Experimental evaluation of deep learning method in reticulocyte enumeration in peripheral blood.

INTRODUCTION: Reticulocytes (RET) are immature red blood cells, and RET enumeration in peripheral blood has important clinical value in diagnosis, treatment efficacy observation, and prognosis of anemic diseases. For RET enumeration, flow cytometric reference method has shown to be more precise than the manual method by light microscopy. However, flow cytometric method generates occasionally spurious RET counts in some situations. The manual method, which is subjective, imprecise, and tedious, currently remains as an accepted reference method. As a result, there is a need for manual method to be more objective, precise, and rapid. METHODS: 40 EDTA-K2 anticoagulated whole blood samples were randomly selected for the study. 784 microscopic images were taken from blood slides as dataset, and all mature RBCs and RETs in these images were located and labeled by experienced experts. Then, we leverage a Faster R-CNN deep neural network to train a RET detection model and evaluate the model. RESULTS: Both the recall and precision rate of the model are more than 97%, and average analysis time of a single image is 0.21 seconds. CONCLUSION: The deep learning method shows outstanding performance including high accuracy and fast speed. The experimental results show that the deep learning method holds the potential to act as a rapid computer-aid method for manual RET enumeration for cytological examiners.

Use of Sysmex XN-10 red blood cell parameters for screening of hereditary red blood cell diseases and iron deficiency anaemia.

INTRODUCTION: In daily practice in haematology laboratories, red blood cell (RBC) abnormalities are frequent and their management is a real challenge. The aim of this study is to establish a "decision tree" using RBC and reticulocyte parameters from the SYSMEX XN-10 analyser to distinguish between patients with a hereditary RBC disease from iron deficiency anaemia and other patients. METHODS: We analysed results of complete RBC counts in a cohort composed of 8217 adults divided into 5 different groups: iron deficiency anaemia (n = 120), heterozygous haemoglobinopathy (n = 92), sickle cell disease syndrome (n = 56), hereditary spherocytosis (n = 18) and other patients (n = 7931). A Classification And Regression Tree (CART) analysis was used to obtain a two-step decision tree in order to predict these previous groups. RESULTS: Five parameters and the calculated RBC score were selected by the CART method: mean corpuscular haemoglobin concentration, percentage of microcytes, distribution width of the RBC histogram, percentage of nucleated red blood cells, immature reticulocytes fraction and finally RBC Score. When applying the tree and recommended flowchart, 158/166 of the RBC hereditary disease patients and 114/120 iron deficiency anaemia patients are detected. Overall, the correct classification rate reached 99.4%. Sensitivity and specificity for RBC disease detection were 95.2% and 99.9%, respectively. These results were confirmed in an independent validation cohort. CONCLUSION: Based on the XN-10 RBC and reticulocyte parameters, we propose a two-step decision tree delivering a good prediction and classification of hereditary RBC diseases. These results can be used to optimize additional reticulocyte analysis and microscopy review.

Interferents of Automated Reticulocyte Analysis Integrated with Relevant Clinical Cases.

BACKGROUND: Reticulocyte count (RET) has been used for many years to estimate the erythropoietic activity of the bone marrow. Fully automated methods not only provide enhanced precision and accuracy, but also enable reliable measurements of mRNA content and cellular indices. However, problems still exist, such as interference. The aim of the present study was to investigate the interferents of Sysmex XN 9000 reticulocyte analysis and ensure the accuracy of the results. METHODS: We collected a total of 510 specimens from normal control patients and patients with various diseases including anemias, leukemias, infectious diseases, immune diseases, kidney disease, etc. Correlation of the agreement for reticulocytes between the new methylene blue (NMB) visual microscopy method and automated reticulocyte counting was evaluated by paired sample method according to the CLSI-ICSH document H44-A2-Methods for Reticulocyte Count. Blood smear microscopic examination was carried out on the disturbed samples, and the interferents were analyzed with the medical history, flagging algorithms, the warning information, and the microscopic examination. RESULTS: A total of 44 (8.6%) cases exhibited interference. The main interferents of spuriously high reticulocyte count were caused by parasites, such as malaria, as well as suspicious autofluorescence due to drugs, while the main interferents of spuriously low reticulocyte count were caused by RBC fragments. CONCLUSIONS: Detection of potential interferences may be accomplished through alarm information and flagging algorithms incorporated into the instrument and by examination of a blood film to ensure absence of relevant interferences.

A comparison of manual counting of rabbit reticulocytes with ADVIA 2120i analyzer counting.

We compared manual counting of reticulocytes in rabbits with automatic counting using an ADVIA 2120i analyzer. Reproducibility and the influence of different anticoagulants (EDTA and Li-heparin) were also examined. Blood samples of 331 rabbits (method comparison, n = 289; reproducibility, n = 33; comparison of anticoagulants, n = 9) were tested. The reticulocyte numbers of each specimen were manually determined twice for method comparison. Passing-Bablok regressions, Bland-Altman plots, and the coefficient of variation (CV) were used to evaluate statistical significance. Good correlation (rs = 0.81) was observed between manual reticulocyte counting (groups 1-4) and the ADVIA 2120i. Quantification with the ADVIA 2120i was reproducible for relative reticulocyte numbers (EDTA, CV = 4.24%; Li-heparin, CV = 3.63%) and absolute reticulocyte numbers (EDTA, CV = 5.64%; Li-heparin, CV = 3.81%). The absolute and relative reticulocyte numbers were significantly higher in Li-heparin samples than in EDTA samples (absolute, p = 0.009; relative, p = 0.016). The ADVIA 2120i is suitable for counting reticulocytes in rabbit blood samples, but reticulocyte numbers are higher by manual counting than by ADVIA 2120i counting. Therefore, microscopic confirmation of quantifications is recommended when high numbers of reticulocytes are observed. The anticoagulant of choice is EDTA.

Diagnostic performances of manual and automated reticulocyte parameters in anaemic cats.

Objectives The objective of this study was to evaluate the diagnostic performances of manual and instrumental measurement of reticulocyte percentage (Ret%), reticulocyte number (Ret#) and reticulocyte production index (RPI) to differentiate regenerative anaemia (RA) from non-regenerative anaemia (NRA) in cats. Methods Data from 106 blood samples from anaemic cats with manual counts (n = 74; 68 NRA, six RA) or instrumental counts of reticulocytes (n = 32; 25 NRA, seven RA) collected between 1995 and 2013 were retrospectively analysed. Sensitivity, specificity and positive likelihood ratio (LR+) were calculated using either cut-offs reported in the literature or cut-offs determined from receiver operating characteristic (ROC) curves. Results All the reticulocyte parameters were significantly higher in cats with RA than in cats with NRA. All the ROC curves were significantly different ( P <0.001) from the line of no discrimination, without significant differences between the three parameters. Using the cut-offs published in literature, the Ret% (cut-off: 0.5%) was sensitive (100%) but not specific (<75%), the RPI (cut-off: 1.0) was specific (>92%) but not sensitive (<15%), and the Ret# (cut-off: 50 × 10³/µl) had a sensitivity and specificity >80% and the highest LR+ (manual count: 14; instrumental count: 6). For all the parameters, sensitivity and specificity approached 100% using the cut-offs determined by the ROC curves. These cut-offs were higher than those reported in the literature for Ret% (manual: 1.70%; instrumental: 3.06%), lower for RPI (manual: 0.39; instrumental: 0.59) and variably different, depending on the method (manual: 41 × 10³/µl; instrumental: 57 × 10³/µl), for Ret#. Using these cut-offs, the RPI had the highest LR+ (manual: 22.7; instrumental: 12.5). Conclusions and relevance This study indicated that all the reticulocyte parameters may confirm regeneration when the pretest probability is high, while when this probability is moderate, RA should be identified using the RPI providing that cut-offs <1.0 are used.

Reticulocyte count and extended reticulocyte parameters by Mindray BC-6800: Reference intervals and comparison with Sysmex XE-5000.

INTRODUCTION: In this study, analytic performance (imprecision, carryover, time stability) and diagnostic efficiency of Mindray BC-6800 analyzer to quantify reticulocytes and extended reticulocyte parameters was evaluated. Moreover, reference intervals on adult population were determined. Results were compared with those obtained by Sysmex XE-5000 analyzer. METHODS: One hundred and eighty-four healthy adults of both sexes, and 368 subjects affected by various pathologic conditions (nutritional anemias before and after treatment, hemolytic and posthemorragic anemias, acute and chronic inflammations, malignancy under therapy, and beta thalassemia trait) were selected. RESULTS: Reference intervals were as follows: reticulocytes (×109 /L): 23.2-93.2; immature reticulocyte fraction: 0.015-0.14; mean reticulocyte hemoglobin equivalent (RHE) (pg): 30.9-35.7; mean reticulocyte volume (fL): 97.8-118. Imprecision on reticulocyte count at all concentrations was close to analytic goal based on within-subject biological variation. Carryover (2.3%) was negligible, and time-stability was excellent up to 8 hours. CONCLUSION: When compared with XE-5000, BC-6800 shows a good overall correlation on counting despite evidence of difference in the upper limit of reference intervals (93.2 vs 101.3). Comparison of diagnostic efficiency of extended parameters shows a good total agreement of RHE (91.6%).

Spurious reticulocyte profiles in a dog with babesiosis.

A 9-year-old, female Maltese dog was referred to the Veterinary School of Toulouse with a 2-day history of anorexia and weakness. On clinical examination, the dog had hyperthermia (39.7°C), abdominal discomfort, and polypnea. Significant laboratory findings included pigmenturia, hyperbilirubinemia, hypercreatininemia, hyperfibrinogenemia, abnormal Snap canine pancreas-specific lipase, and pancytopenia with a nonregenerative anemia. A peripheral blood smear revealed numerous intraerythrocytic large Babesia but no polychromasia. There was a discrepancy between the absolute automated reticulocyte count (Sysmex reticulocyte count: 60 × 109 /L; RI 19.4-150.1 × 109 /L) and the manual reticulocyte count (3.6 × 109 /L) as well as the absence of polychromasia. The optical red blood cell scattergram showed an abnormal isolated reticulocyte cluster at the location of low-fluorescence ratio cells. These findings were interpreted as erythrocytes parasitized by large Babesia. The discrepancy between the Sysmex reticulocyte count and the manual reticulocyte count has been reported previously in people with falciparum malaria and numerous intra-erythrocytic Plasmodium falciparum organisms. This spurious reticulocyte profile and reticulocyte count were observed with the Sysmex XT-2000iV and the ProCyte using the same fluorescent dye polymethine but not with the LaserCyte using new methylene blue which does not stain Babesia organisms on a blood smear performed for manual reticulocyte counting.

Spurious reticulocyte profiles in dogs with large form babesiosis: a retrospective study.

BACKGROUND: Erroneously high reticulocyte counts (pseudoreticulocytosis) have been reported in dogs with leukemia. Pseudoreticulocytosis and an abnormal reticulocyte profile were observed in a dog with large form babesiosis presented at our institution. OBJECTIVES: The aims of this retrospective study were to determine if dogs with babesiosis and other dogs had abnormal reticulocyte profiles, and to correlate these profiles with the primary diagnosis. METHODS: All canine CBCs obtained with the Sysmex XT-2000iV or Procyte DX were reviewed. Cases of large form babesiosis were identified and their reticulocyte dot plots were analyzed. Dogs with abnormal reticulocyte profiles but without microscopically apparent intraerythrocytic Babesia piroplasms were identified. The reticulocyte profiles and fluorescence ratios of dogs with and without babesiosis were compared. RESULTS: Twenty of 92 dogs with babesiosis had abnormal reticulocyte profiles, including 8 with a separation between the reticulocyte and mature RBC plots or a continuum of reticulocytes from the RBC plot but with a higher density of dots in the middle of the "comet tail" than in the left quarter of the dot plot. Thirteen of 6980 dogs without Babesia on the blood smear had abnormal reticulocyte profiles, including 3 with leukemia. The medium-fluorescence reticulocyte ratios tended to be higher in dogs with babesiosis and abnormal dot plots than in other dogs, whereas the high-fluorescence ratio was higher in one dog with leukemia. CONCLUSION: Abnormal reticulocyte dot plots and atypical reticulocyte fluorescence ratios may occur in dogs with babesiosis and alert clinical pathologists to consider this diagnosis.

Analytic errors in Sysmex-generated hematology results in blood from a dog with chronic lymphocytic leukemia.

A blood sample from a 14-year-old dog was submitted to the veterinary diagnostic laboratory of the University of Milan for marked leukocytosis with atypical cells. A diagnosis of chronic T-cell lymphocytic leukemia (CLL) was made based on blood smear evaluation and flow cytometric phenotyping. A CBC by Sysmex XT-2000iV revealed a moderate normocytic normochromic anemia. Red blood cells counted by optic flow cytometry (RBC-O) resulted in a higher value than using electrical impedance (RBC-I). The relative reticulocyte count based on RNA content and size was 35.3%, while the manual reticulocyte count was < 1%. The WBC count of 1,562,680 cells/µL was accompanied by a flag. Manual counts for RBC and WBC using the Bürker chamber confirmed the Sysmex impedance results. Finally the manual PCV was lower than HCT by Sysmex. While Sysmex XT can differentiate between RBC and WBC by impedance, even in the face of extreme lymphocytosis due to CLL, RBC-O can be affected by bias, resulting in falsely increased RBC and reticulocyte numbers. Overestimation of RBC-O may be due to incorrect Sysmex classification of leukemic cells or their fragments as reticulocytes. This phenomenon is known as pseudoreticulocytosis and can lead to misinterpretation of regenerative anemia. On the other side PCV can be affected by bias in CLL due to the trapping of RBC in the buffy coat, resulting in a pink hue in the separation area. As HGB concentration is not affected by flow cytometric or other cell-related artifacts it may represent the most reliable variable to assess the degree of anemia in cases of CLL.

Quantitative detection of target cells using unghosted cells (UGCs) of DxH 800 (Beckman Coulter).

BACKGROUND: In the Retic channel of DxH 800 (Beckman Coulter), the red blood cells (RBCs) resistant to hemoglobin clearing are counted as unghosted cells (UGCs). The aim of this study was to evaluate that the UGC is a surrogate marker for both the detection and counting of target cells. METHODS: In total, 1181 samples including 22 from iron deficiency anemia (IDA) patients, 95 from jaundice, 2 from sickle cell anemia, 3 from thalassemia, 1 cord blood, and 269 from normal controls were analyzed. Slides were prepared from all samples except normal controls and target cells were counted for correlation analysis of target cell counts to UGCs. RESULTS: The normal control samples showed 0.01% (0%-0.01%) UGCs, and the reference range was set at ≤0.02%. The IDA samples showed 0.015% (0.01%-0.03%) UGC count and 0.05% (0%-0.2%) target cell count. The jaundice samples showed 0.98% (0.1%-5.36%) UGC count, and 1.4% (0.1%-7.0%) target cell count. The two sickle cell anemia samples showed 0.41% and 3.74% UGC counts and 0.4% and 11.5% target cell counts. A cord blood sample showed 0.01% UGCs and 0% target cells. The three thalassemia samples showed 0.01%, 1.99%, and 7.82% UGC counts and 0%, 1.4%, and 15.5% target cell counts. The samples showing poikilocytosis other than target cells showed normal UGC count (≤0.02%). The positive predictive value of UGCs was 58.2% (124/213) and the negative predictive value was 96.8% (674/696). The UGC counts were well correlated to the manual target cell counts (r=0.944, p=0.000). CONCLUSIONS: This study demonstrates for the first time in the literature that a hematological parameter obtained automatically every time a reticulocyte counting is performed can be used to both screen for the presence of target cells and reliably quantify them.

Reticulocyte count in red-blood-cell units stored in AS-1.

BACKGROUND AND OBJECTIVES: Previous data that showed maintenance of reticulocyte percentage in whole blood stored in CPDA-1 have led to the assumption that reticulocyte maturation becomes arrested during refrigerated storage. However, reticulocyte behaviour in red-blood-cell units stored in additive solutions has not yet been studied. This study was thus aimed at determining reticulocyte count and reticulocyte subtypes in red-blood-cells units stored in AS-1. MATERIALS AND METHODS: Reticulocyte percentage and subtypes were determined by flow cytometry with thiazole orange in six red-blood-cells units stored in AS-1. RESULTS: Reticulocyte count was 26.8 ± 4.6 × 10(9) /l at week 0.5 and 8.2 ± 2.9 × 10(9) /l at week 6. Total haemolysis during storage was 0.19 ± 0.08%. High-fluorescence reticulocytes were 2.0 ± 3.2 × 10(9) /l at week 0.5 and decreased by weeks 2, 4 and 6. Low-fluorescence reticulocytes were 22.1 ± 3.1 × 10(9) /l at week 0.5 and decreased by weeks 4 and 6. CONCLUSION: A significant decrease in reticulocytes occurred during red-blood-cells units' storage in AS-1. Even if it were assumed that all of haemolysed cells during storage were reticulocytes, there are a number of them whose disappearance cannot be explained by this mechanism. Changes observed in reticulocyte subtypes suggest that they mature during storage.

[Automated hematology analysers and spurious counts Part 3. Haemoglobin, red blood cells, cell count and indices, reticulocytes]. / Anomalies et erreurs de détermination de l'hémogramme avec les automates d'hématologie cellulaire Partie 3. Hémoglobine, hématies, indices érythrocytaires, réticulocytes.

Several situations lead to abnormal haemoglobin measurement or to abnormal red blood cells (RBC) counts, including hyperlipemias, agglutinins and cryoglobulins, haemolysis, or elevated white blood cells (WBC) counts. Mean (red) cell volume may be also subject to spurious determination, because of agglutinins (mainly cold), high blood glucose level, natremia, anticoagulants in excess and at times technological considerations. Abnormality related to one measured parameter eventually leads to abnormal calculated RBC indices: mean cell haemoglobin content is certainly the most important RBC parameter to consider, maybe as important as flags generated by the haematology analysers (HA) themselves. In many circumstances, several of the measured parameters from cell blood counts (CBC) may be altered, and the discovery of a spurious change on one parameter frequently means that the validity of other parameters should be considered. Sensitive flags allow now the identification of several spurious counts, but only the most sophisticated HA have optimal flagging, and simpler ones, especially those without any WBC differential scattergram, do not share the same capacity to detect abnormal results. Reticulocytes are integrated into the CBC in many HA, and several situations may lead to abnormal counts, including abnormal gating, interference with intraerythrocytic particles, erythroblastosis or high WBC counts.

Evaluation of the automated hematology analyzer Sysmex XT-2000iV ™ compared to the ADVIA ® 2120 for its use in dogs, cats, and horses. Part II: Accuracy of leukocyte differential and reticulocyte count, impact of anticoagulant and sample aging.

The automated laser-based hematology analyzer Sysmex XT-2000iV™ provides a 5-part differential count and specific cytograms that are of great interest for large veterinary laboratories. The aim of the study was to validate the Sysmex XT-2000iV compared to the laser-based hematology analyzer ADVIA® 2120 and manual differential in dogs, cats, and horses as well as the impact of anticoagulant (heparin, ethylenediamine tetra-acetic acid [EDTA], and citrate) and storage at 22°C and 4°C. Consecutive fresh K(3)-EDTA blood samples from 216 cats, 314 dogs, and 174 horses were included. The impact of anticoagulant and sample storage was assessed in specimens obtained from an additional 9 cats, 10 dogs, and 10 horses. Agreement between both analyzers was excellent to good except for monocytes and canine reticulocytes. Spearman rank correlation coefficients (r (s)) between Sysmex XT-2000iV and manual differential were good to fair and ranged from 0.91 (cat lymphocytes) to 0.44 (cat monocytes). Hematocrit value (Hct), mean corpuscular hemoglobin (MCH), MCH concentration (MCHC; all: P < 0.001), and mean corpuscular volume (MCV; P < 0.01) were higher in canine citrated blood compared to heparin and EDTA. In cats, lymphocytes and monocytes were lower in heparinized blood compared to EDTA (P < 0.05), whereas in horses no significant effect was seen. Regarding storage time and temperature, white and red blood cell counts, hemoglobin, and MCH were stable. Hct, MCV, and MCHC were influenced by erythrocyte swelling. Differential count remained stable for 24 hr (22°C) and nearly 72 hr (4°C) except for monocytes. The overall performance of the Sysmex XT-2000iV was excellent and compared favorably with that of the ADVIA 2120. A special strength was the excellent detection of feline eosinophils.

Manual and automated reticulocyte counts.

Manual reticulocyte counts were examined under light microscopy, using the property whereby supravital stain precipitates residual ribosomal RNA versus the automated flow methods, with the suggestion that in the latter there is greater precision and an ability to determine both mature and immature reticulocyte fractions. Three hundred and forty-one venous blood samples of patients were analyzed of whom 224 newborn and the rest adults; 51 males and 66 females, with ages between 0 and 89 years, as part of the laboratory routine for hematological examinations at the Clinical Laboratory of the Hospital Universitário do Oeste do Paraná. This work aimed to compare manual and automated methodologies for reticulocyte countings and evaluate random and systematic errors. The results obtained showed that the difference between the two methods was very small, with an estimated 0·4% systematic error and 3·9% random error. Thus, it has been confirmed that both methods, when well conducted, can reflect precisely the reticulocyte counts for adequate clinical use.

Flow cytometric reticulocyte counting: a comparison between two methods.

The peripheral reticulocyte count is commonly used as an indicator of the erythropoietic activity of the bone marrow. Manual counting provides results with a high degree of inaccuracy and imprecision. Automation of counting is therefore needed. The increase in the number of methods available requires however that the results from the various methods agree with one another. The aim of our study was to evaluate the analytic performance of two automated hematology analyzers by a parallel study. We compared the analyzers between them and with manual counting. We enrolled in our study a total of 100 healthy subjects and an additional 80 patients affected by various hematological diseases. Difference between methods is statistically significant: the reference intervals of ADVIA2120 are higher than the Sysmex XE-2100. The correlation between methods and correlation with the microscopic method are excellent and statistically significant. In conclusion, we can affirm that total automation of reticulocyte counts represents a definite improvement over microscopic counts. This study confirms the diversity of the reference intervals still exists in the new automated hematology analyzers.

Under-filled blood collection tubes containing K2EDTA as anticoagulant are acceptable for automated complete blood counts, white blood cell differential, and reticulocyte count.

Current laboratory standards from Clinical Laboratory Standards Institute (CLSI) and manufacturer's (Becton Dickinson) data indicate that under-filling K(2)EDTA blood collection tubes can result in erroneous hematology values. To accommodate under-filled tubes and reduce collection volumes while optimizing our automation, we explored the acceptable limit of under-filled tubes for hematology values. We collected 8.0 ml of blood from 30 normal adult volunteers. Each donation was aliquoted in the following volumes: 4.0, 2.0, 1.0, 0.5 ml x 2. These samples were analyzed within 1 h of blood collection on Sysmex XE-2100 (Sysmex America Inc., Mundelein, IL, USA) for complete blood count, reticulocyte, and white blood cell differentials. Results of the under-filled tubes were compared to those of the standard volume. The Deming regression analysis show excellent correlation for all parameters between each under-filled blood collection volume compared to a standard 4 ml volume. The Bland and Altman analysis shows good agreement between both 1.0 and 2.0 ml compared to a 4.0 ml volume. The 0.5 ml compared to a 4.0 ml volume, however, shows increased variation on many parameters. In addition all three collection volumes show negative bias compared to the standard volume for platelet count, but the difference is considered insignificant with a percent difference of 5.5%, 3.2%, and 1.5% for 0.5, 1.0, and 2.0 ml collection volume respectively. Finally for 0.5 ml collection volume we noticed a low level of false positive flagging rate for white blood cell. Acceptable complete blood count values of under-filled powdered K(2)EDTA tubes can be obtained with as little as 1.0 ml of blood.