Andrology Lab Corner*: Reflections on CASA After 25 Years
2004; Wiley; Volume: 25; Issue: 3 Linguagem: Inglês
10.1002/j.1939-4640.2004.tb02793.x
ISSN2047-2927
Autores Tópico(s)Ovarian function and disorders
ResumoLike the Journal of Andrology, computer assisted sperm analysis (CASA) systems did not arise de novo. Although both are celebrating 25th anniversaries, the Journal evolved over several years and drew on predecessors. In contrast, today's CASA systems represent third-generation devices for visualization and analysis of sperm motion. Modern CASA evolved some 300 years after the first-generation device was placed into use. This device was the light microscope, which von Leeuwenhoek used to first visualize sperm in 1678. The concept and principles underlying such a device did not change until shortly before World War II, when European opticians developed phase-contrast optics. Such second-generation devices were first used by pioneering andrology labs in the mid-1950s, and phase-contrast microscopes remain the primary instruments for observation of living sperm. Phase-contrast optics are integral to every contemporary CASA system because they enable high-contrast visualization and edge detection of each translucent cell. Important publications on quantifying sperm motion appeared between 1940 and 1970 (summarized by Boyers et al, 1989), and provided a foundation for CASA systems. However, the approaches in these studies were not at all automatic, and most used manual cartography. By the early 1970s, convergence of technology and government priorities set the stage for development of CASA. Federal and private investments in tracking rockets and diverse objects on the macroscale increased significantly. Computer technology, user friendliness, and cost began to improve exponentially. Video cassettes replaced the original reel-to-reel technology. As a result, computerized video image digitization, recognition, and quantification technologies began to emerge, with substantial cost savings over prior methodologies. Applications to the microscope followed and primitive CASA systems emerged. Here we consider the motives of individuals and companies pioneering CASA, and comment upon whether their expectations were realistic (with the benefit of hind-sight) and met. We note the positive and negative impacts of CASA in sperm biology, clinical medicine, and epidemiology. We also reflect on the role of high-technology devices in the practice of andrology. We restrict our considerations to measurement of sperm motion, although current CASA systems can measure number of sperm per unit volume and can be modified to capture data appropriate for morphologic classification of each sperm examined. CASA refers to an automated system (hardware and software) to visualize and digitize successive images of sperm, process and analyze the information, and provide accurate, precise, and meaningful information on the kinematics of individual cells, and also population summary statistics, that is, mean values. Early systems required operator intervention, but preferred systems would require the operator only to insure that the system is functioning properly, place the sample into the instrument, and examine/store output data. Underlying concepts of CASA are illustrated in Boyers et al (1989). Beginning in the 1940s and continuing for decades, a few university scientists recognized the need to obtain objective data (ie, bias-free) on percentage of motile sperm and, ideally, velocity of movement of spermatozoa. They were driven by the desire to establish standards useful to retrain or train individuals making subjective evaluations in a commercial setting (ie, animal genetics companies), and for objective data to enhance research on sperm function. Routine use in clinical andrology was not an immediate goal, although some clinicians had recognized limitations of visual observations of sperm motility. It was believed that if precise and accurate data on sperm movement could be obtained, this information could be used to predict the potential fertility of a male or select a "best procedure" for sperm preparation. As early as the 1950s, it was appreciated that electronic technology could be developed or adapted to the measurement of sperm motion (Van Demark et al, 1958). Several different technologies were developed to infer estimates of average velocity of sperm in a suspension without actually identifying the swimming trajectory or measuring velocity of individual cells. These approaches included assessment of disruption of light passing through a pinhole by moving sperm heads (van Duijn and Rikmenspoel, 1960), analysis of scattering of light from a laser directed at a sperm suspension (Dubois et al, 1974), and use of an early image-analysis computer to count fluctuations in sperm numbers in a fixed volume (Katz and Dott, 1975). However, these all were indirect methods that did not identify and track individual sperm cells. In the late 1940s, Lord Rothschild introduced the use of time-exposure photomicrographs, using dark-field illumination, to create images of the swimming trajectories of spermatozoa, which could be manually analyzed to determine swimming velocity (Rothschild and Swann, 1949; Rothschild, 1953). During the 1960s and 1970s, this technique was used in a number of contexts, including analysis of bull (Elliott et al, 1973) and human (Janick and MacLeod, 1970; Overstreet et al, 1979) sperm. This approach has been modernized by use of a digital camera (JL Schenk and RP Amann, personal communication). There also were a number of studies that identified sperm trajectories via frame-by-frame projection of cine films (eg, Rikmenspoel, 1957; Katz et al, 1978) obtained micrographically. However, these studies and similar ones using videotapes still required extensive manual work for raw data acquisition as well as subsequent analysis. These manual cartographic studies established two important points: 1) human observers were biased when estimating percentage of motile sperm; and 2) information on pattern and velocity of sperm motion indeed was of biological significance and possibly clinical utility. Acceptance of these conclusions provided motivation for seeking automated cartographic analysis of sperm trajectories. In 1973, Jecht and Russo reported that a motion-analysis system developed for the National Aeronautics and Space Administration at the Jet Propulsion Laboratory could track human sperm. Videotape interfaced a microscope with the analysis system, and operator input was obligatory. Although there was no comparison of multiple samples or a follow-up publication, this paper included concepts still used today (eg, determination of sperm centroids, linear and angular velocities, and linear and angular displacements). At this time, Amann at Penn State recognized the need for automated quantitative measurement of percentage of motile sperm. He approached commercial bull studs in Pennsylvania with a proposal to make computerized measurements, and his colleague Hammerstedt sought assistance with the requisite computer programming. With additional local and federal funding, plus efforts of a dedicated student and several technicians, the first presentation of a system designed to track sperm motion was made at the Third International Conference on the Spermatozoon in Woods Hole, Mass, in 1978 (Amann, 1979) and it utilized software developed by Liu and Warme (1977). Because real-time video-capture boards and high-speed recording hardware cost >$200 000, the Penn State team recorded primary data on motion picture film (a step back from Jecht and Russo, 1973). The film was projected frame-by-frame on a screen so that a video camera could capture each stationary image over several seconds (reasonably priced reel-to-reel video recorders introduced image distortion) and move data for digitization and storage on a 33-cm diameter hard disc (1.2 mb). The computer (16 kb memory) required 3 minutes to analyze a sample. Output data based on 4 or 5 frames were considered adequate to gave meaningful data on percentage motile sperm and velocity. This system was subjected to comprehensive validation (Amann and Hammerstedt, 1980), and that paper set a standard for validations of other systems. As summarized later, this system was the first using computerized cartography, rather than manual cartography, to provide output data for production of training/educational aids or publish linkage with fertility of individual males. Schoevaret-Brossault (1984) introduced a similar approach with human sperm, and his system analyzed 30 frames and provided more comprehensive output on sperm movement characteristics. The first system enabling direct transfer of video information from a microscope into a video-capture board, followed by automatic image processing and data output, was described by Katz et al (1985). The heart of this system was an Expert Vision™ system developed by Motion Analysis Corporation for study of macroscale (human ambulation) and microscale (marine microorganisms) movements. The authors emphasized that there was useful information in measures of vigor and pattern of sperm motion (eg, curvilinear velocity, average path velocity, and linerarity) as well as percentage of motile sperm. Profit was the goal of commercial developers of CASA systems, and this was predicated, primarily, on sales to clinical laboratories serving human patients or processing human or animal sperm for use in artificial insemination (AI). The pioneering Expert Vision system (Motion Analysis Corp; see above), and its descendents, several generations of CellTrack™ systems, apparently were not widely used, despite certain technological advantages over then competing systems. The first commercial CASA system developed specifically for evaluation of sperm motion was the CellSoft™ system (CRYO Resources Ltd), which sold and distributed a number of units starting in 1985. Over the next several years, a flurry of publications described use of the CellSoft system with bull, human, mouse, and rat sperm (eg, Mathur et al, 1986; Working and Hurtt, 1987; Budworth et al, 1988; Mack et al, 1988). At this time, it began to be recognized that CASA data had potential application in detecting effects of environmental and, later, occupational hazards on sperm function (Toth et al, 1989). Both the Expert Vision and CellSoft systems used free-standing phase-contrast microscopes with heated stages, conducted real-time video capture, and provided unattended analysis after image capture. The second commercial system developed specifically for evaluation of sperm motion was the HTM-2000® (Hamilton-Thorn Research), a "system in a box," introduced in 1986. The impetus for development of this system was quantification of changes in stallion sperm during 1–5 days of storage in a shipping container then in development (marketed as Equitainer) by a physicist, with profit a secondary motivation. This system had technical advances including an integrated optical system and video display, keypad controls, and automated positioning of the sample to predetermined locations. The near-infrared illumination and dark-field optics of the initial system soon were replaced, in the HTM-S®, by visible-light illumination and phase-contrast optics. Focusing directly on a video screen eliminated the problem of accommodation of the human eye, an adaptation making cells above or below the 16 μm depth of field of the objective appear in focus even when they would not present a sharp image to the video camera. User inputs later led to special chambers with a fixed depth of 20 μm, wherein all cells are sufficiently sharp to allow digitization. The CellSoft and HTM-S systems both provided meaningful output data (Gill et al, 1988), although terminology and calculations were different, and direct comparisons of numerical values could be problematic. Despite the many elements of automation in these systems, there still was a need for user intervention to teach the instruments the "best" settings for software parameters to track sperm from a given species under conditions used in that laboratory (eg, Knuth et al, 1987). These settings critically affect the outputs from an instrument. Technological issues (hardware and software) inherent in automated capture and processing of image data were considered by Boyers et al (1989), and most remain addressed by compromise. Characteristics of the Expert-Vision, CellSoft, and HTM-2000 systems were compared in Amann (1988), along with validation and experimental data for bull or stallion sperm with the latter 2 systems. In due course, CRYO Resources ceased operation, and Motion Analysis Corporation abandoned the area of sperm analysis. Hamilton-Thorn introduced the IVOS® integrated system in 1992, and the companion CEROS® for use with an available microscope and computer. Innovations of the IVOS included: strobed light-emitting diode illumination to provide sharp images and, therefore, more accurate and precise image digitization; presets to enable tracking of rodent sperm; automated classification of sperm undergoing hyperactivated movement; and optional use of internal fluorescent illumination and fluorochrome—DNA-stained sperm, so that these cells could be distinguished unambiguously from other objects. In parallel with development of commercial systems, university-based researchers developed systems in the 1980s (eg, Stephens et al, 1988). Today, in addition to the IVOS system, at least 2 other commercial CASA systems are in use. The SM-CMA system (MTG GmbH), developed for sperm analysis early in the 1990s, is the only system known to utilize detection of the sperm middle piece as a secondary factor to decide if an immotile object indeed is an intact spermatozoon, or to incorporate algorithms allowing proper extensions of the paths of 2 cells whose trajectories intersect or enter a region of uncertainty. The Hobson Sperm Tracker (Hobson Sperm Tracking, Ltd) was introduced in the mid 1990s, and apparently it evolved from software developed for use with microorganisms. The IVOS system has evolved to allow concurrent classification of each sperm as motile or nonmotile and also permeant or exclusionary to a vital dye (D Douglas-Hamilton, personal communication). Volume 8 of the Journal of Andrology included several "firsts." Although it is possible that we missed an earlier abstract of a CASA presentation at an annual meeting of the Society, there were such presentations in 1987. More importantly, full-length publications reported use of the ExpertVision system with human sperm (Katz and Davis, 1987) and the CellSoft system with rat sperm (Working and Hurtt, 1987). The index for volume 8 was the first to include "computer," thus linking computers and sperm motility. One goal of pioneers or early users of CASA systems was to obtain motility data free of bias, to establish standards for training, testing, and retraining of personnel making subjective evaluations. There is mixed evidence on achievement of this goal. The system described by Liu and Warme (1977) was used to prepare a videocassette tape with scenes of bull sperm with known/stated average percentage of motile sperm and average cell velocity for retraining laboratory technicians, plus other scenes lacking outcome information for testing. Although this project was funded by bull studs, apparently they were dissatisfied with the product or decided that elimination of bias among or within laboratories was unneeded or not worth the effort. This negative response was duplicated by bull stud personnel and a veterinary school after preparation of a far more modern and comprehensive teaching/testing tape in the mid 1980s, using a then-current CASA system (RP Amann, personal observations). Nonuse of CASA-based training aids undoubtedly is due to the fact that animal genetics companies and practicing veterinarians are not required to demonstrate competency in sperm analysis to a regulatory group. With respect to human sperm motility, a videotape training aid on semen evaluation was marketed (Fertility Solutions Inc) in 2002, and production costs have been recovered (S Rothmann, personal communication). This vendor is preparing to introduce an enhanced teaching and quality-control "calibration" product specific for sperm motility, in a compact disc format. Both federal regulations on quality control and Clinical Laboratory Improvement Act oversight now include competency testing. This requirement will increasingly motivate laboratories evaluating human semen to use digital video disc-based images of human sperm with "correct" or "gold standard" observations on the basis of CASA (but see comments on "gold standards" below) for demonstration of proficiency. Hence, the early goal of educational use of CASA soon could be widespread. Although routine use of CASA in clinical andrology was not an immediate goal, this application was the major force behind product development. CASA systems never would have attained their current use if each was "homemade." Product introduction, improvement, standardization, and marketing by companies were driven by the implicit goal of profit. Today, systems are in approximately 1200 sites worldwide, primarily in human andrology laboratories and often in conjunction with an in vitro fertilization facility. Hamilton Thorne Biosciences has units in 342 sites in the United States and >750 in foreign countries (D Douglas-Hamilton, personal communication), with most market growth in large human clinics and for "line-speed" evaluations in large animal genetics companies (eg, pigs). Actual market penetration of CASA in US laboratories performing human semen analysis might be approximately 2%, because most semen evaluations are performed in general clinical pathology laboratories and not in andrology laboratories. Below, we discuss why operation of an andrology laboratory without a CASA system is not unreasonable. Placement of CASA systems in veterinary laboratories has been nil, although recently there has been entry into bull, boar, and stallion stud farms as well as large equine clinics. It is likely that the clinical market will grow slowly and the semen processing market (both human and animal) might approach saturation. One company apparently dominates the US market and shares the European and Asian markets with possibly 2–3 others. Presumably at least 1 company has met their goal of profit from sales of CASA systems. The second impetus for developing CASA was acquisition of objective data to enhance research on sperm function, selection of a "best procedure" for sperm preparation, or to predict the potential fertility of a male. After the precision and accuracy of CASA were established in numerous publications, it was logical to use CASA to measure effects of medium or processing procedure on aspects of sperm function. This use is especially important for species in which direct, prospective fertility testing is impractical (eg, horse, dog, endangered species) or ethically not possible (ie, human). A number of such studies have been published. Typically, authors consider treatments resulting in the highest percentage of motile sperm with the highest velocity and most linear movement to be "best." There is no unequivocal biological logic to support this assumption. In interpreting CASA data, we cannot unequivocally state whether a high mean value or a minimum dispersion of values about the mean is best. Once it was evident that precise and accurate data on sperm motion could be obtained, prediction of potential fertility of a male, on the basis of CASA data, became a major focus of research. Apparently the first attempt to link outcome data from CASA with pregnancy rates for individual males involved the system described by Liu and Warme (1977). Straws of the same semen were used for CASA evaluation (as in Amann and Hammerstedt, 1980) and for AI of dairy cattle. Relative fertility of 9 bulls was established by parentage after heterospermic AI of equal numbers of sperm from 2 bulls, in different combinations (this approach has more statistical power than conventional AI). Correlations between CASA data and relative fertility were not substantially better than those based on other visual or subjective measures of sperm quality (Saacke et al, 1980; O'Connor et al, 1981). Subsequent evaluations of other straws of the same semen with a CellSoft system, after careful validation, revealed only modest correlations between the competitive fertility index and either percentage of motile sperm or curvilinear velocity (Budworth et al, 1988). However, multiple correlation analyses, each including 6 parameters measured by CellSoft at 0 or 1.5 hours after thawing the semen, gave correlation coefficients of ≥0.94. The same paper included a study of the correlation between percentage of cattle pregnant 75 days after commercial AI with semen from 1 of 10 dairy bulls. Although data were based on 620–900 females/bull and pregnancy rates encompassed a range of 18 percentage units, no correlation was significant. This contrast in magnitude of correlations between CASA data and fertility data likely was due, in part, to the manner used to measure "fertility," as discussed by the authors and also in Amann (1989). The many other papers reporting linkage of CASA and fertility data, for animals or humans, are not reviewed because with the benefit of hindsight it now is obvious that it was unrealistic to expect any one or combination of attributes of sperm motion to be predictive of pregnancy rate achieved by any individual male. Note that prediction cannot be achieved via correlation analysis. The underlying problem is that malfunction of any one of many essential and independent sperm attributes can render a given spermatozoon incapable of fertilizing an oocyte. Thus, adequate motility is a necessary characteristic, but alone is insufficient to insure fertilizing potential of a spermatozoon or population of cells. Different sperm fail for different reasons. Hence, quality of sperm motion in the laboratory has limited value because it provides little information on adequacy of other known and unknown attributes (Amann, 1989; Amann and Hammerstedt, 1993). Further, and equally important, such linkage demands precise measurement of fertility under conditions allowing expression and detection of male-associated differences in fertilizing potential of sperm in a seminal sample. As detailed elsewhere (Muller 2000; Amann and Hammerstedt, 2002), pregnancy is not a good measure of fertilizing potential of sperm because nonsperm factors dictate if a 1-cell embryo becomes a fetus or living young. This problem is compounded by AI of so many sperm that male-to-male differences often are masked, even when averages are based on sufficient females (Amann and Hammerstedt, 2002). Many of the original expectations of CASA have been met, and in general the impact has been favorable. There is no doubt that, when properly calibrated and used with appropriate software parameter settings, a CASA system can provide both accurate and precise data on sperm location in successive video images. Especially when image capture is at 60 Hz and the number of sperm in the field of view is appropriately low (to avoid crossing of cell trajectories), the path of each cell can be computed with appropriate accuracy and precision. However, algorithms for secondary calculations (eg, smoothed path velocity, linearity of motion, amplitude of lateral head displacement, beat cross frequency) are compromises that provide understandable but perhaps not the most accurate or meaningful data. Percentage of motile sperm can be established accurately and precisely, but the threshold demarcation between a nonmotile or static spermatozoon and a motile spermatozoon apparently is set arbitrarily (perhaps based on common sense; eg, 10 μm/s) rather than on the uncertainty (upper 99% or 99.5% confidence limit) for replication of location of the centroid for a series of truly immotile (killed) sperm in a sample. Similarly, demarcations between slow and medium or medium and fast sperm are arbitrary. Accuracy and precision of CASA systems have allowed detection of subtle changes in sperm motion and, hence, improved discrimination among treatments in a laboratory study of new seminal extenders, cryoprotectants, or other steps in processing (eg, centrifugation) on aliquots of a given sperm suspension. However, there has been general reliance on mean values for each parameter although the distributions for many data sets for >200 sperm evaluated per sample are not normal (van Djiun and Rikmenspoel, 1960). Appropriate transformations (eg, Gladen et al, 1991) always should be used, and comparing distribution plots might be more informative than comparing averages (eg, Toth et al, 1989; Vantman et al, 1989). More importantly, in a typical study, data for 6–12 attributes for sperm in the same samples are subjected to statistical analysis without consideration that this increases likelihood of a type-I error (incorrectly concluding that a difference is significant). Further, many of the measured attributes are not independent. Finally, there is no experimental basis for deciding that a change of a given magnitude in a given measure of sperm motion is of biological importance as contrasted to the change being statistically significant. The former is much more important, yet the precision of CASA systems enhances the probability of detecting changes unlikely to have "real world impact" on success of a given spermatozoon in its quest to fertilize an oocyte. Although there is no answer to this dilemma, the discussion of "how much is enough" in Amann and Hammerstedt (1993) is appropriate in this context. Application of multivariate analysis to CASA data does not eliminate this latter problem, despite utility to increase statistical significance of relations (eg, Gladen et al, 1991). The accuracy and precision of a properly operated CASA system should give it a strong role for calibration of scenes of swimming sperm used to prepare aids for implementation of quality-control (QC)/quality-assurance (QA) programs. It has taken 20–25 years for this important role to emerge, at least for human sperm, and it likely will take years for such calibration products to enter use in most laboratories evaluating human semen. In our opinion, preparation of CASA-based calibration standards should emphasize percentage of progressively motile sperm and, hence, consider the combined effect of a "robust" velocity with a "reasonably linear" path of swimming. At least as important in teaching visual sperm analysis is control of temperature at 37°C (CASA data show a marked difference in percentage of motile sperm at 20– 22°C as compared to 37°C), use of a phase-contrast microscope, and preparation of the slide to be viewed. An emerging and very logical use of CASA is in laboratories processing human or animal semen for use in AI, especially via cryopreservation. This allows more confidence in adjusting number of motile sperm/dose to maximize number of doses prepared from each ejaculate. Even more important is the capability of CASA to discriminate small differences in sperm quality during postthaw evaluation of "test doses" from each batch processed. Even if CASA data are not highly predictive of fertilizing potential of a given batch of semen (see discussion elsewhere herein), the discrimination power is greater than that provided by a human observer, and this might have biological impact via decisions to cull or use. Perhaps the greatest unanticipated impact of CASA has been in the parallel areas of reproductive toxicology and epidemiological studies of semen from men exposed to putative occupational or environmental hazards. Such use is a logical extension of early goals, unanticipated only because general concern about impact of diverse chemicals on male reproductive function did not emerge until approximately 1980, when the Environmental Protection Agency convened an important meeting. Attendees discussed the need to use objective measures of sperm motion, and the proceedings (Christian et al, 1983) cite one early CASA paper. Early on, the CellSoft system was validated for use with rat sperm (Working and Hurtt, 1987). Toth et al (1989) discussed operational parameters and which endpoints might be meaningful when applying CASA to evaluate effects of epichlorohydrin on rats. Their detailed considerations of alternative statistical approaches and minimal detectable changes remain pertinent. Refinements of software for the IVOS system (eg, Slott et al, 1993; Cancel et al, 2000) have enhanced utility with rat sperm and inclusion of strobed fluorescent capability allows detection of sperm labeled with Hoechst 33342 from detritus or other particles in semen (eg, granules in rabbit semen) or an extender. Early epidemiological studies incorporating CASA were retrospective and usually lacked robust dosimetry data. Nevertheless, they provided evidence that CASA could be useful in reproductive toxicology of humans, and a consensus approach was published (Schrader et al, 1992). Longitudinal studies now have been reported (eg, Schrader et al, 1991). Use of CASA in studies of human epidemiology will increase. Some will consider the most obvious unmet expectation to be placement of CASA systems in ≤2% of laboratories evaluating human semen, and 20 output values have clinical utility. Certainly unbiased measures of percentage of progressively motile sperm (cells with a velocity greater than some threshold value, and moving forward) and accurate determination of number of sperm/mL of semen have clinical importance. Average path velocity or percentage of total motile sperm also might be clinically important (eg, Barratt et al, 19
Referência(s)