Analytic Performance of the iq200 Automated Urine Microscopy Analyzer and Comparison With Manual Counts Using Fuchs-Rosenthal Cell Chambers

Transcription

1 Clinical Chemistry / PERFORMANCE OF THE IQ2 URINALYSIS SYSTEM Analytic Performance of the iq2 Automated Urine Microscopy Analyzer and Comparison With Manual Counts Using Fuchs-Rosenthal Cell Chambers David T. Wah, Porntip K. Wises, and Anthony W. Butch, PhD Key Words: Automated urinalysis; Casts; Cells; Instrumentation; Manual urinalysis DOI: 1.139/VNGU9Q5V932D74NU Abstract Automated instruments that can examine urine for cells and particles have reduced the need for laborintensive manual microscopy. We evaluated the performance of the Iris iq2 Automated Urine Microscopy Analyzer (Iris Diagnostics, Chatsworth, CA) and compared results with manual cell and particle counts using Fuchs-Rosenthal counting chambers. Within-run imprecision (coefficient of variation) of the iq2 for urine samples ranged from 3.% to 45% for RBC counts between 1,29 and cells/l, 3.4% to 4% for WBC counts between 1,6 and cells/l, and 8.9% to 35% for epithelial cell counts between 93 and cells/l. Between-run imprecision was 3.3% at 1, cells/l and 19.2% at cells/l. There was good agreement between the iq2 and manual cell counts (r.94); however, the iq2 produced lower results based on slopes of.92 (RBC count),.81 (WBC count), and.94 (epithelial cell counts). The iq2 has satisfactory performance and correlates well with manual cell counts. Most urine samples containing RBCs, WBCs, and epithelial cells can be reported without review of captured images. Chemical testing of urine samples (reagent strip method) and identification and counting of particles is performed routinely to identify and monitor diseases of the kidney and urinary tract. 1 Despite the introduction of an automated system to count particles more than 2 decades ago, there is still a need for manual microscopic examination of urinary sediment owing to the limitations of existing automated counting systems. 2-5 Unfortunately, manual microscopy is extremely time consuming and has poor precision owing to variations in sediment preparation and counting technique. 6-9 To improve standardization, the European Confederation of Laboratory Medicine published guidelines for manual counting of particles in uncentrifuged and concentrated urine samples. 1,11 However, given the large volume of requests for routine urinalysis, improvements in automated systems still are needed to eliminate laborintensive manual microscopy in the clinical laboratory. The 2 systems currently available to automate manual microscopy are based on different technologies. One is an image-based analysis system that uses a video camera and strobe lamp (stops fluid motion) to detect and sort particles based on predetermined particle dimensions. The first of these systems was introduced in the mid-198s. 12 The other type of system was developed in the mid-199s and is based on the principle of flow cytometry. 13 This system classifies particles based on fluorescent intensity (cells are stained with 2 fluorescent dyes), electrical impedance, and forward angle light scatter. Both of these systems have improved precision and sample throughput compared with manual microscopy. 2,7 The major drawback of older image-based systems has been that they require continuous monitoring of images by the operator and do not have walk-away capability. Although flow cytometric systems do not have to be monitored continuously, they 29 Am J Clin Pathol 25;123: DOI: 1.139/VNGU9Q5V932D74NU

2 Clinical Chemistry / ORIGINAL ARTICLE provide only scattergram analysis, limiting the number of particles that can be differentiated in urine samples. A next-generation automated image-based urinalysis system, the iq2 (Iris Diagnostics, Chatsworth, CA), recently received US Food and Drug Administration clearance. The iq2 is a walk-away system that uses flow imaging analysis technology and Auto-Particle Recognition (APR, Iris Diagnostics) software to classify particles based on multiple parameters. Images are stored and can be viewed on the workstation screen, thereby eliminating the need for manual microscopy in most cases. In the present study, we evaluated the analytic performance of the iq2 and compared results with those from manual microscopy using standardized Fuchs-Rosenthal counting chambers. Materials and Methods Urine Specimens Specimens submitted to the outpatient laboratory at the University of California Los Angeles Medical Center during a 5-week period for routine urinalysis were selected for the study. The majority of urine samples were from the renal clinic. Approximately 1 urine samples with sufficient volume for additional testing (15-mL volume after routine testing) were identified each day as having normal or abnormal cell counts by the Yellow IRIS urinalysis workstation (Iris Diagnostics). After analyzing about 5 normal and 5 abnormal urine specimens, additional samples were selected with specific concentrations of RBCs, WBCs, or squamous epithelial cells (ECs) to cover a wide range of concentrations. An effort was made to select a representative number of urine samples with abnormal cell counts near the upper limit of normal. All samples were tested without dilution. For the study, 166 urine samples were analyzed for cells and particles by manual and automated microscopy. All urine samples were tested the same day of collection. The study was reviewed and approved by the human studies committee at the University of California Los Angeles. Iris Diagnostics iq2 The iq2 is an automated urine microscopy analyzer that uses digital imaging and a trained neural network (APR software) to classify and quantitate cells and formed particles in native, uncentrifuged urine. The iq2 requires a minimum of 3 ml of urine for testing and quantitates particles in 2 µl of sample. The instrument can analyze 6 urine samples per hour. Calibration was performed at the beginning of the study and as needed based on a 1-month calibration stability. Before testing of urine samples each day, iq Focus, iq Negative, and iq Positive control samples (Iris Diagnostics) were run according to the manufacturer s instructions. Each urine sample was divided into aliquots in 2 tubes, analyzed for cells and particles, and expressed as the average of duplicate cell and particle counts. Unless otherwise indicated, the APR software was used to classify and count cells and particles without manual editing of images. Manual Microscopy For manual cell counts 3.2-µL Fuchs-Rosenthal counting chambers (catalog No. 372, Hausser Scientific, Horsham, PA) were used and counted by bright-field microscopy using a 1 ocular. Each side of the chamber has 16 large squares, with each square being mm deep (.2 mm 3 ), for a total volume of.2 µl. The large squares are further divided into 64 smaller squares. Two technologists loaded both sides of separate counting chambers with well-mixed native urine or control cells. Particles were allowed to settle undisturbed for 5 minutes, and then both sides of the chamber were scanned using a 1 objective to verify an even distribution of particles throughout the counting area. All 16 large squares on both sides of the chamber (6.4 µl) were examined using a 1 objective to count ECs and casts. The technologists then examined 16 (of 64) evenly distributed small squares in each of the 4 large center squares on both sides of the chamber under high magnification (4 objective) to count RBCs and WBCs. This was equivalent to 1 large square being counted per side (64 squares/1,24 total squares 3.2 µl =.2 µl per side), for a total volume of.4 µl. Each technologist then counted the other person s chamber using the same procedure. For correlation with the iq2, results were expressed as the average of the 4 separate counts for each cell and particle type (2 separate chambers counted by 2 technologists). Precision Studies Within-run imprecision was evaluated by counting cells in individual urine samples a total of 2 times during a single day. The mean and coefficient of variation (CV) for each sample were calculated, and results for different urine samples with similar RBC, WBC, or EC counts were grouped together and expressed as a range within each of the groups. Each group contained results from 5 urine samples. Imprecision also was evaluated at low cell numbers by making dilutions of urine samples with initial cell counts around /L using a cell-free urine pool. Each urine sample was tested before dilution and at each dilution a total of 2 times during a single day. For between-run imprecision, glutaraldehyde-preserved human RBCs (positive control cells provided by the manufacturer) were diluted in Iris Diluent (Iris Diagnostics), and an aliquot was tested in duplicate each day for 12 days. Statistical Analysis EP Evaluator-CLIA software (David G. Rhoads Associates, Kennett Square, PA) was used for Deming regression analysis and supporting statistics. Am J Clin Pathol 25;123: DOI: 1.139/VNGU9Q5V932D74NU 291

3 Wah et al / PERFORMANCE OF THE IQ2 URINALYSIS SYSTEM Results Variability in Manual Cell Counts To determine imprecision in manual cell counts, 2 technologists loaded both sides of a Fuchs-Rosenthal counting chamber with glutaraldehyde-preserved human RBCs and counted the equivalent of 1 large square on each side of the chamber (.4 µl total volume). The technologists then performed a cell count on the other person s chamber. This resulted in 2 counts for each of 2 counting chambers. This process was repeated a total of 5 times for each RBC concentration. Imprecision (expressed as CV %) was less than 5.9%, less than 8.2%, and less than 22.8% at mean cell counts of 983, 255, and /L, respectively. Variability in manual urine cell counts between technologists was examined by Deming regression analysis, and the statistics are shown in Table 1. The slopes of the regression lines were not different from 1. for RBC and WBC counts. A slight bias in EC manual counts was observed based on a slope of.9. There was good correlation among counts for all 3 cell types based on r values of more than.97. RBC counts had the most dispersion around the regression line (S y/x, 4.1); this was most likely due to more samples containing higher numbers of RBCs compared with the other cell types. Precision of the iq2 Within-run imprecision of the iq2 was evaluated by analyzing urine samples a total of 2 times during the same day for RBCs, WBCs, and ECs. As shown in Table 2, imprecision for RBC counts was 5.6% (CV) or less at concentrations ranging from 786 to 1, /L and increased with decreasing cell numbers up to 29.6% for counts between 17 and /L. Within-run CVs were slightly better for WBC counts with values of 3.4% or less and 22.3% or less at concentrations ranging from 794 to 1,6 1 6 /L and 16 to /L, respectively. ECs had the highest CV of 31.3% in the concentration range of 13 to /L. Between-run CVs for glutaraldehyde-preserved RBCs was lowest (3.3%) at a concentration of 1, /L and highest (19.2%) at the lowest cell concentration tested of /L Table 3. Imprecision was examined further in urine samples at low cell concentrations. Overall, CVs increased with decreasing cell numbers and approached 45% for RBCs, 4% for WBCs, and 35% for ECs at the lowest cell concentrations Figure 1. CVs of approximately 2% were found at RBC, WBC, and EC concentrations of 25, 18, and /L, respectively (Figure 1). Linearity and Carryover To determine linearity, serial dilutions of urine samples with high concentrations of each cell type were made using a cell-free urine pool, and each dilution was tested in triplicate by the iq2. RBC, WBC, and EC counts were linear up to Table 1 Correlation Between Manual Counts * Cell No. of Type Counts Slope Intercept r S y/x RBC WBC EC EC, epithelial cell. * Two technologists loaded separate Fuchs-Rosenthal counting chambers with the same urine sample, performed a cell count, and then counted the other person s chamber. The number of cell counts performed. The number of urine samples is half this value because urine samples were counted twice by each technologist. Table 2 Within-Run Imprecision of the iq2 Automated Urine Microscopy Analyzer in Counting Various Cell Types * Cell Type Mean Cell Range ( 1 6 /L) CV Range (%) RBC 786-1, WBC 794-1, EC CV, coefficient of variation; EC, epithelial cell. * Each cell range contains 5 urine samples that were analyzed 2 times during a single day. For proprietary information, see the text. Table 3 Between-Run Imprecision of the iq2 Automated Urine Microscopy Analyzer * Mean Cell Count ( 1 6 /L) SD CV (%) 1, CV, coefficient of variation. * Glutaraldehyde-preserved RBCs were analyzed in duplicate for a total of 12 days. For proprietary information, see the text. 1,, 9, and 1, respectively, based on slopes of 1. ±.1 and measured concentrations ± 1% of expected target values. Unfortunately, specimens with high numbers of ECs were not available during the study to validate linearity beyond 1 ( 1 6 /L) and confirm the manufacturer s linearity claim of 1,. Carryover was determined by counting the negative control sample immediately after the positive control sample (fixed RBCs with a count of 1, 1 6 /L) each day for 9 days and was found to be.2% or less. 292 Am J Clin Pathol 25;123: DOI: 1.139/VNGU9Q5V932D74NU

4 Clinical Chemistry / ORIGINAL ARTICLE A 5 B CV (%) 3 2 CV (%) Mean RBC ( 1 6 /L) Mean WBC ( 1 6 /L) C CV (%) Figure 1 Within-run imprecision of the iq2 at low cell concentrations. Urine samples initially containing approximately /L RBCs (A), WBCs (B), or ECs (C) were diluted with a cell-free urine pool and analyzed 2 times during the same day. CV, coefficient of variation; EC, epithelial cell. For proprietary information, see the text Mean EC ( 1 6 /L) Correlation Studies Correlation data for 166 urine samples with cell counts of /L or more by both manual microscopy and the iq2 are shown in Figure 2. Overall, there was good agreement between the iq2 and manual Fuchs-Rosenthal chamber counts for cellular elements based on r values of.94 to.959. The iq2 exhibited a mean negative bias of 11.3% for RBCs, 24.7% for WBCs, and 4.3% for ECs compared with manual counts. Of the 166 urine samples tested, only 2 (12%) had cell counts outside the 95% confidence interval of the regression lines (open circles, Figure 2). iq2-stored images for these samples were reviewed manually to determine possible causes for the discrepancies. Seven samples had discrepant RBC counts, with 4 lower and 3 higher by the iq2 Table 4. Review of the images revealed that RBC discrepancies were due to abnormal RBC morphologic features (ghost and dysmorphic cells), trapping of RBCs in mucus strands, oval calcium oxalate crystals, and yeast. Ghost RBCs, mucus, and calcium oxalate crystals resulted in low or high iq2 RBC counts, whereas dysmorphic RBCs and yeast produced low and high RBC counts, respectively. For WBCs, there were 8 discrepancies. Five were low and 3 were high by the iq2. The low WBC counts were due to disintegrating and atypical WBCs that were classified as artifact by the iq2. All high WBC counts were due to WBC clumps. For ECs, there were 5 discrepant counts, with EC clumps accounting for 4. The remaining EC discrepancy was due to unusually wide hyaline casts being classified as ECs by the iq2. Clinical Performance of the iq2 The ability of the iq2 to correctly classify urine samples as having normal or abnormal RBC, WBC, and EC counts was examined using manual Fuchs-Rosenthal chamber counts as the comparison method. As shown in Table 5, the agreement rate for the iq2 ranged from 92.5% to 1% for samples classified as normal for each of the cell types. The agreement rate for abnormal samples was lower and ranged from 76.7% to 86.4%. Overall, the iq2 had the highest agreement rate for ECs. Receiver operating characteristic area values for RBCs, WBCs, and ECs were.941,.931, and.97, respectively, indicating high diagnostic accuracy (data not shown). Am J Clin Pathol 25;123: DOI: 1.139/VNGU9Q5V932D74NU 293

5 Wah et al / PERFORMANCE OF THE IQ2 URINALYSIS SYSTEM A B 1, 5 iq2 RBC ( 1 6 /L) iq2 WBC ( 1 6 /L) , Manual RBC ( 1 6 /L) Manual WBC ( 1 6 /L) iq2 EC ( 1 6 /L) C Figure 2 Deming regression analysis of RBC (A), WBC (B) and EC (C) counts in native urine samples quantitated by manual chamber counts and the iq2. A total of 166 urine specimens were tested with cell counts of /L or more by both methods. A, 141 data points; y =.92x 2.94; r =.959. B, 152 data points; y =.81x 3.2; r =.94. C, 8 data points; y =.94x +.34; r =.951. Open circles represent data points outside the 95% confidence interval of the regression lines. EC, epithelial cell. For proprietary information, see the text Manual EC ( 1 6 /L) Cast Identification In 24 samples, there were between 1 and /L hyaline casts by both counting methods. The iq2 detected more hyaline casts based on a Deming regression line slope of 1.74 (data not shown). The correlation coefficient was.91. Table 4 Summary of Discordant Results * iq2 Result iq2 Image Review RBC count Low (n = 4) Ghost RBCs, 1; dysmorphic RBCs, 1; mucus, 1; calcium oxalate crystals and mucus, 1 High (n = 3) Ghost RBCs, 1; calcium oxalate crystals and mucus, 1; yeast and mucus, 1 WBC count Low (n = 5) Disintegrating WBCs, mucus, and yeast, 4; atypical WBCs, mucus, and yeast, 1 High (n = 3) WBC clumps and mucus, 3 EC count Low (n = 2) EC clumps, 2 High (n = 3) EC, epithelial cell. * For proprietary information, see the text. EC clumps, 1; EC clumps, mucus, bacteria, and crystals, 1; numerous wide hyaline casts, 1 Manual editing of stored iq2 images revealed that the APR overestimated hyaline cast counts in 2 samples with small numbers of hyaline casts ( /L). Otherwise there was good agreement between counts generated by the APR software and after manual review of stored images. Nine samples had nonhyaline casts that were placed in the unclassified pathologic cast category by the APR software. The APR software detected more pathologic casts than manual counting in 7 of 9 samples. There was good agreement between cast counts produced by the APR software and after manual review of images for all samples. In general, the quality of stored images was sufficient for subclassification of casts with the exception of cellular casts, in which the cellular detail was not always adequate for differentiation into cellspecific subcategories. Discussion The iq2 is a new, automated, digital imaging-based system for routine identification of cells and particles in native 294 Am J Clin Pathol 25;123: DOI: 1.139/VNGU9Q5V932D74NU

6 Clinical Chemistry / ORIGINAL ARTICLE Table 5 Agreement Between the iq2 and Manual Counts Based on Cutoff Values * % Agreement Cell Type Upper Limit of Normal ( 1 6 /L) No. of Samples Normal Abnormal RBC WBC EC EC, epithelial cell. * For proprietary information, see the text. Results were considered normal/abnormal based on manual cell counts. Of these samples, 8, 92, and 58 had normal RBC, WBC, and EC counts, respectively, based on manual cell counts. urine. Features of individual images such as size, shape, contrast, and texture are used by the neural network pattern recognition software for classification into 1 of 12 particle categories. The results can be reported without operator review, or stored images in each category can be displayed on the screen for verification and/or manual editing. The iq2 does not require continuous monitoring and can be connected to a reagent strip chemistry analyzer to provide a fully integrated chemical and microscopic urinalysis workstation. 14 Between-run imprecision of the iq2 using fixed RBC counts was excellent at high cell counts based on CVs less than 7.5%. At lower cell counts, between-run imprecision was good, approaching 2% at counts below /L. Withinrun precision for RBCs, WBCs, and ECs in urine samples was satisfactory at several cell concentrations, producing CVs of 25%, 15%, and 16% at the upper limits of normal for RBCs ( /L), WBCs ( /L), and ECs ( /L), respectively. Overall, CVs for quantitation of cells by the iq2 were smaller than those reported for manual cell counts. 5,12,15 Compared with other automated instruments, the iq2 displayed similar 16,17 or higher 2,4,18 CVs depending on the study design. The iq2 was linear over a wide range of cell concentrations and had negligible carryover. Although there was good correlation between methods for RBC, WBC, and EC counts (r values ranging from.94 to.959), the iq2 identified fewer cells compared with manual cell counts using Fuchs-Rosenthal counting chambers. On average, cell counts by the iq2 were 4.3% to 24.7% lower than manual counts depending on the cell type. The reasons for these differences in cell counting are unclear, but we considered that it could reflect inaccuracies associated with manual cell counting methods. 6,7,9 To address this, we minimized the imprecision of standardized manual chamber counts by averaging 4 separate counts (2 technologists counts of 2 counting chambers). There was excellent correlation for urine cell counts between technologists (Table 1). Thus, we do not believe that inaccuracies associated with the manual counting method can adequately explain the differences in cell counts between methods. It is interesting that the UF-1 automated instrument (TOA Medical Electronics, Kobe, Japan) also has been reported to produce lower cell counts than those obtained using Fuchs-Rosenthal counting chambers. 3 During this study, 2 urine samples were identified with cell counts falling outside the 95% confidence interval of the regression lines. Abnormal RBCs (some dysmorphic RBCs and ghost cells) and disintegrating WBCs accounted for many of these cases. The iq2 counted distorted and disrupted cells as artifact, whereas this judgment would be highly subjective and would vary among technologists performing manual counts. Likewise, yeast and oval calcium oxalate crystals, which accounted for 2 of the discordant counts, could be confused with RBCs by automated and manual counting methods. It is interesting that one of the discrepancies due to calcium oxalate crystals resulted in a falsely high iq2 count and the other a falsely high manual count. Cell clumping also is problematic and can result in lack of agreement between methods depending on sample mixing before analysis. The iq2 automatically resuspends particles in solution before testing by injecting a bolus of air into the sample. This might dissociate some of the cellular clumps, resulting in counts that are higher than those obtained by manual counting methods. It is noteworthy that almost all of the discrepant results occurred at very high cell concentrations that would not have influenced clinical sensitivity. In fact, we found that the sensitivity of the iq2 for detecting abnormal levels of RBCs, WBCs, and ECs was between 76% and 86%, whereas the specificity exceeded 92% for all 3 cell types. Overall, the iq2 detected more hyaline casts compared with manual cell counts. Although the reason for this is unclear, it might be related to the use of bright-field microscopy for manual cell counting. Hyaline casts are translucent, and some might have been missed by the reader, leading to falsely low cast counts. 11 Review of iq2-stored images confirmed higher hyaline cell counts by the iq2. In a handful of cases, the iq2 overestimated the number of hyaline casts in samples with 1 to 3 casts ( 1 6 /L) by manual counts. In view of this, we agree with the manufacturer s recommendation that all casts should be verified for accuracy Am J Clin Pathol 25;123: DOI: 1.139/VNGU9Q5V932D74NU 295

7 Wah et al / PERFORMANCE OF THE IQ2 URINALYSIS SYSTEM by manual review of stored images. Other types of casts also were identified, but owing to the small number of positive samples, additional studies are needed to evaluate the performance of the iq2 in identifying pathologic casts. The iq2 has satisfactory analytic performance for quantitation of RBCs, WBCs, and ECs and correlates well with Fuchs-Rosenthal manual chamber counts using uncentrifuged urine samples. Based on our findings, the majority of automated counts for RBCs, WBCs, and ECs can be reported without manual review of captured images. Exceptions include urine samples containing crystals and/or yeast that would require review of crystals and/or yeast images for confirmation, along with editing of the RBC, WBC, and EC images for accurate cell counts. This should require less than a minute per sample. In addition, WBC clumps will need to be verified by image review because several artifacts could be classified as cellular clumps. Last, based on limited data, we propose that the artifact category be reviewed for urine samples with 2+ protein or higher to detect small numbers of pathologic casts and should require no more than an additional 1 to 3 seconds. The iq2 can be connected physically and electronically to a dry-reagent strip chemistry analyzer so that urine chemical analyses such as protein appear on the monitor next to the microscopic results. From the Department of Pathology and Laboratory Medicine, David Geffen School of Medicine at the University of California Los Angeles. Supported by Iris Diagnostics, Chatsworth, CA. Address reprint requests to Dr Butch: Dept of Pathology and Laboratory Medicine, David Geffen School of Medicine at UCLA, 1833 Le Conte Ave, Mailroom A7-149 CHS, Los Angeles, CA Acknowledgment: We thank Linda Escobar for excellent clerical assistance. References 1. Fuller CE, Threatte GA, Henry JB. Basic examination of urine. In: Henry JB, ed. Clinical Diagnosis and Management by Laboratory Methods. 2th ed. Philadelphia, PA: Saunders; 21: Okada H, Sakai Y, Kawabata G, et al. Automated urinalysis: evaluation of the Sysmex UF-5. Am J Clin Pathol. 21;115: Regeniter A, Haenni V, Risch L, et al. Urine analysis performed by flow cytometry: reference range determination and comparison to morphological findings, dipstick chemistry and bacterial culture results: a multicenter study. Clin Nephrol. 21;55: Ottiger C, Huber AR. Quantitative urine particle analysis: integrative approach for the optimal combination of automation with UF-1 and microscopic review with KOVA cell chamber. Clin Chem. 23;49: Roe CE, Carlson DA, Daigneault RW, et al. Evaluation of the Yellow IRIS: an automated method for urinalysis. Am J Clin Pathol. 1986;86: Winkel P, Statland BE, Jorgensen K. Urine microscopy, an illdefined method, examined by multifactorial technique. Clin Chem. 1974;2: Elin RJ, Hosseini JM, Kestner J, et al. Comparison of automated and manual methods for urinalysis. Am J Clin Pathol. 1986;86: Gadeholt H. Quantitative estimation of urinary sediment, with special regard to sources of error. Br Med J. 1964;1: Ferris JA. Comparison and standardization of the urine microscopic examination. Lab Med. 1983;14: Kouri TT, Gant VA, Fogazzi GB, et al. Towards European urinalysis guidelines: introduction of a project under European Confederation of Laboratory Medicine. Clin Chim Acta. 2;297: European Confederation of Laboratory Medicine. European urinalysis guidelines: summary. Scand J Clin Lab Invest Suppl. 2;231: Deindorfer FH, Gangwer JR, Laird CW, et al. The Yellow IRIS urinalysis workstation: the first commercial application of automated intelligent microscopy. Clin Chem. 1985;31: Hyodo T, Kumano K, Haga M, et al. Detection of glomerular and non-glomerular red blood cells by automated urinary sediment analyzer [in Japanese]. Nippon Jinzo Gakkai Shi. 1995;37: Butch A. Preliminary experience in automated urine microscopy with the iqtm2 [abstract]. Clin Chem Lab Med. 23;41(suppl):S Carlson DE, Statland BE. Automated urinalysis. Clin Chem Lab Med. 1988;8: Ben-Ezra J, Bork L, McPherson RA. Evaluation of the Sysmex UF-1 automated urinalysis analyzer. Clin Chem. 1998;44: Kouri TT, Kahkonen U, Malminiemi K, et al. Evaluation of the Sysmex UF-1 urine flow cytometer vs chamber counting of supravitally stained specimens and conventional bacterial cultures. Am J Clin Pathol. 1999;112: Fenili D, Pirovano B. The automation of sediment urinalysis using a new urine flow cytometer (UF-1TM). Clin Chem Lab Med. 1998;36: Am J Clin Pathol 25;123: DOI: 1.139/VNGU9Q5V932D74NU