Background: Cytomorphology is the gold standard for quick assessment of peripheral blood and bone marrow samples in hematological neoplasms. It is a broadly-accepted method for orchestrating more specific diagnostics including immunophenotyping or genetics. Inter-/intra-observer-reproducibility of single cell classification is only 75 to 90%. Only a limited number of cells (100 - 500 cells/smear) is read in a time-consuming procedure. Machine learning (ML) is more reliable where human skills are limited, i.e. in handling large amounts of data or images. We here tested ML to differentiate peripheral blood leukocytes in a high throughput hematology laboratory. Aim: To establish an ML-based cell classifier capable of identifying healthy and pathologic cells in digitalized peripheral blood smear scans at an accuracy competitive with or outperforming human expert level. Methods: We selected >2,600 smears out of our unique archive of > 250,000 peripheral blood smears from hematological neoplasms. Depending on quality, we scanned up to 1,000 single cell images per smear. For image acquisition, a Metafer Scanning System (Zeiss Axio Imager.Z2 microscope, automatic slide feeder and automatic oiling device) from MetaSystems (Altlussheim, GER) was used. Areas of interest were defined by pre-scan in 10x magnification followed by high resolution scan in 40x to generate cell images for analysis. Average capture times for 300/500 cells were 3:43/4:37 min We set up a supervised ML-learning model using colour images (144x144 pixels) as input, outputting predicted probabilities of 21 predefined classes. We used ImageNet-pretrained Xception as our base model. We trained, evaluated and deployed the model using Amazon SageMaker on a subset of 82,974 images randomly selected from 514,183 cells captured and labelled for this study. 20 different cell types and one garbage class were classified. We included cell type categories referring to the critical importance of detecting rare leukemia subtypes (e.g. APL). Numbers of images from respective 21 classes ranged from 1,830 to 14,909 (median: 2,945). Minority classes were up-sampledto handle imbalances. Each picture was labelled by highly skilled technicians (median years practicing in this laboratory: 5) and two independent hematologists (median years at microscope: 20). Results: On a separate test set of 8,297 cells, our classifier was able to predict any of the five cell types occurring in the peripheral blood of healthy individuals (PMN, lymphocytes, monocytes, eosinophils, basophils) at very high median accuracy (97.0%) Median prediction accuracy of 15 rare or pathological cell types was 91.3%. For six critical pathological cell forms (myeloblasts, atypical/bilobulated promyelocytes in APL/APLv, hairy cells, lymphoma cells,plasma cells), median accuracy was 93.4% (sensitivity 93.8%). We saw a very high "T98 accuracy" for these cell types (98.5%) which is the accuracy of cell type predictions with prediction probability >0.98 (achieved in 2231/2417 cases), implicating that critical cells predicted with probability <0.98 should be flagged for human expert validation with priority. For all 21 classes median accuracy was 91.7%. Accuracy was lower for cells representing consecutive steps of maturation, e.g. promyelo-/myelo-/metamyelocytes, reproducing inconsistencies from the human-built phenotypic classification system (s.Fig.). Conclusions: We demonstrate an automated workflow using automatic microscopic cell capturing and ML-driven cell differentiation in samples of hematologic patients. Reproducibility, accuracy, sensitivity and specificity are above 90%, for many cell types above 98%. By flagging suspicious cells for humanvalidation, this tool can support even experienced hematology professionals, especially in detecting rare cell types. Given an appropriate scanning speed, it clearly outperforms human investigators in terms of examination time and number of differentiated cells. An ML-based intelligence can make its skills accessible to hematology laboratories on site or after upload of scanned cell images, independent of time/location. A cloud-based infrastructure is available. A prospective head to head challenge between ML-based classifier and human experts comparing sensitivity and accuracy for detection of all cell classes in peripheral blood will be tested to proof suitability for routine use (NCT 4466059). Figure Disclosures Heo: AWS: Current Employment. Wetton:AWS: Current Employment. Drescher:MetaSystems: Current Employment. Hänselmann:MetaSystems: Current Employment. Lörch:MetaSystems: Current equity holder in private company.
Background: Machine Learning (ML) offers automated data processing substituting various analysis steps. So far it has been applied to flow cytometry (FC) data only after visualization which may compromise data by reduction of data dimensionality. Automated analysis of FC raw matrix data has not yet been pursued. Aim: To establish as proof of concept an ML-based classifier processing FC matrix data to predict the correct lymphoma type without the need for visualization or human analysis and interpretation. Methods: A set of 6,393 uniformly analyzed samples (Navios cytometers, Kaluza software, Beckman Coulter, Miami, FL) was used for training (n=5,115) and testing (n=1,278) of different ML models. Entities were chronic lymphatic leukemia (CLL) 1103 (training) and 279 (testing), monoclonal B-cell lymphocytosis (MBL, 831/203), CLL with increased prolymphocytes (CLL-PL, 649/161), lymphoplasmacytic lymphoma (LPL, 560/159), hairy cell leukemia (HCL, 328/88), mantle cell lymphoma (MCL, 259/53), marginal zone lymphoma (MZL, 90/28), follicular lymphoma (FL, 84/16), no lymphoma (1211/291). Three tubes comprising 11 parameters per tube were applied. Besides scatter signals analyzed antigens included: CD3, CD4, CD5, CD8, CD10, CD11c, CD19, CD20, CD22, CD23, CD25, CD38, CD45, CD56, CD79b, CD103, FMC7, HLA-DR, IgM, Kappa, Lambda. Measurements generated LMD files with 50,000 rows of data for each of the 11 parameters. After removing the saturated values (≥ 1023) we produced binned histograms with 16 predefined frequency bins per parameter. Histograms were converted to cumulative distribution functions (CDF) for respective parameters and concatenated to produce a 16x11 matrix per each tube. Following the assumption of independence of parameters this simplification of concatenating CDFs represents the same information as if they were jointly distributed. The first matrix-based classifier was a decision tree model (DT), the second a deep learning model (DL) and the third was an XGBoost (XG) model, an implementation of gradient boosted decision trees ideal for structured tabular data (such as LMD files). The first set of analyses included only three classes which are readily separated by human operators: 1) CLL, 2) HCL, 3) no lymphoma. The second set included all nine entities but grouped into four classes: 1) CD5+ lymphoma (CLL, MBL, CLL-PL, MCL), 2) HCL, 3) other CD5- lymphoma (LPL, MZL, FL), 4) no lymphoma. The third set included each of the nine entities as its own class. Results: Analyzing the three classes from the first set (CLL, HCL, no lymphoma) the models achieved accuracies of 94% (DT), 95% (DL) and 96% (XG) when including all cases. By analysis of cases with prediction probabilities above 90%, DT now reached 97%, DL 97% and XG 98% accuracy, whilst losing 38%, 8% and 6% of samples, respectively. We further observed that accuracy was also dependent on the size of the pathologic clone, which is in line with the experiences from human experts with very small clones (≤ 0.1% of leukocytes) representing a major challenge regarding their correct classification. Focusing on cases with clones > 0.1% but considering all prediction probabilities accuracies were 96% (DT), 97% (DL) and 98% (XG), with loss of 5% of samples for each model. Considering cases only with prediction probabilities > 90% and clones > 0.1% accuracies were 97% (DT), 99% (DL) and 99% (XG) whilst losing 38%, 9% and 9% of samples, respectively. Further analyses were performed applying the best model based on results above, i.e. XG. Analyzing four classes in the second set of analyses (CD5+ lymphoma, HCL, other CD5- lymphoma, no lymphoma) and considering cases only with prediction probabilities > 95% and clones > 0.1% accuracy was 96% while losing 28% of samples. In the third set of analyses with each entity assigned its own class and again considering cases only with prediction probabilities > 95% and clones > 0.1% accuracy was 93% while losing 28% of samples. Conclusions: This first ML-based classifier using the XGboost model with transforming FC matrix data to concatenated distributions, is capable of correctly assigning the vast majority of lymphoma samples analyzing FC raw data without visualization or human interpretation. Cases that need further attention by human experts will be flagged but will not account for more than 30% of all cases. This data will be extended in a prospective blinded study (clinicaltrials.gov NCT4466059). Disclosures Heo: AWS: Current Employment. Wetton:AWS: Current Employment.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.