| Abstract|| |
As we approach the aftermath of a global pandemic caused by Severe Acute Respiratory Syndrome-Corona Virus (SARS-CoV-2), the importance of quickly developing rapid screening tests has become very clear from the point of view of containment and also saving lives. Here, we present an explorative study to develop a telepathology-based screening tool using peripheral blood smears (PBS) to identify Coronavirus Disease (COVID-19)-positive cases from a group of 138 patients with flu-like symptoms, consisting of 82 positive and 56 negative samples. Stained blood smear slides were imaged using an automated slide scanner (AI 100) and the images uploaded to the cloud were analyzed by a pathologist to generate semi-quantitative leukocyte morphology-related data. These telepathology data were compared with the data generated from manual microscopy of the same set of smear slides and also the same pathologist. Besides good correlation between the data from telepathology and manual microscopy, we were able to achieve a sensitivity and specificity of 0.83 and 0.71, respectively, for identifying positive and negative COVID-19 cases using a six-parameter combination associated with leukocyte morphology. The morphological features included plasmacytoid cells, neutrophil dysplastic promyelocyte, neutrophil blast-like cells, apoptotic cells, smudged neutrophil, and neutrophil-to-immature granulocyte ratio. Although Polymerase Chain Reaction (PCR) and antibody tests have a superior performance, the PBS-based telepathology tool presented here has the potential to be an interim screening tool in resource-limited settings in underdeveloped and developing countries.
Keywords: COVID-19, leukocyte morphology, peripheral blood smear, telepathology
|How to cite this article:|
Savitha K A, Prasad V, Manjunath K H, Govind E N, Manjula S J, Renu E, Lokanathan R A, Neha D, Shanthinee R, Tathagato R D. A telepathology based screening tool for COVID-19 by leveraging morphological changes related to leukocytes in peripheral blood smears. Indian J Pathol Microbiol 2022;65:886-90
|How to cite this URL:|
Savitha K A, Prasad V, Manjunath K H, Govind E N, Manjula S J, Renu E, Lokanathan R A, Neha D, Shanthinee R, Tathagato R D. A telepathology based screening tool for COVID-19 by leveraging morphological changes related to leukocytes in peripheral blood smears. Indian J Pathol Microbiol [serial online] 2022 [cited 2022 Dec 7];65:886-90. Available from: https://www.ijpmonline.org/text.asp?2022/65/4/886/346687
| Introduction|| |
Ever since the COVID-19 outbreak started in 2019, the number of deaths has crossed 2.4 million and the total infections have surpassed 104 million globally., Numerous antigen-based and antibody-based COVID-19 diagnostic tests have aided in the management of this pandemic in terms of containing the spread and also saving lives. Besides developing molecular and immuno-diagnostic tests, numerous studies have been reported that utilize changes in blood parameters, especially leukocyte- and inflammation-related characteristics to predict at an early stage of COVID-19 infection which of the infected patients are likely to develop severe symptoms.,,,, As in the case of COVID-19, there is a significant lag between the moment a novel pathogenic outbreak is detected and the time when large-scale testing kits become available to manage a pandemic. During such a lag period, an easily accessible screening tool could be very helpful in managing the spread of infections. PBS is a highly accessible diagnostic tool and PBS-based diagnostic cues associated with COVID-19 have been investigated by various research groups as a potential screening tool, with their main conclusions pointing toward a way to identify those SARS-COV-2-infected individuals who might experience severe symptoms as the infection progresses.,,,,,, For example, Zhang et al. reported the qualitative observation that COVID-19-associated PBS showed larger, atypical, vacuolated monocytes. Chong et al. reported an elevated presence of reactive lymphocytes in the case of COVID-19-positive cases. One common observation from many reports was the neutrophil-to-lymphocyte ratio, which showed the strongest correlation with the severity of COVID-19 prognosis.,,, The neutrophil Pseudo-Pelger-Huët Anomaly was one other widely reported qualitative observation related to the PBS of COVID-19-related samples.,, With the advent of whole-slide scanners connected to the cloud such as AI100, it becomes easy to add the feature of telepathology to PBS-based diagnosis., Telepathology is increasingly becoming an indispensable tool in resource-limited settings, and, especially amidst the travel restrictions during the pandemic.,, Here, we present a study involving the development of a telepathology screening tool to identify the positive and negative COVID-19 cases using PBS. As part of this study, a cross-verification is performed to compare the results of telepathology and manual microscopy. The entire study is performed using smears from 138 patients consisting of 82 COVID-19-positive and 56 negative samples, everyone with flu-like symptoms. We quantitatively analyze at least 15 parameters associated with PBS and conclude that a 6-parameter combination is able to achieve the optimal PBS-based detection of COVID-19 among patients with flu-like symptoms. The 6-parameter combination included plasmacytoid cells (PCC), neutrophil dysplastic promyelocyte (NDP), neutrophil blast-like cells (NBC), apoptotic cells (AC), smudged neutrophils (SN), and neutrophil-to-immature granulocyte ratio (neutrophil-to- Immature Granulocytes (IG) ratio). Besides immediate relevance to COVID-19, the overall study design to arrive upon a multiparameter combination based on the analysis of PBS slides can be applied to the other outbreaks in the future, provided a smear-based diagnosis is feasible.
| Materials and Methods|| |
The PBS was prepared by the standard manual method to get a monolayer smear, followed by staining using the Leishman Stain. For each blood sample from the group of 138 patients, one stained blood smear was prepared. The same set of blood smears were analyzed by manual microscopy and also by the automated whole-slide scanner AI100 with the aid of telepathology.
A semi-quantitative morphological examination of the stained smears was performed by manual microscopy using a 40X objective. A total of at least 100 cells was counted on each slide to ascertain the morphological characteristics of the slides. A typical smear consisted of three zones, namely head (thick region), body, and tail (thin end). The cell counting was performed starting from the region between body and tail wherein the red blood cells began to overlap. The smear was examined by moving the field of view from one region to the other without any overlap. Based on the morphological features, the specific type of leukocytes was recorded. The semi-quantitative counting was performed by classifying a feature to be of 'Occasional' occurrence if a count between 1 and 2 was observed per 100 cells counted and if the count was 3 or more, then the same was classified as 'Yes' for that specific feature. It must be mentioned that the correlation analysis of the semi-quantitative data by assigning weightage proportional to the extent of the occurrence did not yield any valuable inference, therefore, those results have been left out of this report. Overall, the correlation with the COVID-19 status as positive or negative was performed using just the qualitative information regarding the morphological characteristics. The specific details about the list of morphological features considered for correlation shall be discussed in the subsequent section.
Telepathology using AI100 whole-slide scanner
All 138 PBS slides analyzed by manual microscopy were also scanned using the automated whole-slide scanner AI100 equipped with 40X objective and white LED to record 120 sharp images from the smear region which exclusively has a monolayer of cells., The patches containing the individual cells were generated from each of the 120 images, along with the 120 images associated with each PBS slide, and were uploaded to the cloud platform called Mandara. Subsequently, a pathologist accessed the patches and Field of View (FOVs) from Mandara to semi-quantitatively evaluate the presence of 15 morphological characteristics associated with the leukocytes. Besides manual visual analysis, an AI image analysis algorithm classified the patches to help evaluate the neutrophil-to-IG ratio for each PBS slide. Once this information was available, the correlation with COVID-19 status as positive or negative was performed using just the qualitative information regarding the morphological characteristics. The specific aspects detailing the list of morphological features considered shall be discussed in the Image Classification and Data Analysis Section.
Image classification and data analysis
The manual image classification and counting were performed by a pathologist using a microscope with a 40X objective and a manual cell counter. The pathologist counted a total of at least 100 leukocytes while looking for 15 morphological features. After the correlation analysis, only 7 of the 15 features were retrospectively found to be helpful in predicting the COVID-19 positive or negative status of the sample. These seven morphological features include plasmacytoid cells, neutrophil dysplastic promyelocyte, neutrophil blast-like cells, AC, smudged neutrophils, neutrophil small myelocyte, and neutrophil Pseudo-Pelger-Huët Anomaly. The representative images of these seven features generated by the whole-slide scanner AI100 are shown in [Figure 1]. In the case of telepathology, the patch generated by AI100 from each smear slide was shared with the pathologist over the cloud and these patches were manually classified into one of the 15 morphological features considered in this study and subsequently, only 7 parameters listed above were used for the COVID-19 predictive analysis. Besides the qualitative visual analysis, two quantitative derivative parameters were also calculated, which included the neutrophil-to-IG ratio and neutrophil-to-lymphocyte. After the correlational analysis, it was observed with specific relation to predicting positive or negative COVID-19 status of the samples, the neutrophil-to-IG ratio was useful in improving the prediction capability while the neutrophil-to-lymphocyte ratio was not as helpful. Therefore, in this report, the analysis results related to the neutrophil-to-IG ratio will be presented in combination with the seven relevant morphological parameters depicted in [Figure 1].
|Figure 1: Representative images of the morphological features generated by AI100. Plasmacytoid cells - PCC (a), neutrophil dysplastic promyelocyte - NDP (b), neutrophil blast-like cells - NBC (c), apoptotic cells - AC (d), smudged neutrophils - SN (e), neutrophil small myelocyte - NSM (f), Neutrophil Pseudo-Pelger-Huët Anomaly - NPPH (g)|
Click here to view
Each of the seven morphological characteristics listed in the previous paragraph was counted purely from a qualitative perspective and five different combinations (presented in [Table 1]) using these seven features were ascertained for each smear slide. For instance, in the case of combination 1, which consisted of three features: PCC, NDP, and NBC, if any one of these were qualitatively positive, then the sample was ascertained to be COVID-19-positive. Likewise, the qualitative assessment of the samples for five different combinations was done, and finally, the sensitivity and specificity of each of these combinations were ascertained by comparing the predictions of these combinations with that of the ground truth from Reverse Transcription Polymerase Chain Reaction (RT-PCR) for all 138 PBS. Besides the combinations of seven morphological features, one additional condition neutrophil-to-IG ratio >35 (necessary condition to be positive) was applied to the best of the five combinations which is combination 3 and the resulting combination was termed combination 3.1. With respect to combination 3.1, it must be noted that the condition neutrophil-to-IG ratio >35 was applied only to the negative samples and not to the positive samples, as this enabled the recovery of a significant number of false-negative samples, while minimally affecting the number of false positives. Needless to say, all these conclusions to achieve the optimal predictability were arrived upon retrospectively. The performance of these six combinations shall be presented in the Results Section.
|Table 1: This is a list of various combinations involving seven morphological characteristics presented in Figure 1 including the plasmacytoid cells (PCC), neutrophil dysplastic promyelocyte (NDP), neutrophil blast-like cells (NBC), apoptotic cells (AC), smudged neutrophils (SN), neutrophil small myelocyte (NSM), Neutrophil Pseudo-Pelger-Huët Anomaly (NPPH), and one derivative parameter neutrophil-to-Ig ratio >35|
Click here to view
| Results|| |
Analysis of individual morphological parameters
As described in the Methods section, after the initial analysis, 15 morphological parameters, 7 parameters were chosen based on the positive and negative correlation with the COVID-19 status of 138 PBS slides. The prevalence of these seven features in the positive and negative COVID-19 samples as identified through manual microscopy and AI100-based telepathology are presented in [Table 2] and [Table 3], respectively. Among the seven morphological features SN, NSM, and NPPH were found to have a high correlation with positive COVID-19 status, as inferred from both the manual and telepathology data. Overall, by comparing the qualitative results, it can be concluded that A100-based telepathology can replace manual microscopy-based qualitative analysis of PBS smears. As the AI100 instrument is capable of classifying and counting the various subtypes of leukocytes, it was possible to calculate one additional parameter neutrophil-to-IG ratio in case of telepathology-based analysis, and based on the preliminary analysis, a condition neutrophil-to-IG ratio >35 was chosen to be added to the list of parameters to be tracked to enable PBS-based accurate classification of the samples into COVID-19-positive or -negative. Once the seven most relevant morphological features were identified, the correlational analysis was performed by considering various combinations of these seven parameters in the base of manual microscopy and one additional parameter including neutrophil-to-IG ratio in case of telepathology, as shown in [Table 1]. The results of the performance of these combinations shall be presented in the next section.
|Table 2: Imaging performed through manual microscopy. Percentage of COVID-19-positive and -negative samples (out of the total 138 PBS) qualitatively exhibiting the following morphological including plasmacytoid cells (PCC), neutrophil dysplastic promyelocyte (NDP), neutrophil blast-like cells (NBC), apoptotic cells (AC), smudged neutrophils (SN), neutrophil - small myelocyte (NSM), and neutrophil - Pseudo-Pelger-Huët Anomaly (NPPH)|
Click here to view
|Table 3: Imaging performed using AI100 and patches shared through telepathology. The percentage of COVID-19-positive and -negative samples (out of the total 138 PBS) qualitatively exhibiting the following morphological including plasmacytoid cells (PCC), neutrophil dysplastic promyelocyte (NDP), neutrophil blast-like cells (NBC), apoptotic cells (AC), smudged neutrophils (SN), neutrophil - small myelocyte (NSM), Neutrophil - Pseudo-Pelger-Huët Anomaly (NPPH), and condition neutrophil-to-Ig ratio >35|
Click here to view
Performance of combinations to predict the COVID-19 status
A list of five combinations in case of manual microscopy-based analysis and six combinations in the case of telepathology-based analysis were quantitatively ascertained from the point of view of predicting the positive or negative status of the PBS slides by calculating the sensitivity, specificity, and total percentage of the outliers which represents the number of instances where the status of the sample was wrongly predicted. The performance for manual microscopy-based assessment is presented in [Table 4], while that corresponding to the telepathology is presented in [Table 5]. At the outset, with regards to combinations 1–5, the performance of these combinations did not vary significantly when the manual microscopy basis analysis was compared with the telepathology-based analysis. In both these cases, as one moves from combinations 1–5, it can be observed that the specificity improves significantly at the cost of sensitivity, and a good balance between both was achieved for combination 3 consisting of PCC, NDP, NBC, AC, and SN which achieved little over 0.7 for both sensitivity and specificity. This observation was true for both manual microscopy and telepathology. Subsequently, in the case of telepathology, the additional condition neutrophil-to-IG ratio >35 necessary for the sample to be COVID-19-positive was applied to the negative samples and this enabled the conversion of a significantly larger number of false-negative samples into true positives when compared to the number of conversions of true negative to false positives, therefore, improving the sensitivity by 0.09 against a decline in specificity by 0.04 or in other words an overall drop in the percentage of outliers by 3.7%, as seen from [Table 5].
|Table 4: Performance of manual microscopy-based prediction of positive or negative status of 138 PBS slides using various combinations of morphological features. The % outliers include both false positives and false negatives|
Click here to view
|Table 5: Performance of telepathology-based prediction of positive or negative status of 138 PBS slides using various combinations of morphological features. The % outliers include both false positives and false negatives|
Click here to view
| Conclusions|| |
We present an evaluation study to ascertain the performance of a telepathology-based screening tool to identify the COVID-19-positive or -negative status of blood samples using PBS-derived morphological characteristics of the leukocytes. The study was performed using a total sample size of 138 samples consisting of 82 COVID-19-positive samples and 56 negative samples, all having flu-like symptoms. In the case of manual microscopy as well as telepathology-based analysis, a specific combination of the morphological characteristics consisting of PCC, NDP, NBC, AC, and SN was identified as the best among all the possible combinations at predicting the COVID-19 status of a sample. This 5-parameter combination achieved the sensitivity and specificity of a little over 0.7. Upon adding the condition neutrophil-to-IG ratio >35 to negative samples associated with telepathology data, the number of outliers decreased by 3.7%. Overall, we present a moderately accurate screening tool to identify COVID-19-positive or -negative status using leukocyte morphological information derived from the telepathology-based analysis of PBS performed using an automated whole-slide scanner.
| Discussion|| |
As per the literature related to PBS and COVID-19 covered in the Introduction Section, reactive lymphocytes, vacuolated monocytes, and neutrophil-to-lymphocyte ratio were found to be the most important characteristics from the point of view of the severity of the symptoms. It was surprising that none of these three parameters featured in the best combination of morphological parameters ascertained by the study presented here. One significant differentiating factor pertaining to this study is the correlation with COVID-19-positive or -negative status of PBS rather than the severity of the symptoms, especially focusing on the events of hospitalization and fatality. Compared to a highly accurate RT-PCR and antibody-based COVID-19 diagnosis tests, the PBS-based screening tool presented here may be just moderately accurate, but this level of performance could be of immense help in the initial phase of the pandemic when affordable accurate kits were yet to become available. The robustness of the starting materials associated with PBS including glass slide, Leishman stain, and slide scanner or microscope from the point of view of storage makes this a very versatile diagnostic tool compared to the molecular or immunodiagnostic kits which require special temperature conditions. Therefore, a PBS-based screening tool would be of immense help in resource-limited settings, especially in undeveloped and developing countries. Besides specific relevance to COVID-19, the overall exercise presented here involving the correlation study through multiparameter morphological analysis and subsequent association with the positive or negative status of a novel outbreak must be part of a first-level response in the future. In the likelihood of high accuracy for PBS-based screening solution, it would be a quick globally implementable tool to manage future outbreaks.
We wish to acknowledge the support of Ms. Pooja C at Sigtuple Technologies Pvt. Ltd. towards conducting the clinical study.
Financial support and sponsorship
Conflicts of interest
There are no conflicts of interest.
| References|| |
Huang C, Wang Y, Li X, Ren L, Zhao J, Hu Y, et al
. Clinical features of patients infected with 2019 novel coronavirus in Wuhan, China. Lancet 2020;395:497-506.
Ravi N, Cortade DL, Ng E, Wang XS. Diagnostics for SARS-CoV-2 detection: A comprehensive review of the FDA-EUA COVID-19 testing landscape. Biosens Biolectron 2020;165:112454.
Zhang D, Guo R, Lei L, Liu H, Wang Y, Wang Y, et al
. Frontline Science: COVID-19 infection induces readily detectable morphologic and inflammation-related phenotypic changes in peripheral blood monocytes. J Leukoc Biol 2021;109:13-22.
Xie G, Ding F, Han L, Yin D, Lu H, Zhang M. The role of peripheral blood eosinophil counts in COVID-19 patients. Allergy 2021;76:471-82.
Qin C, Zhou L, Hu Z, Zhang S, Yang S, Tao Y, et al
. Dysregulation of immune response in patients with coronavirus 2019 (COVID-19) in Wuhan, China. Clin Infect Dis 2020;71:762-8.
Yang AP, Liu JP, Tao WQ, Li HM. The diagnostic and predictive role of NLR, d-NLR and PLR in COVID-19 patients. Int Immunopharmacol 2020;84:106504.
Liu J, Li S, Liu J, Liang B, Wang X, Wang H, et al
. Longitudinal characteristics of lymphocyte responses and cytokine profiles in the peripheral blood of SARS-CoV-2 infected patients. EBioMedicine 2020;55:102763.
Sadigh S, Massoth RL, Christensen BB, Stefely JA, Keefe J, Sohani AR. Peripheral blood morphologic findings in patients with COVID-19. Int J Lab Hematol 2020;42:e248-51.
Nazarullah A, Liang C, Villarreal A, Higgins RA, Mais DD. Peripheral blood examination findings in SARS-CoV-2 infection. Am J Clin Pathol 2020;154:319-29.
Zini G, Bellesi S, Ramundo F, d'Onofrio G. Morphological anomalies of circulating blood cells in COVID-19. Am J Hematol 2020;95:870-2.
Berber I, Cagasar O, Sarici A, Berber NK, Aydogdu I, Ulutas O, et al
. Peripheral blood smear findings of COVID-19 patients provide ̇information about the severity of the disease and the duration of hospital stay. Mediterr J Hematol Infect Dis 2021;13:e2021009.
Chong VCL, Lim KGE, Fan BE, Chan SSW, Ong KH, Kuperan P. Reactive lymphocytes in patients with COVID-19. Br J Haematol 2020;189:844.
El Jamal SM, Salib C, Stock A, Uriarte-Haparnas NI, Glicksberg BS, Teruya-Feldstein J, et al
. Atypical lymphocyte morphology in SARS-CoV-2 infection. Pathol Res Pract 2020;216:153063.
Lüke F, Orsó E, Kirsten J, Poeck H, Grube M, Wolff D, et al
. Coronavirus disease 2019 induces multi-lineage, morphologic changes in peripheral blood cells. ejHaem 2020;1:376-83.
Cantu MD, Towne WS, Emmons FN, Mostyka M, Borczuk A, Salvatore SP, et al
. Clinical significance of blue-green neutrophil and monocyte cytoplasmic inclusions in SARS-CoV-2 positive critically ill patients. Br J Haematol 2020;190:e89-92.
Dastidar RT, Ethirajan R. Whole slide imaging system using deep learning-based automated focusing. Biomed Opt Express 2019;11:480-91.
Lutnick B, Manthey D, Pinaki Sarder P. Atool for user friendly, cloud based, whole slide image segmentation. Electr Eng Syst Sci 2021:2101.07222-2.
Weinstein RS, Graham AR, Richter LC, Barker GP, Krupinski EA, Lopez AM, et al
. Overview of telepathology, virtual microscopy, and whole slide imaging: Prospects for the future. Hum Pathol 2009;40:1057-69.
Hitchcock CL. The future of telepathology for the developing world. Arch Pathol Lab Med 2011;135:211-4.
Henriksen J, Kolognizak T, Houghton T, Cherne S, Zhen D, Cimino JP, et al
. Rapid validation of telepathology by an academic neuropathology practice during the COVID-19 pandemic. Arch Pathol Lab Med 2020;144:1311-20.
Bain JB. Blood Cells: Blood Sampling and Blood Film Preparation and Examination. 5th
ed. John Wiley Sons Ltd; 2014.
Mundhra D, Cheluvaraju B, Rampure J, Dastidar RT. Analyzing microscopic images of peripheral blood smear using deep learning. Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support. 2017;10553:178-85.
R Arcot Lokanathan
Sigtuple Technologies Pvt. Ltd., L-162, 14th Cross Rd, Sector 6, HSR Layout, Bengaluru, Karnataka - 560102
Source of Support: None, Conflict of Interest: None
[Table 1], [Table 2], [Table 3], [Table 4], [Table 5]