Computerized semen analysis: objective measurement of semen characteristics is biased by subjective parameter setting

Transcription

1 FERTILITY AND STERILITY Coyright c Vol No 48, 1, July The American Fertility Society Printed in USA Comuterized semen analysis: objective measurement of semen characteristics is biased by subjective arameter setting Ulrich A Knuth, Dr med Ching-Hei Yeung, PhD Eberhard Nieschlag, Professor Dr med* Max Planck Clinical Research Unit for Reroductive Medicine and Institute of Reroductive Medicine, University of Munster, Munster, Federal Reublic of Germany To test the influence of arameter settings on results of automated semen analysis, microscoic images of 4 semen samles were videotaed Taes were analyzed by the CellSoft system (CRYO Resources Ltd, New York, NY) at different arameter combinations for, number of frames analyzed, and minimum number of frames analyzed for velocity measurements The data indicate that values obtained for concentration, ercentage of motile serm, velocity, and linearity deend, to varying degrees, on the settings of the comuter It is concluded that results of automated semen analysis may not be comarable among different laboratories unless identical arameter settings are used The need for general agreement of these details among different users is stressed Fertil Steril 48:118, 1987 Assessment of male fertility is based mainly on subjective evaluation by a technician Estimation of serm motility is esecially rone to human error and detailed analysis of additional motion characteristics such as velocity, linearity, and lateral head dislacement are not ossible by microscoic observation alone To overcome these aarent insufficiencies, research has focused on the establishment of objective methods for semen analysis1-9 High cost and time requirements usually limited alication of these methods to secial research laboratories Only recently have commercial systems based on ersonal comuters at affordable rices become available, which allow motion analysis of semen samles to be incororated into routine clinical andrology Received December 19, 198; revised and acceted March 11, 1987 * Rerint requests: Professor Dr med Eberhard Nieschlag, Max Planck Clinical Research Unit for Reroductive Medicine, University of Munster, Steinfurter Str 17, D-4 Munster, FR Germany The first ublished studies described an objective and reroducible automatic measurement of semen characteristics using such comuterized systems111 However, to a great extent, results seem to deend on arameter settings, which are comletely at the discretion of the user The resent study analyzes in detail the influence of arameter settings on the erformance of the commercially available CellSoft system (CRYO Resources, Ltd, New York, NY) MATERIALS AND METHODS Semen samles from 4 atients were used in the study The median of serm concentration according to the WHO rocedure was 2 million/ml, with a lower quartile of 19 and an uer quartile of 5 million/mi Values for motility were 8% (quartiles, % and 58%, resectively) The samles were obtained by masturbation at the infertility clinic of the Institute of Reroductive Medicine after 48 hours to 7 days of abstinence Semen samles were allowed to liquefy in an incubator at 7 C 118 Knuth et al Comuterized semen analysis

2 before microscoic images were recorded by a Panasonic VCR NV -85 (Matsushita Electric Industrial Co, Ltd, Osaka, Jaan) for automatic image analysis at a later time using a Makler counting chamber (Sefi-Medical Instruments, Haifa, Israel) Video recordings (VHS, 25 frames/sec) were made by a high-resolution camera mounted on an Olymus hase contrast BHS-microscoe BH2 (Olymus Otical Co, Ltd, Tokyo, Jaan) with a hotograhic eyeiece of 7X magnification and a lox S-lan-PL objective Taes were evaluated by a comuterized semen analysis system (IBM, Armonk, NY) consisting of an IBM AT ersonal comuter and two high-resolution black and white video monitors Measurements were erformed around the grid of the Makler chamber in a clockwise fashion starting at the uer left corner An average of cells were measured for each analysis To evaluate the influence of different adjustments on the outcome of the automated semen analysis, the following arameters were varied Gray Scale The comuterized analysis of the CellSoft system is erformed using a digitized, gray-level image Particles that do not aear luminous through hase contrast microscoy (ie, light against a dark background) will not be "seen" by the comuter Adjustment of this setting is a subjective rocedure and is done by finding an otimal setting that eliminates most debris without dimming serm images Samling Frequency and Number of Frames All comuterized measurements were erformed at a samling frequency of 5 video frames er second This reresents every other frame of the Euroean video standard of 25 frames er second Accordingly, serm were tracked for 9 and 48 seconds Minimum Requirements for Motility and Measurements To evaluate whether a cell is motile, a change in osition on consequent video frames is used By choosing a certain limit of frames, the comuter can be instructed to consider only those motile cells that change their osition on the secified number of frames For examle, if the minimum samling rate for motile cells is set at five, only cells with changing ositions on five or more frames will be counted as motile During all measurements, this setting was ket at one frame A similar adjustment is required for velocity and linearity In this case, velocity of a cell is only comuted and ut into the final analysis if it had been tracked for the secified number of frames All samles were measured at a minimum setting for velocity measurement of three and six frames A threshold velocity of 1 flm/sec was used to set the lower bound of cellular velocity This arameter defines the minimum velocity at which a serm cell is considered motile Any cell moving slower than the threshold velocity will be considered nonmotile This limit hels to exclude nonsecific movement of debris due to collision with motile cells or flow effects Each taed semen samle was analyzed at three different settings for Within different settings, serm were tracked for and frames Within both frame settings, the influence of minimal frame requirement for velocity measurement was tested for and frames Since these individual factors were arranged in a arametric design, each samle was measured at different combinations of the three arameters involved Table 1 indicates individual combinations of arameter settings for a series of measurements Parameters of Serm Motility Measured The system records the number of motile and nonmotile cells analyzed and comutes their concentrations Percentage of motile serm is given as motility In addition to these classical semen arameters, serm motion characteristics are also comuted, in micrometers er second, is calculated by adding the distances between every two Table 1 Individual Combinations of Parameter Settings: Series of Measurements Series No of Minimum frames for velocity Gray level Otimal Otimal Otimal Otimal Vol 48, No 1, July 1987 Knuth et al Comuterized semen analysis 119

3 successive video frames analyzed on the cell's track and dividing that sum by the total time interval for which the cell was tracked Linearity is a measure of the deviation of the actual serm cell track from the straight line con necting the oint of origin and end oint of the cell's track Linearity is exressed as a ratio and is calculated by dividing the length of the straight line distance by the actual track distance The more the actual track deviates from the straight line, the lower the linearity value In the resent study, linearity is given as ercentage Statistics Storage of data and statistical analysis was han dled by the same IBM AT ersonal comuter that was used for the automatic analysis of semen sam les Generated data of the Cell Soft system were stored online using the database management sys tem segment of an integrated software rogram (Enable, DBMS-Module, The Software Grou, Ballston Lake, NY) Raw data were transferred via 1 Motilit y % ASCII files into aroriate statistical rograms for further analysis To detect overall differences between individual settings, analysis of variance was erformed followed by a Student-Newman Keul test with a confidence limit of 95% Comari son of two-samle data sets were comared by t test for related samles Grahical dislays in Figures 1 to 4 use box-and whiskers lots modified by a notch indicating the 95% confidence limit of the median1 2 These lots allow quick determination of the range of data, the median, and the skew toward the uer end The central box covers the middle 5% of the data values, between the lower and uer quartile The "whiskers" extend out to the extremes (minimum and maximum values), while the central line is the median When unusual values occur at great vari ance from the bulk of the data, they are lotted as searate oints The whiskers extend only to those oints that are within 15 times the interquartile range For these comutations, the Statgrahics rogram was used (STSC, Rockville, MD) arameter setting 5 1 % arameter setting Linearit y rrrr rrrrrrrrrrrrrrrrrr J arameter setting Figure 1 Influence of different arameter CellSoft system Reetitive measurements of "," "analyzed number of frames," Details of individual arameter settings are Statistics 1 Knuth et al Comuterized semen analysis ara meter setting settings on selected semen values measured by the 4 semen samles at different arameter settings for and "velocity minimum" using the CellSoft system given in Table 1 For exlanation of box lots, see

4 Motility ( 1 /ml) L"""'I"'""T""-r',,""T"""T"""--I"""'T""",,, otimal + - otimal ( 1 o - m/sec) + Linearity rt,, rt 5 1 O LLa & a - 1 _ L _ :I - o timal + - otimal + 1 Figure 2 lnflue ce o adjustment on selected semen values measured by the CellSoft system Resul s f mdi du al meas rements of Figure 1 collased for "" Different shadings of box lots mdicate Significant differences at the level indicated for individual arameters For exlanation of box lots, see Statistics RESULTS The influence of general arameter settings is shown overall in Figure 1 To allow a more detailed analysis of the three different arameters tested, results of the 51 different measurements were re groued according to single factors Gray Scale Figure 2 shows the influence of the adjustment on the concentration of serm from identical recordings A reduction of units from he setting, considered subjectively to be otimal, mcreases the concentration of serm 7 times based on the geometric means A shift of similar magnitude to the "lighter" end of the affected measured values by a factor of 1 Percentage of motile serm deended less on a roerly adjusted system A lighter icture did not effect the measured values significantly when comared with the otimal setting A shift to a darker icture increased the estimated ercentage of motile serm by 45% Vol 48, No 1, July 1987 Mean velocity of serm differed by + 1% when a darker icture was analyzed, and by +78% for the other extreme Estimated linearity of motile serm was lower at both ends of the evaluated (-14% and -1%, resectively) Number of Analyzed Frames Values measured for concentration and motility were indeendent of the number of frames used for analysis (Fig ) However, the mean velocity mea sured was slightly higher (79%, P = 4), at a setting of comared with the samling rate of frames Measured linearity showed an oosite ef fect with a lower reading ( - % ) at a higher samling rate Minimum Number of Frames for Measurement and ercentage of motile serm (Fig 4) were estimated significantly higher when the minimum number of frames for the motility Knuth et al Comuterized semen analysis 1

5 ( 1 /ml) 1 = ( %) number o f frames analyzed n u m ber o f fra mes analyzed Linearity ( 1o- m/sec) Motility rt ' nu mber o f frames analyzed number of fra mes analyzed ] Figure Influence of frame number on selected semen values measured by the CellSoft system Results of individual measurements of Figure 1 collased for "number of frames analyzed" Identical shadings of box lots indicate absence of significant differences at the level indicated for individual arameters For exlanation of box lots, see Statistics criteria increased from three to six (+24% and + 158%, resectively) readings decreased at the higher setting by - 175%, when comared with measurements at a setting of three Measured linearity was similarly lowered by -24% DISCUS SION The data resented demonstrate that results of automated semen analysis deend, to a great ex tent, on the arameter setting, which is decided on subjectively The overall dislay in Figure 1 under lines the influence of reset values on final results Adjustment of the is of crucial imor tance since even small variations of this arameter, articularly to the darker side, seem to influence measurement of concentration and motility consid erably Variations in velocity and linearity readings are more robust with biologically negligible varia tions around 1% at the most extreme settings As exected from theoretical considerations, concentration and motility were not influenced by 2 Knuth et al Comuterized semen analysis the number of frames analyzed Problems with debris are more imortant than selected samling rate Too short a samling rate, however, leads to an erroneously high linearity reading From basic geometry it is evident that the extreme low, a sam ling rate of two frames, would lead to a linearity of 1%, which clearly cannot reresent the motion attern of a single serm cell The more oints that are samled, the better the real track will be char acterized, even though analysis is slowed down In this resect, samling frequency becomes an im ortant factor In contrast to the setting of 5 Hz used, newer versions of the CSN system allow samling frequencies of 25 Hz or even 5 Hz ( Hz and Hz, resectively, in areas using the NTSC video standard, eg, the United States and Can ada) With the original version of the system, we were not able to test the influence of this arameter variation in the same arametric design Theoreti cally, a higher samling frequency will imrove the accuracy of recorded motion characteristics In combination with aroriate duration of analysis, single samling oints will reflect the actual ath

6 1 1 I Lo I &J Q L& I LL &L LOJL&& minimal nu mber of frames I""T'"'Ir-T"'I-r-r'"'T"T, ( 1o- m/ sec), I 5 1 Linearity minimal number of frames minimal number of frames,,, + + minimal number of frames Figure 4 Influence of minimal samling requirement on selected semen values measured by the CellSoft system Results of individual measurements of Figure 1 collased for "velocity minimum" Different shadings of box lots indicate significant differences at the level indicated for individual arameters For exlanation of box lots, see Statistics of the serm even more closely This, however, will lead to some decrease in linearity and higher values for velocity The CellSoft system allows secification of the minimal number of consequent ositions of a serm cell required for velocity comutation If a serm cell is tracked for a lower number of video frames, it is considered to be motile but its velocity will not be included in the comutation of average velocity and linearity When we increased this set ting from three to six frames, a significantly lower velocity was recorded Serms with high velocity seem to have a greater chance of leaving the area of analysis before the inclusion criteria of a high num ber of frames is met This would bias the results toward the slower-moving serm remaining in the field longer Conversely, too low a requirement may not give a reresentative value for the velocity and linearity over a certain eriod This roblem be comes articularly imortant when hyeractivated serm with much higher velocity values are mea sured This is the case with, for examle, serm washed and reared for in vitro fertilization All settings seem to reresent a certain comromise Vol 48, No 1, July 1987 between samling frequency, velocity evaluation, and linearity measurement Amlitude of lateral head dislacement and beat/cross frequency, available now with the Cell Soft system, were not installed with the early ro gram version used in this study and could not be evaluated The results of our exeriment demonstrate that data roduced by the CellSoft or similar systems at different laboratories may not be comarable Al though the system is suosed to allow objective evaluation of semen samles, results deend to a great degree on arameter settings, which so far have not been standardized Thus, results for mo tion characteristics should not be looked uon as absolute values but must be interreted in the light of arameter settings and samles analyzed Mea surements of serm velocity after washing and in cubation may require different settings than semen analysis from unreared samles to gain results aroaching "reality" as close as ossible How ever, one should not be too critical on urely sta tistical grounds and should kee biologic reality in mind Knuth et al Comuterized semen analysis

7 Based on our ast exerience, we have started to analyze routine samles for 1 second at a frequency of 25 Hz using a minimum requirement for velocity comutation of frames (this covers the same real time interval as frames at 5 Hz) To avoid erroneous inclusion of nonmotile debris into motility, the estimate for the motility limit recording was increased to frames Within the limits reviously outlined, the Cell Soft system has initiated a new era of routine semen analysis The great advantage of routine velocity measurements, however, will only be reaed if an agreement on standard arameter settings can be reached to allow comarison of results among laboratories For standard semen evaluation, this goal has been achieved by the WHO laboratory manual for the examination of human semen and semen/cervical mucus interactionp A similar agreement will be necessary for automated semen analysis More detailed comarisons between high seed videomicrograhy and results from comuterized semen analysis achieved with different settings must hel to base this agreement on facts Once an agreement has been reached, it might even be feasible to send taed images of semen samles at regular intervals to articiating laboratories for external quality control REFERENCES 1 Jouannet P, Volochine B, Deguent P, David G: Study of human sermatozoa motility arameters by scattered light serm action Prog Rerod Biol 1:28, Makler A: A new multile exosure hotograhy method for objective human sermatozoal motility determination Fertil Steril :192, 1978 Overstreet JW, Katz DF, Hanson FW, Fonseca JR: A simle inexensive method for objective assessment of human serm movement characteristics Fertil Steril 1:12, Levin RM, Greenberg SH, Wein AJ: Clinical use of the turbidimetric analysis of serm motility: comarison with visual techniques Fertil Steril 5:2, Katz DF, Overstreet JW: Serm motility assessment by videomicrograhy Fertil Steril 5:188, 1981 Aitken RJ, Best FSM, Richardson DW, Djahanbakhch, Mortimer D, Temleton AA, Lees MM: An analysis of serm function iri cases of exlained infertility: conventional criteria, movement characteristics and fertilizing caacity Fertil Steril 8:2, Makler A, Maclusky NJ, Chodos A, Haseltine F, Decherney A: Raid microcomuter-based analysis of semen characteristics from hotograhs taken by the me method Arch Androl :91, Holt WV, Moore HDM, Hillier SG: Comuter-assisted measurement of serm swimming seed in human semen: correlation of results with in vitro fertilization Fertil Steril 44:1, Earnshaw JC, Munroe G, Thomson W, Traub AI: Automated laser light scattering system for assessment of serm motility Med Biol Eng Comut 2:2, Mathur S, Carlton M, Ziegler J, Rust PF, Williamson HO: A comuterized serm motion analysis Fertil Steril 4:484, Tash JS, Hidaka H, Means AR: Axokinin hoshorylation by camp-deendent rotein kinase is sufficient for activation of serm flagellar motility J Cell Biol 1:49, 198 McGill R, Tukey JW, Larsen WA: Variation of box lots Am Statistician 2:, Belsey MA, Eliasson R, Gallegos AJ, Moghissi KS, Paulson CA, Prasad MRN: Laboratory Manual for the Examination of Human Semen and Semen-Cervical Mucus Interaction Singaore, Press Concern, 19 4 Knuth et al Comuterized semen analysis