to a uniform surface potential on the order of 1000 volts. When this photoreceptor is exposed to x-rays, its surface charge is depleted by photoconduction in amounts proportional to the radiation intensity striking it. A residual potential pattern which is a shadow graph of the object being x-rayed is then formed. After exposure, the electrostatic image on the plate is developed by a powder cloud technique in which charged, pigmented particles are blown over the plate and adhere to it in the pattern of the image. This powder or toner is the ink which defines the image. Following development, the toner is transferred to paper and fixed into a permanent image by low temperature heat fusing. The purpose of this paper is to present the principles of zero radiography, to describe practical equipment utilizing these principles for x-ray imaging in the medical and industrial fields, and to show some examples of the types of images produced by the system.
Xeroradiographic and radiographic film image analysis through application of the radiographic modulation transfer function defined in this paper is a powerful analytical tool. This tool has been found to predict spatial and densitometric characteristics of an x-ray image for an arbitrary set of input object exposures and contrast variations with sufficient accuracy to allow meaningful comparisons of different types of radiographic imaging systems. This paper presents a simple computerized technique for the derivation of the radiographic modulation transfer function from the image of a knife edge test object. Techniques used for assessing the degree of system linearity over the output response range of interest and the role of the system phase response are also presented. Typical xeroradiographic, screened and unscreened film radiographic modulation transfer functions derived with the program are compared and discussed in the conclusion to the paper.
Radiological images constitute an important source of diagnostic information in medical practice. The radiological images presently available to the medical profession generally suffer from low contrast and poor resolution. Many factors contribute to the poor image quality. Some of them are inherent in the nature of the object and the basic radiological imaging process and some of them are due to imperfections or degradations of the system. In the past, little could be done in the way of restoring or enhancing the image once it had been recorded. Now, recently developed techniques allow the image to be made more intelligible to the human observer by correction or compensation of the degradations which it has undergone. Based on the knowledge of some identifiable characteristics of degrading influences and some known characteristics of the object, these techniques perform a sophisticated mathematical operation on a recorded image to undo the effects of whatever degradations may be present and restore, as nearly as possible, the image which would have been recorded in the absence of degradation. They correct various geometrical, photometric and spatial frequency distortions. Sometimes, even with a perfectly recorded or restored image the information content cannot be fully extracted or retrieved because of the limitations of human vision. Image processing techniques will allow the image to be purposely distorted in frequency domain and/or gray levels to suppress background, to emphasize some special features, etc.
Close-range photogrammetry has experienced considerable growth during the past few years, and have shown potential for even. greater growth in the future. The increased acceptance of properly calibrated non-metric data acquisition systems such as simple (non-metric) cameras, T.V., and X-rays in photogrammetric work is expected to open the door for many engineers and scientists to make full use of the technical and economical advantages of photogrammetry. The author presents a number of examples to illustrate the metric potentials of simple cameras in close-range photogrammetry, and discusses the various approaches used. To ensure reliable and precise results, three basic requirements must be fulfilled: careful calibration of the camera, sufficient control in the object space, and the use of the analytical approach in data reduction.
Photogrammetry was applied to help biological research in determining the growth rate of living jelly fish colonies bred under different laboratory conditions. The examined species of jelly fish is a tiny roundshaped animal with diameter up to three milimeters. It is grouped and reproduced in colonies. Individual colonies are kept on slides submerged in sea water. Pairs of close range photographs were obtained from a press camera that did not meet all photogrammetric requirements. It was, therefore, indispensable to calibrate each individual picture. A special self-calibrating system was designed for this purpose. This made it possible to perform the calibration directly from the measurements on the photographs of the colonies. The techriique of photogrammetric evaluation on the Wild A-7 plotter was adapted to cope with the specific two-media conditions of the photography (salt solution and air between the object and the camera). The theoretical analysis of the case led to the use of a relatively simple photogrammetric treatment of photographs with affine-transformed bundles of projecting rays in the plotter.
The anatomical sciences have traditionally relied primarily upon descriptive terms in analyzing the form and shape of various parts of the body. Although much insight has been gained by this approach as to the functional and dynamic roles of the various body components, the lack of quantitative data in defining these same structures has become a definite hindrance in attempting to understand more precisely their mechanisms of action. The application of photogrammetry to the analysis of body form and shape provides us for the first time with an opportunity to represent any external portion of the body as a system of coordinate points of the X, Y and Z axes. Thus, with appropriate mathematical manipulations, these values permit one to calculate the volume of a given structure and make comparisons of position and symmetry between different structures of the body as well as members of various sized populations of people. Some specific examples of this application carried out to date by Dr. Herron and his associates include the determination of the volume of amputated stumps of either the upper of lower extremity as an aid in fitting prostheses, the efficacy of various modes of therapy for lymphedema as a result of surgical removal of the breast, and the early detection of changes in the curvature of the spine of children in anticipation of preventing the development of idiopathic scoliosis.
The recent development of the scanning electron microscope (SEM) has added a new dimension to the field of microscopy. The SEM produces an image having considerable depth of focus; for example, approximately 400 times that of a light microscope at equivalent magnifications. It can be operated continuously between magnifications of 20 and 100,000X with a resolution of approximately 100 Å.
Historically, heart research has been directed at characterizing the heart as a pump. The hemodynamic parameters used in early studies have included pressure, flow, stroke volume, and cardiac output, to name a few. In the last decade, however, a new emphasis has been placed on studying the heart as a muscle. It has become increasingly apparent that a large number of compensatory mechanisms are at work simultaneously to stabilize the hearts's pump parameters, and that only after failure of one or more of these mechanisms is it possible to measure significant changes in these hemodynamic parameters. A growing body of evidence supports the concept that an early, sensitive measure of cardiac 'health' is provided by direct measurement of the dynamics of the muscular walls of the heart. The work in our laboratories is directed toward this goal.
A method for the quantitative assessment of contractional patterns of the urinary bladder in dogs is described. It involves a mirror well in which the bladder is suspended. 44 photograph of the well gives five orthogonal projections of the bladder surface. Sequential pictures allow movements to be detected and analysed.
Reseachers seeking a histological determination of microscopic specimen morphology face the basic problem of analyzing three-dimensional microscopic specimens with two-dimensional imaging devices. This tedious task can be done at the microscope but reproduction of images in hard copy or digital form entails the loss of depth information. Quantitative measurements must generally be made on 2-D images. Several methods have been proposed and implemented to obtain and display 3-D images. In this paper we discuss an approach combining digital image processing techniques with stereo display techniques. This allows control of image quality and selection of different 3-D characteristics for display.
A frequent problem in evaluating a technique for measurement of the human body is the lack of a standard of comparison. Even though determination of volume by water displacement, an unwieldy procedure for large objects, can yield accurate measurements, the difficulties of maintaining comparable volumes in the lungs during water immersion and photography of a living human call for a more stable calibration object. Regular objects may be used to test a system's capability under ideal conditions, and a calibration involving volume determination of a regular object of known dimensions was also carried out as described below. Yet the system must be evaluated in terms of its performance in measuring the human body form. Therefore, our laboratory tests were centered around the photogrammetric determination of the volume of a full size male mannequin, whose volume was also measured by immersion.
One rather unexpected result of the past decade's manned space flights has been the growing public awareness that the earth is akin to a space ship; both are small, life-sustaining bodies surrounded by an environment hostile to life; both are closed ecological systems that require diligent house keeping and both have finite, exhaustible resources. At the same time, the public has been warned that the earth's ecological system is in a state of stress, that some natural resources are close to depletion, and that with increasing population, our quality of life may face a decline within the lifespan of generations now living. To prevent such a decline and to maintain a beneficial environment far more information about the earth's resources and the effects of industrialization and urbanization are required than now possessed. This accumulated information must be continuously updated and maintained by more effective management systems in the future than in the past. Both aircraft and spacecraft will be used. Sequential use of aircraft sensor data will provide baselines for studying regional problems; larger areas could be covered by earth orbiting satellites.
In the past few decades the medical community has become increasingly adept at, as well as increasingly dependent upon, diagnosis by radiography. However, there exists a growing concern over the possibility that cellular and genetic damage is being done by the large amounts of x-radiation turned upon the populations of medically advanced countriesI . The state of the art has considerably reduced the dosages required to obtain satisfactory radiographic images in an attempt to reduce doses to a safe level. However, radiobiologists have studied the effects of low radiation doses on different cell lines in vivo in the hopes that some threshhold would be found below which permanent damage ceases to occur. Unfortunately there are now grounds for believing that the probability of genetic change follows the dose down to nearly zero.
The problem of mating the pictorial world of bio-medical imagery with the digital one of computers is a difficult one. Basicly, the problem is one of assimilating massive quantities of raw data into the computer and then operating on that data at a rate which is fast enough for clinical decisions.
The production of three-dimensional images using the techniques of optical holography is a well-established phenomena. The invention of the laser in 1962 as an intense source of monochromatic, coherent illumination precipitated much research in the use of such fields in imaging. Specifically, Dennis Gabor's approach to imaging using coherent illumination and recording the resulting interference pattern or hologram as he called it, has become a familiar technique to individuals working in the field of ootics.
Since the audience for this discussion is at least partially comprised of life scientists, it may be appropriate to begin by mentioning some facts of microscopy and physical optics with which they may not be completely familiar. These facts impose constraints under which any endeavors in magnifying and imaging, whether holographic or otherwise, must labor.
The relationship between holo-graphy and stereo-photography is described with emphasis on the metric characteristics of holographic imagery. A mapping and mensuration system, with a self-illuminated measuring mark, is used to extract quantitative information from three-dimensional holographic virtual images. Quantitative results and plots are presented for a variety of objects which are recorded holographically at close range. Results of mensuration and mapping from holographic images of dental castings and bone joints demonstrate the practicality and potential of the discussed techniques. Holographic stereomodels from stereo-pairs of photographs have been successfully produced and work is in progress to apply the mapping techniques to them.
Acoustical holography is a relatively new method for imaging which is beginning to emerge as a promising technique for the nonionizing, noninvasive visualization of soft tissue structures in biological subjects. With further improvement it could become a valuable method of visualizing internal organs and soft tissue structures in medical diagnostic imaging. Acoustical holography evolved as a logical extension of the principles of optical holography to sound waves. In this paper the basic principles of acoustical holography are explained in physical terms so that the nonmathematically inclined reader new to the field can obtain a physical understanding of how holography works. Some of the methods which are currently being used successfully to obtain images of biological subjects are described.
Thermography is concerned with the pictorial representation of infrared radiation emanating from an object. The instrument that produces visible images of this thermal radiation is called a thermograph, and the permanent recordings of these images are called thermograms. The thermograph may thus be characterized as being an electro-optical imaging device that operates as a wavelength converter, translating spatial and thermal information about an object to the visible spectrum from that of the infrared. In addition, special processing techniques may be employed to enhance or identify certain signal characteristics as an aid for the user in evaluating the displayed data.
This report describes current results of a research and development project whose goals are to develop a prototype system for the automated analysis of dental radiographs. This system uses a flying spot (cathode-ray) scanner connected on-line with a computer for the purpose of automatically performing the pattern recognition of features found on panoramic dental radiographs.
As a photographic system, television has three unique advantages: 1) It provides a real-time, remote, visual display of the object field; 2) a hard copy of the imagery can be obtained at the remote station without actual access to the TV camera; and 3) the video signal from the TV camera can be digitized directly and thus providing an excellent potential for real-time geometric measurement. Largely because of these reasons, television has been used extensively in space exploration, in medical, biological and ballistic research, and in the supervision of industrial processes. Because of its high sensitivity, television is also used in astronomical and underwater observations. However, in all of these applications, TV imageries have been used primarily for qualitative analysis.
Some of the basic knowledge and skills of photo-optical engineering can often be applied in a broad manner to generalized problems where such applicability might be only casually suspected. The authors have been concerned with such problems in muscle physiology. Representative examples are:
In the last few years there has been a dramatic increase in the interest of the medical community with human cytogenetics. This interest has been generated by the demonstrated relationship between chromosome abnormalities and human pathology, and there is the expectation that this discipline will provide an even better understanding of the human organism.
This paper is concerned with the automatic calculation of left ventricular volumes from a cineangiogram by a digital computer. It will first describe three digital techniques (1) to restore the original dimension of the objects by a logarithmic transformation, (2) to subtract images before and after the injection of radio opaque agent, and (3) to detect boundaries by a threshold method based on statistical principles and heuristics. Experimental results on several frames will be presented to demonstrate the feasibility of these techniques. Once the boundary of the ventricle is determined, the computation of the volume is performed by the usual Simpson's method.
An interactive computer system for analyzing biomedical pictures which operates under the IBM PLAN (problem language analyzer) System is described. Various programs commonly used in picture processing such as the filtering program and image transforming programs are integrated in the system and selected to load at the execution time through problem-oriented commands. A CRT display unit is used to monitor the processed image data to pick up the objective entities and to set up the processing parameters.
Vladimir Kratky: I am extremely pleased and honored indeed to be here. What I have to say is necessarily from the point of view of a photogrammetrist and I hope you will keep this in mind when you respond to my remarks.