Automating the Detection of IV Fluid Contamination Using Unsupervised Machine Learning.

BACKGROUND: Intravenous (IV) fluid contamination is a common cause of preanalytical error that can delay or misguide treatment decisions, leading to patient harm. Current approaches for detecting contamination rely on delta checks, which require a prior result, or manual technologist intervention, which is inefficient and vulnerable to human error. Supervised machine learning may provide a means to detect contamination, but its implementation is hindered by its reliance on expert-labeled training data. An automated approach that is accurate, reproducible, and practical is needed. METHODS: A total of 25 747 291 basic metabolic panel (BMP) results from 312 721 patients were obtained from the laboratory information system (LIS). A Uniform Manifold Approximation and Projection (UMAP) model was trained and tested using a combination of real patient data and simulated IV fluid contamination. To provide an objective metric for classification, an "enrichment score" was derived and its performance assessed. Our current workflow was compared to UMAP predictions using expert chart review. RESULTS: UMAP embeddings from real patient results demonstrated outliers suspicious for IV fluid contamination when compared with the simulated contamination's embeddings. At a flag rate of 3 per 1000 results, the positive predictive value (PPV) was adjudicated to be 0.78 from 100 consecutive positive predictions. Of these, 58 were previously undetected by our current clinical workflows, with 49 BMPs displaying a total of 56 critical results. CONCLUSIONS: Accurate and automatable detection of IV fluid contamination in BMP results is achievable without curating expertly labeled training data.

Organizational preparedness for the use of large language models in pathology informatics.

In this paper, we consider the current and potential role of the latest generation of Large Language Models (LLMs) in medical informatics, particularly within the realms of clinical and anatomic pathology. We aim to provide a thorough understanding of the considerations that arise when employing LLMs in healthcare settings, such as determining appropriate use cases and evaluating the advantages and limitations of these models. Furthermore, this paper will consider the infrastructural and organizational requirements necessary for the successful implementation and utilization of LLMs in healthcare environments. We will discuss the importance of addressing education, security, bias, and privacy concerns associated with LLMs in clinical informatics, as well as the need for a robust framework to overcome regulatory, compliance, and legal challenges.

A Single Reference Interval for Interpreting Serum Free Light Chains across Patients with Varying Renal Function.

BACKGROUND: Serum free light chain (sFLC) assays are interpreted using a sFLC-ratio-based reference interval (manufacturer's interval) that was defined using a cohort of healthy patients. However, renal impairment elevates the sFLC-ratio, leading to a high false positive rate when using the manufacturer's interval. Prior studies have developed renal-specific reference intervals; however, this approach has not been widely adopted due to practical limitations. Thus, there remains a critical need for a renally robust sFLC interpretation method. METHODS: Retrospective data mining was used to define patient cohorts that reflect the spectrum of renal function seen in clinical practice. Two new reference intervals, one based on the sFLC-ratio and one based on a novel principal component analysis (PCA)-based metric, were developed for the FREELITE assay (Binding Site) on the Roche Cobas c501 instrument (Roche). RESULTS: Compared to the manufacturer's reference interval, both new methods exhibited significantly lower false positive rates and greater robustness to renal function while maintaining equivalent sensitivity for monoclonal gammopathy (MG) diagnosis. While not significantly different, the point estimate for sensitivity was highest for the PCA-based approach. CONCLUSION: Renally robust sFLC interpretation using a single reference interval is possible given a reference cohort that reflects the variation in renal function observed in practice. Further studies are needed to achieve sufficient power and determine if the novel PCA-based metric offers superior sensitivity for MG diagnosis. These new methods offer the practical advantages of not requiring an estimated glomerular filtration rate result or multiple reference intervals, thereby lowering practical barriers to implementation.

Optimizing Equity: Working towards Fair Machine Learning Algorithms in Laboratory Medicine.

BACKGROUND: Methods of machine learning provide opportunities to use real-world data to solve complex problems. Applications of these methods in laboratory medicine promise to increase diagnostic accuracy and streamline laboratory operations leading to improvement in the quality and efficiency of healthcare delivery. However, machine learning models are vulnerable to learning from undesirable patterns in the data that reflect societal biases. As a result, irresponsible application of machine learning may lead to the perpetuation, or even amplification, of existing disparities in healthcare outcomes. CONTENT: In this work, we review what it means for a model to be unfair, discuss the various ways that machine learning models become unfair, and present engineering principles emerging from the field of algorithmic fairness. These materials are presented with a focus on the development of machine learning models in laboratory medicine. SUMMARY: We hope that this work will serve to increase awareness, and stimulate further discussion, of this important issue among laboratorians as the field moves forward with the incorporation of machine learning models into laboratory practice.

Comparison of the Friedewald equation with Martin and Sampson equations for estimating LDL cholesterol in hypertriglyceridemic adults.

Low density lipoprotein cholesterol (LDL-C) is traditionally calculated using the Friedewald (LDL-F) equation. New equations by Martin (LDL-M) and Sampson (LDL-S) have improved accuracy relative to LDL-F for samples with high triglycerides (TG) or low LDL-C. However, most labs still rely on LDL-F and few studies have examined the accuracy and impact of contemporary LDL-C equations applied to a retrospective dataset. 934 lipid panels with a concurrent direct enzymatic LDL-C (dLDL-C) result were extracted from the laboratory information system. LDL-F, LDL-M, and LDL-S were calculated and the accuracy of each equation determined in a predominantly hypertriglyceridemic population. The impact of implementing each equation was compared by analyzing the LDL-C treatment group miscategorization rate relative to dLDL-C. The slope for the LDL-F, LDL-M and LDL-S were 0.59, 0.78, and 0.94, relative to dLDL-C. The three equations performed comparably for samples with TG <4.52 mmol/L (<400 mg/dL). The LDL-C treatment group miscategorization rate was 48.6 % for LDL-F, 28.8 % for LDL-M and 37.2 % for LDL-S in specimens with TG ≥4.52 mmol/L (≥400 mg/dL) (n = 817). LDL-S underestimated treatment group category (31.3 %, 95 % CI 17.2-22.4) relative to LDL-M (9.0 %, 4.39-7.41, P < 0.001). 5.9 % of samples were overestimated for treatment group category by LDL-S vs 19.8 % for LDL-M (P = 0.1883). LDL-M and LDL-S demonstrate reduced bias with a dLDL-C method compared to LDL-F in samples with TG ≥4.52 mmol/L (≥400 mg/dL). LDL-M reduces LDL-C treatment group miscategorization rate leading to fewer underestimations of risk overall compared to LDL-S; however, neither may be sufficiently accurate to report LDL-C in patients with TG ≥4.52 mmol/L (≥400 mg/dL).

Use of laboratory data for illicit drug use surveillance and identification of socioeconomic risk factors.

BACKGROUND: Drug overdose is the leading cause of death among people 25-44 years of age in the United States. Existing drug surveillance methods are important for prevention and directing treatment, but are limited by delayed reporting and lack of geographic granularity. METHODS: Laboratory urine drug screen and complete metabolic panel data from patients presenting to the emergency department was used to observe long-term and short-term temporal and geospatial changes at the zip code-level in St. Louis. Multivariate linear regression was performed to investigate associations between zip code-level socioeconomic factors and drug screening positivity rates. RESULTS: An increase in the fentanyl positive drug screens was seen during the initial COVID-19 shutdown period in the spring of 2020. A decrease in cocaine positivity was seen in the fall and winter of 2020, with a return to baseline coinciding with the second major COVID-19 shutdown in the summer of 2021. These changes appeared to be independent of changes in emergency department utilization as measured by complete metabolic panels ordered. Significant short-term changes in fentanyl and cocaine positivity rates between specific time periods were able to be localized to individual zip codes. Zip code-level multivariate analysis demonstrated independent associations between socioeconomic/demographic factors and fentanyl/cocaine positivity rates as determined by laboratory drug screening data. CONCLUSIONS: Analyzing clinical laboratory drug screening data can enable a more temporally and geographically granular view of population-level drug use surveillance. Additionally, laboratory data can be utilized to find population-level socioeconomic associations with illicit drug use, presenting a potential avenue for the use of this data to guide public health and healthcare policy decisions.

State estimation-based control of COVID-19 epidemic before and after vaccine development.

In this study, a nonlinear robust control policy is designed together with a state observer in order to manage the novel coronavirus disease (COVID-19) outbreak having an uncertain epidemiological model with unmeasurable variables. This nonlinear model for the COVID-19 epidemic includes eight state variables (susceptible, exposed, infected, quarantined, hospitalized, recovered, deceased, and insusceptible populations). Two plausible scenarios are put forward in this article to control this epidemic before and after its vaccine invention. In the first scenario, the social distancing and hospitalization rates are employed as two applicable control inputs to diminish the exposed and infected groups. However, in the second scenario after the vaccine development, the vaccination rate is taken into account as the third control input to reduce the susceptible populations, in addition to the two objectives of the first scenario. The proposed feedback control measures are defined in terms of the hospitalized and deceased populations due to the available statistical data, while other unmeasurable compartmental variables are estimated by an extended Kalman filter (EKF). In other words, the susceptible, exposed, infected, quarantined, recovered, and insusceptible individuals cannot be identified precisely because of the asymptomatic infection of COVID-19 in some cases, its incubation period, and the lack of an adequate community screening. Utilizing the Lyapunov theorem, the stability and bounded tracking convergence of the closed-loop epidemiological system are investigated in the presence of modeling uncertainties. Finally, a comprehensive simulation study is conducted based on Canada's reported cases for two defined timing plans (with different treatment rates). Obtained results demonstrate that the developed EKF-based control scheme can achieve desired epidemic goals (exponential decrease of infected, exposed, and susceptible people).

An Extensive Set of Kinematic and Kinetic Data for Individuals with Intact Limbs and Transfemoral Prosthesis Users.

This paper introduces an extensive human motion data set for typical activities of daily living. These data are crucial for the design and control of prosthetic devices for transfemoral prosthesis users. This data set was collected from seven individuals, including five individuals with intact limbs and two transfemoral prosthesis users. These data include the following types of movements: (1) walking at three different speeds; (2) walking up and down a 5-degree ramp; (3) stepping up and down; (4) sitting down and standing up. We provide full-body marker trajectories and ground reaction forces (GRFs) as well as joint angles, joint velocities, joint torques, and joint powers. This data set is publicly available at the website referenced in this paper. Data from flexion and extension of the hip, knee, and ankle are presented in this paper. However, the data accompanying this paper (available on the internet) include 46 distinct measurements and can be useful for validating or generating mathematical models to simulate the gait of both transfemoral prosthesis users and individuals with intact legs.

Robust Cell Detection and Segmentation for Image Cytometry Reveal Th17 Cell Heterogeneity.

Image cytometry enables quantitative cell characterization with preserved tissue architecture; thus, it has been highlighted in the advancement of multiplex immunohistochemistry (IHC) and digital image analysis in the context of immune-based biomarker monitoring associated with cancer immunotherapy. However, one of the challenges in the current image cytometry methodology is a technical limitation in the segmentation of nuclei and cellular components particularly in heterogeneously stained cancer tissue images. To improve the detection and specificity of single-cell segmentation in hematoxylin-stained images (which can be utilized for recently reported 12-biomarker chromogenic sequential multiplex IHC), we adapted a segmentation algorithm previously developed for hematoxlin and eosin-stained images, where morphological features are extracted based on Gabor-filtering, followed by stacking of image pixels into n-dimensional feature space and unsupervised clustering of individual pixels. Our proposed method showed improved sensitivity and specificity in comparison with standard segmentation methods. Replacing previously proposed methods with our method in multiplex IHC/image cytometry analysis, we observed higher detection of cell lineages including relatively rare TH 17 cells, further enabling sub-population analysis into TH 1-like and TH 2-like phenotypes based on T-bet and GATA3 expression. Interestingly, predominance of TH 2-like TH 17 cells was associated with human papilloma virus (HPV)-negative status of oropharyngeal squamous cell carcinoma of head and neck, known as a poor-prognostic subtype in comparison with HPV-positive status. Furthermore, TH 2-like TH 17 cells in HPV-negative head and neck cancer tissues were spatiotemporally correlated with CD66b+ granulocytes, presumably associated with an immunosuppressive microenvironment. Our cell segmentation method for multiplex IHC/image cytometry potentially contributes to in-depth immune profiling and spatial association, leading to further tissue-based biomarker exploration. © 2019 International Society for Advancement of Cytometry.

Deep learning based Nucleus Classification in pancreas histological images.

Tumor specimens contain a variety of healthy cells as well as cancerous cells, and this heterogeneity underlies resistance to various cancer therapies. But this problem has not been thoroughly investigated until recently. Meanwhile, technological breakthroughs in imaging have led to an explosion of molecular and cellular profiling data from large numbers of samples, and modern machine learning approaches including deep learning have been shown to produce encouraging results by finding hidden structures and make accurate predictions. In this paper, we propose a Deep learning based Nucleus Classification (DeepNC) approach using paired histopathology and immunofluorescence images (for label), and demonstrate its classification prediction power. This method can solve current issue on discrepancy between genomic- or transcriptomic-based and pathology-based tumor purity estimates by improving histological evaluation. We also explain challenges in training a deep learning model for huge dataset.

Quantitative Multiplex Immunohistochemistry Reveals Myeloid-Inflamed Tumor-Immune Complexity Associated with Poor Prognosis.

Here, we describe a multiplexed immunohistochemical platform with computational image processing workflows, including image cytometry, enabling simultaneous evaluation of 12 biomarkers in one formalin-fixed paraffin-embedded tissue section. To validate this platform, we used tissue microarrays containing 38 archival head and neck squamous cell carcinomas and revealed differential immune profiles based on lymphoid and myeloid cell densities, correlating with human papilloma virus status and prognosis. Based on these results, we investigated 24 pancreatic ductal adenocarcinomas from patients who received neoadjuvant GVAX vaccination and revealed that response to therapy correlated with degree of mono-myelocytic cell density and percentages of CD8+ T cells expressing T cell exhaustion markers. These data highlight the utility of in situ immune monitoring for patient stratification and provide digital image processing pipelines to the community for examining immune complexity in precious tissue sections, where phenotype and tissue architecture are preserved to improve biomarker discovery and assessment.

BREAST CANCER HISTOPATHOLOGY IMAGE ANALYSIS PIPELINE FOR TUMOR PURITY ESTIMATION.

The translation of genomic sequencing technology to the clinic has greatly advanced personalized medicine. However, the presence of normal cells in tumors is a confounding factor in genome sequence analysis. Tumor purity, or the percentage of cancerous cells in whole tissue section, is a correction factor that can be used to improve the clinical utility of genomic sequencing. Currently, tumor purity is estimated visually by expert pathologists; however, it has been shown that there exist vast inter-observer discrepancies in tumor purity scoring. In this paper, we propose a quantitative image analysis pipeline for tumor purity estimation and provide a systematic comparison between pathologists' scores and our image-based tumor purity estimation.

Quantitative analysis of histological tissue image based on cytological profiles and spatial statistics.

The cellular heterogeneity and complex tissue architecture of most tumor samples is a major obstacle in image analysis on standard hematoxylin and eosin-stained (H&E) tissue sections. A mixture of cancer and normal cells complicates the interpretation of their cytological profiles. Furthermore, spatial arrangement and architectural organization of cells are generally not reflected in cellular characteristics analysis. To address these challenges, first we describe an automatic nuclei segmentation of H&E tissue sections. In the task of deconvoluting cellular heterogeneity, we adopt Landmark based Spectral Clustering (LSC) to group individual nuclei in such a way that nuclei in the same group are more similar. We next devise spatial statistics for analyzing spatial arrangement and organization, which are not detectable by individual cellular characteristics. Our quantitative, spatial statistics analysis could benefit H&E section analysis by refining and complementing cellular characteristics analysis.