Cytometry Part B: Clinical Cytometry最新文献

Machine-learning (ML) models in flow cytometry have the potential to reduce error rates, increase reproducibility, and boost the efficiency of clinical labs. While numerous ML models for flow cytometry data have been proposed, few studies have described the clinical deployment of such models. Realizing the potential gains of ML models in clinical labs requires not only an accurate model but also infrastructure for automated inference, error detection, analytics and monitoring, and structured data extraction. Here, we describe an ML model for the detection of Acute Myeloid Leukemia (AML), along with the infrastructure supporting clinical implementation. Our infrastructure leverages the resilience and scalability of the cloud for model inference, a Kubernetes-based workflow system that provides model reproducibility and resource management, and a system for extracting structured diagnoses from full-text reports. We also describe our model monitoring and visualization platform, an essential element for ensuring continued model accuracy. Finally, we present a post-deployment analysis of impacts on turn-around time and compare production accuracy to the original validation statistics.

T lymphoblastic leukemia/lymphoma (T-ALL) is a malignancy composed of proliferating T lymphoblasts. T lymphoblasts can be identified by flow cytometric analysis through the detection of aberrant antigen expression and/or immaturity marker expression. An ideal marker would be expressed brightly on T lymphoblasts but absent in mature T lymphocytes. One such marker is protein tyrosine kinase-7 (PTK7). However, PTK7 has not been widely adopted in clinical flow cytometry labs or incorporated into any T-ALL flow cytometry best practice recommendations. To this end, we demonstrate the utility of PTK7 in flow cytometry panels for T-ALL diagnosis, minimal/measurable residual disease (MRD) detection, and relapse. We retrospectively evaluated flow cytometry data on 175 patients. PTK7 was classified as positive, showing a near two-fold difference in brightness versus background mature T cells, in 87.76% of T-ALL cases at initial diagnosis, 75% of T-ALL at MRD, and 100% of T-ALL at relapse. PTK7 expression remained intact in cases of CD34 and/or TdT negative T-ALL (p = 0.992) and while expression was dimmer at MRD (72% decrease, p = 0.0313), PTK7 remained intact at relapse (33% increase, p = 0.8125). PTK7 should be included in flow cytometry panels when evaluating for T-ALL, both at initial diagnosis, relapse, and for the presence of MRD.

Implementing a new laboratory information system (LIS) presents an opportunity to improve operational efficiency and streamline reporting by refining workflows by utilizing LIS functionality. Flow cytometry laboratories face unique challenges because the specimen and test results may be categorized under clinical pathology (CP), anatomic pathology (AP), or both. We describe the design and implementation of reporting flow cytometry results within the Epic Beaker CP module, its interface with the Epic Beaker AP module, and integrated reporting for AP/CP cases at an academic institution. This manuscript emphasizes the challenges and steps needed to integrate anatomic and clinical pathology workflows by leveraging LIS functionality to implement electronic and predominantly paperless workflows within a flow cytometry laboratory.

Blast quantification in the bone marrow (BM) is crucial for evaluating myeloid neoplasms, with cytomorphology being the only recognized analysis. The CD34 myeloid cell (CD34M) count by flow cytometry is promising but impaired by BM hemodilution. A modified version of the Holdrinet index (mHI) is routinely used to assess it, though not yet validated. By analyzing two differently hemodiluted BM samples from 51 patients with suspicion of myelodysplasia, this study confirms mHI accuracy in assessing BM white blood cell purity. mHI-adjusted count by flow cytometry of BM-exclusive cell subsets, such as CD34 myeloid cells, may offer a reliable and practical alternative to cytomorphology analysis, independently from hemodilution.

A reduced proportion of peripheral class-switched memory B cells (CSM-B cells) is presumed to indicate ineffective germinal activity. The extent that this finding corresponds to a plausible germinal center failure pathophysiology in patients not diagnosed with CVID or hyper IgM syndrome is not known. We asked if patients with low CSM-B cells are more likely to demonstrate failure to produce serum IgA and IgG than counterparts with nonreduced class-switched memory B cell levels, regardless of diagnosis. Patients with low CSM-B cell levels regardless of diagnosis were retrospectively compared with their counterparts without CSM-B cell reductions for demographics and serum immunoglobulin levels. Patients were further divided based on whether CSM-B cell levels remained low over time or fluctuated, and these groups were compared for serum immunoglobulin levels and diagnoses. Of 305 patients, those with CSM-B cell (n = 50) reductions were more likely to have low serum IgA and IgG than those without reductions. Of the 78 patients in whom CSM-B cells were measured repeatedly over time, 21 patients had low CSM-B cell levels, but this finding was persistent in only 10. Patients with persistent CSM-B cell reductions universally had severe serum IgA and IgG deficiencies and patients with transient CSM-B cell reduction often did not. These two groups contained divergent diagnoses and likely represent separate pathophysiologic groups. Quantifying CSM-B cells as a percentage of CD27+ B cells repeatedly over time is a robust approach to identifying patients with a plausible germinal center failure endotype.