Identifying Psychosis Episodes in Psychiatric Admission Notes via Rule-based Methods, Machine Learning, and Pre-Trained Language Models.

Early and accurate diagnosis is crucial for effective treatment and improved outcomes, yet identifying psychotic episodes presents significant challenges due to its complex nature and the varied presentation of symptoms among individuals. One of the primary difficulties lies in the underreporting and underdiagnosis of psychosis, compounded by the stigma surrounding mental health and the individuals' often diminished insight into their condition. Existing efforts leveraging Electronic Health Records (EHRs) to retrospectively identify psychosis typically rely on structured data, such as medical codes and patient demographics, which frequently lack essential information. Addressing these challenges, our study leverages Natural Language Processing (NLP) algorithms to analyze psychiatric admission notes for the diagnosis of psychosis, providing a detailed evaluation of rule-based algorithms, machine learning models, and pre-trained language models. Additionally, the study investigates the effectiveness of employing keywords to streamline extensive note data before training and evaluating the models. Analyzing 4,617 initial psychiatric admission notes (1,196 cases of psychosis versus 3,433 controls) from 2005 to 2019, we discovered that the XGBoost classifier employing Term Frequency-Inverse Document Frequency (TF-IDF) features derived from notes pre-selected by expert-curated keywords, attained the highest performance with an F1 score of 0.8881 (AUROC [95% CI]: 0.9725 [0.9717, 0.9733]). BlueBERT demonstrated comparable efficacy an F1 score of 0.8841 (AUROC [95% CI]: 0.97 [0.9580,0.9820]) on the same set of notes. Both models markedly outperformed traditional International Classification of Diseases (ICD) code-based detection methods from discharge summaries, which had an F1 score of 0.7608, thus improving the margin by 0.12. Furthermore, our findings indicate that keyword pre-selection markedly enhances the performance of both machine learning and pre-trained language models. This study illustrates the potential of NLP techniques to improve psychosis detection within admission notes and aims to serve as a foundational reference for future research on applying NLP for psychosis identification in EHR notes.

Identifying Psychosis Episodes in Psychiatric Admission Notes via Rule-based Methods, Machine Learning, and Pre-Trained Language Models.

Early and accurate diagnosis is crucial for effective treatment and improved outcomes, yet identifying psychotic episodes presents significant challenges due to its complex nature and the varied presentation of symptoms among individuals. One of the primary difficulties lies in the underreporting and underdiagnosis of psychosis, compounded by the stigma surrounding mental health and the individuals' often diminished insight into their condition. Existing efforts leveraging Electronic Health Records (EHRs) to retrospectively identify psychosis typically rely on structured data, such as medical codes and patient demographics, which frequently lack essential information. Addressing these challenges, our study leverages Natural Language Processing (NLP) algorithms to analyze psychiatric admission notes for the diagnosis of psychosis, providing a detailed evaluation of rule-based algorithms, machine learning models, and pre-trained language models. Additionally, the study investigates the effectiveness of employing keywords to streamline extensive note data before training and evaluating the models. Analyzing 4,617 initial psychiatric admission notes (1,196 cases of psychosis versus 3,433 controls) from 2005 to 2019, we discovered that the XGBoost classifier employing Term Frequency-Inverse Document Frequency (TF-IDF) features derived from notes pre-selected by expert-curated keywords, attained the highest performance with an F1 score of 0.8881 (AUROC [95% CI]: 0.9725 [0.9717, 0.9733]). BlueBERT demonstrated comparable efficacy an F1 score of 0.8841 (AUROC [95% CI]: 0.97 [0.9580, 0.9820]) on the same set of notes. Both models markedly outperformed traditional International Classification of Diseases (ICD) code-based detection methods from discharge summaries, which had an F1 score of 0.7608, thus improving the margin by 0.12. Furthermore, our findings indicate that keyword pre-selection markedly enhances the performance of both machine learning and pre-trained language models. This study illustrates the potential of NLP techniques to improve psychosis detection within admission notes and aims to serve as a foundational reference for future research on applying NLP for psychosis identification in EHR notes.

Persistent effects of COVID-19 in patients hospitalized during the first wave of the pandemic: The impact of persistent fatigue on quality of life in a cross-sectional study.

COVID-19 can affect physical and mental health long after acute infection. In this descriptive study, 48 individuals hospitalized for COVID-19 between April and May 2020 were interviewed regarding their experience with COVID-19 after hospitalization. The mean age of participants was 51.1 (±11.91) years (range 25-65 years) and 26 (54.2%) were men. Individuals had a mean of 1.2 (±0.94) comorbidities associated with more severe COVID-19, with hypertension (37.5%) being most common. Nineteen (39.6%) individuals required treatment in the intensive care unit. Participants were interviewed a median time of 553 days (IQR, 405.5-589.0) after discharge from the hospital. Thirty-seven (77.1%) individuals had 5 or more persistent symptoms at time of interview with only 3 (6.3%) experiencing none. The most reported persistent symptoms were fatigue (79.2%), difficulty breathing (68.8%), and muscle weakness (60.4%). Poor quality of life was experienced by 39 (81.3%) participants and 8 (16.7%) had a posttraumatic stress disorder (PTSD) score within the clinical range for diagnosis. For multivariable analyses, persistent fatigue was significantly predicted by number of symptoms during acute COVID-19 (t = 4.4, p < 0.001). Number of symptoms during acute COVID-19 was also significantly associated with persistent dyspnea (t = 3.4, p = 0.002). Higher scores on the Chalder fatigue scale after COVID-19 was significantly associated with poor quality of life (t = 2.6, p = 0.01) and PTSD symptomatology (t = 2.9, p = 0.008). More research is needed to highlight the wide range of resources those suffering from Long COVID require long after discharge.

A System for Characterizing Intraoperative Force Distribution During Operative Laryngoscopy.

OBJECTIVE: This study aimed to create and validate an integrated data acquisition system for gauging the force distribution between a laryngoscope and soft-tissue during trans-oral surgery. METHODS: Sixteen piezoresistive force sensors were interfaced to a laryngoscope and custom maxillary tooth guard. A protocol for calibrating the laryngoscope and maxilla sensors was developed using a motor-controlled linear stage and force measurements were validated against a digital scale. The system was initially tested during suspension laryngoscopy on three cadaver heads mounted on a cadaver head-holder. Intraoperative data was also collected from three patients undergoing head and neck tumor resection. RESULTS: Mean calibration error of the scope sensors was less than 150 g (n = 3) and mean maxilla sensor error was less than 200 g (n = 3). Peak scope mag-forces of 8.09 ± 6.61 kg and peak maxilla forces of 7.62 ± 4.57 kg were experienced during the cadaver trials. The peak scope sensor mag-force recorded during the intraoperative cases was 24.7 ± 4.53 kg, and the peak maxilla force was 22.0 ± 4.60 kg. CONCLUSION: The data acquisition system was successfully able to record intraoperative force distribution data. The usefulness of this technology in informing surgeons during trans-oral surgery should be further evaluated in patients with varying anatomic and procedural characteristics. SIGNIFICANCE: Creation of a low-cost, integrated force-sensing system allows for the characterization of retraction forces at anatomic sites including the pharynx and larynx, brain, and abdomen. Real-time force detection provides surgeons with valuable intraoperative feedback and can be used to improve deformation models at various anatomic sites.

Toward Electrical Impedance Tomography Coupled Ultrasound Imaging for Assessing Muscle Health.

This paper establishes for the first time that a coupled ultrasound (US) and electrical impedance tomography (EIT) system can serve as a non-invasive, spatially localized approach to extract clinically relevant muscle properties. The US/EIT system represents a potential enhancement to electrical impedance myography (EIM), which has shown promise as a non-invasive technology that may have important clinical use in indicating neuromuscular disease status and as a diagnostic tool. A 2.5D EIT algorithm evaluated on simulation, measured phantoms, and measured patient data was studied to evaluate US/EIT's ability to distinguish different aspects of muscle tissue. Simulated and phantom experiments revealed the depths of distinguishability of 3.2 and 4.2 mm in simulation for 10% and 20% changes in muscle properties, respectively, and 3.6 mm in measured phantom experiments assuming a 12% muscle conductivity change. Reconstructions from the patient data established that there were consistent differences 1) between longitudinal (along) and transverse (across) muscle conductivity reconstructions at frequencies of 40 and 80 kHz and 2) side-by-side comparison between healthy and diseased tissue in terms of conductivity, permittivity, and phase at 40 and 80 kHz. Comparisons were made between the EIT reconstructed values and electrical impedance spectroscopy (EIS) measurements (an available surrogate in place of standard EIM measurements) made with the US/EIT system, wherein 1) EIS and EIT show similar sensitivity to longitudinal and transverse differences and 2) EIT showed a more consistent ability to differentiate healthy and diseased tissue. These results suggest that US/EIT appears very promising for non-invasive and spatially localized diagnosis of muscle health.

Signal-to-Noise Ratio Analysis of a Phase-Sensitive Voltmeter for Electrical Impedance Tomography.

In this paper, thorough analysis along with mathematical derivations of the matched filter for a voltmeter used in electrical impedance tomography systems are presented. The effect of the random noise in the system prior to the matched filter, generated by other components, are considered. Employing the presented equations allow system/circuit designers to find the maximum tolerable noise prior to the matched filter that leads to the target signal-to-noise ratio (SNR) of the voltmeter, without having to over-design internal components. A practical model was developed that should fall within 2 dB and 5 dB of the median SNR measurements of signal amplitude and phase, respectively. In order to validate our claims, simulation and experimental measurements have been performed with an analog-to-digital converter (ADC) followed by a digital matched filter, while the noise of the whole system was modeled as the input referred at the ADC input. The input signal was contaminated by a known value of additive white Gaussian noise (AWGN) noise, and the noise level was swept from 3% to 75% of the least significant bit (LSB) of the ADC. Differences between experimental and both simulated and analytical SNR values were less than 0.59 and 0.35 dB for RMS values ≥ 20% of an LSB and less than 1.45 and 2.58 dB for RMS values < 20% of an LSB for the amplitude and phase, respectively. Overall, this study provides a practical model for circuit designers in EIT, and a more accurate error analysis that was previously missing in EIT literature.

Rapid single-molecule imaging in cyclic olefin copolymer channels.

Rapid preparation of high quality capture surfaces is a major challenge for surface-based single-molecule protein binding assays. Here we introduce a simple method to activate microfluidic chambers made from cyclic olefin copolymer for single-molecule imaging with total internal reflection fluorescence microscopy. We describe a surface coating protocol and demonstrate single-molecule imaging in off-the-shelf microfluidic parts that can be activated for binding assays within a few minutes. As the first example, biotinylated protein directly captured on the neutravidin-coated surface was detected using fluorescently labeled antibody. We then showed detection of a fusion construct containing green fluorescence protein and verified its single fluorophore behavior by observing stepwise photobleaching events. Finally, a target protein was identified in the crude cell lysate using antibody-sandwich complex formation. In all experiments, controls were completed to ensure that nonspecific binding to the surface was minimal. Based on our results, we conclude that the simple surface preparation described in this paper enables single-molecule imaging assays without time-consuming coating procedures.

Affinity assisted selection of antibodies for Point of Care TSH immunoassay with limited wash.

BACKGROUND: Molecular binding characteristics of several thyroid stimulating hormone (TSH) antibodies were determined for the TSH antigen, along with its closely related endogenous interfering hormones, follicle stimulating hormone (FSH), luteinizing hormone (LH) and chorionic gonadotropin (CG). METHODS: This data was compared to the same antibodies used in the low wash sandwich ELISA immunoassay system, the Point of Care i-STAT® immunoassay. From this information we developed binding criteria useful in the low wash i-STAT® immunoassay to permit good signal generation from TSH and low cross-reactivity from its interfering hormones. For the TSH Assay we have developed characteristics that enable antibody selection in the i-STAT® immunoassay cartridge. Our antibody screening approach used a dot blot approach as a first screen to select for the most useful antibodies. We then compared a FRET (Förster Resonance Energy Transfer) and electrochemical cartridge approach to determine the appropriate antibody combinations. RESULTS: Both methods generated similar data, but the FRET method was not capable of differentiating the antibody with the best characteristics as a capture antibody or a detection conjugate in a sandwich ELISA assay. Finally, we performed binding characterizations of the antibodies using each of the above mentioned glycoproteins. CONCLUSIONS: We found that we need sub-picomolar detection of TSH, and at least 100 fold or higher values for the cross-reacting species.

Simplified confocal microscope for counting particles at low concentrations.

We describe a compact scanning confocal fluorescence microscope capable of detecting particles concentrations less than 100 particles∕ml in ~15 min. The system mechanically moves a cuvette containing ~3 ml of sample. A relatively large confocal volume is observed within the cuvette using a 1 mm pinhole in front of a detection PMT. Due to the motion of the sample, particles traverse the confocal volume quickly, and analysis by pattern recognition qualifies spikes in the emission intensity data and counts them as events. We show linearity of detection as a function of concentration and also characterize statistical behavior of the instrument. We calculate a detection sensitivity of the system using 3 µm fluorescent microspheres to be 5 particles/ml. Furthermore, to demonstrate biological application, we performed a dilution series to quantify stained E. coli and yeast cells. We counted E. coli cells at a concentration as low as 30 cells∕ml in 10 min/sample.

Studying antibody-antigen interactions with fluorescence fluctuation spectroscopy.

Antibodies are excellent binding proteins that have found numerous applications in biological research, biotechnology, and medicine. Characterization of their ligand binding properties has long been, and continues to be, the focus of many researchers. Antibodies are also perfect test systems which can be used for the evaluation of newly introduced biophysical techniques. Working with many different antibodies, we continuously implement the growing arsenal of methods offered by fluorescence fluctuation spectroscopy (FFS) and apply them for antibody research. In this chapter, we will describe applications of FFS for antibody binding characterization and also provide examples how studying of antibodies helps to develop and enhance the tool set offered by FFS technology. In addition to traditional affinity evaluations, we will describe how resolving molecular populations enables determinations of the binding stoichiometry and provides further information about the system. Even though all our examples include antibodies, the same experimental procedures can also serve well for characterizing various proteins and other ligand binding systems.

Introduction of the mass spread function for characterization of protein conjugates.

Traditionally, characterization of protein molecules conjugated with molecular probes is performed by UV-vis spectroscopy. This method determines the average incorporation ratio but does not yield information about the label distribution. Electrospray ionization mass spectroscopy (ESI-MS) allows direct measurement of the fraction of protein containing a given number of labels. However, for a glycosylated protein, this analysis can be severely limited due to spectral overlap of the labels and carbohydrates. To address this problem, we introduce the mass spread function (MSF) for conjugation analysis. By treating the ESI-MS spectrum of conjugated protein as the spectrum before conjugation convolved with the MSF, we are able to quantify the labeled protein population using a binomial distribution function. We first applied this procedure for characterization of labeled antibody F(ab')(2) fragments which do not contain carbohydrates. We then apply the MSF to fit spectra of entire conjugated monoclonal antibodies and quantify the distribution of labels in the presence of glycans.

Determining antibody stoichiometry using time-integrated fluorescence cumulant analysis.

We applied fluorescence fluctuation spectroscopy to resolve the binding heterogeneity of fluorescently labeled ligand derived from brain natriuretic peptide (BNP), a widely used diagnostic marker of heart failure, to a corresponding monoclonal antibody. This system includes three species: (1) free ligand molecules, (2) antibody with a single site occupied, and (3) antibody with both sites occupied. The method we used, time-integrated fluorescence cumulant analysis (TIFCA), utilizes cumulants of fluorescence fluctuations to resolve subpopulations of multiple fluorescent species freely diffusing in a solution. The values of the cumulants depend on the concentration, molecular brightness and diffusion time of the fluorescent molecules. The number of molecules in each species reflects the antibody affinity. We apply TIFCA to successfully establish the stoichiometry of the system, estimate affinity, and identify the presence of an inactive fraction of antigen in a single titration experiment.

Using nonfluorescent Förster resonance energy transfer acceptors in protein binding studies.

The purpose of this article is to highlight the versatility of nonfluorescent Förster resonance energy transfer (FRET) acceptors in determination of protein equilibrium dissociation constants and kinetic rates. Using a nonfluorescent acceptor eliminates the necessity to spectrally isolate the donor fluorescence when performing binding titrations covering a broad range of reagent concentrations. Moreover, random distribution of the donor and acceptor chromophores on the surface of proteins increases the probability of FRET occurring on their interaction. Three high-affinity antibodies are presented in this study as characteristic protein systems. Monoclonal antibody (mAb) 106.3 binds brain natriuretic peptide (BNP)5-13(C10A) and full-length BNP1-32 with the dissociation constants 0.26+/-0.01 and 0.05+/-0.02 nM, respectively, which was confirmed by kinetic measurements. For anti-hCG (human chorionic gonadotropin) mAb 8F11, studied at two incorporation ratios (IRs=1.9 and 3.8) of the nonfluorescent FRET acceptor, K(D) values of 0.04+/-0.02 and 0.059(-0.004)(+0.006) nM, respectively, were obtained. Likewise, the binding of goat anti-hamster immunoglobulin G (IgG) antibody was not affected by conjugation and yielded K(D) values of 1.26+/-0.04, 1.25+/-0.05, and 1.14+/-0.04 nM at IRs of 1.7, 4.7, and 8.1, respectively. We conclude that this FRET-based method offers high sensitivity, practical simplicity, and versatility in protein binding studies.

The use of vasopressors and inotropes in the emergency medical treatment of shock.

Shock is a final common pathway associated with regularly encountered emergencies including myocardial infarction, microbial sepsis, pulmonary embolism, significant trauma, and anaphylaxis. Shock results in impaired tissue perfusion, cellular hypoxia, and metabolic derangements that cause cellular injury. The clinical manifestations and prognosis of shock are largely dependent on the etiology and duration of insult. It is important that emergency physicians, familiar with the broad differential diagnosis of shock, be prepared to rapidly recognize, resuscitate, and target appropriate therapies aimed at correcting the underlying process. This article focuses on the basic pathophysiology of shock states and reviews the rationale regarding vasoactive drug therapy for cardiovascular support of shock within an emergency environment.

Fluorescence fluctuation spectroscopy in the presence of immobile fluorophores.

Fluorescence contributions from immobile sources present a challenge for fluorescence fluctuation spectroscopy (FFS) because the absence of signal fluctuations from stationary fluorophores leads to a biased analysis. This is especially of concern for cellular FFS studies on proteins that interact with immobile structures. Here we present a method that correctly analyzes FFS experiments in the presence of immobile sources by exploiting selective photobleaching of immobile fluorophores. The fluorescence decay due to photobleaching of the immobile species is modeled taking into account the nonuniform illumination volume. The experimentally observed decay curve serves to separate the mobile and immobile fluorescence contribution, which is used to calculate the molecular brightness from the FFS data. We experimentally verify this approach in vitro using the fluorescent protein EGFP as our immobilized species and a diffusing dye of a different color as the mobile one. For this special case, we also use an alternative method of determining the brightness by spectrally resolving the two species. By conducting a dilution study, we show that the correct parameters are obtained using either technique for a wide range of mobile fractions. To demonstrate the application of our technique in living cells, we perform experiments using the histone core protein H2B fused with EGFP expressed in COS-1 cells. We successfully recovered the brightness of the mobile fraction of H2B-EGFP.

Probing nucleocytoplasmic transport by two-photon activation of PA-GFP.

Two-photon activation of photoactivatable green fluorescent protein (PA-GFP) provides a unique tool for probing cellular transport processes, because activation is strictly limited to the sub-femtoliter optical volume of the two-photon spot. We demonstrate two-photon activation of PA-GFP immobilized in a gel and freely diffusing within cells and recover a quadratic power dependence. Illumination at 820 nm allows simultaneous activation and fluorescence monitoring by two-photon excitation. Alternatively, we activate PA-GFP using two-photon excitation and monitor the fluorescence of the photoconverted product with one-photon excitation. We probe nucleocytoplasmic transport through the nuclear pore complex of COS-1 cells, by observing the time-dependent fluorescence at various locations within the cell after two-photon activation of PA-GFP in the nucleus and in the cytoplasm. Two-photon activation of a tandem construct of two PA-GFPs showed a markedly slower rate of crossing through the nuclear pore. Analysis based on a restricted diffusion model yields a nuclear pore radius of 4.5 nm, which is in good agreement with previously reported values. This application demonstrates the attractive features of two-photon photoactivation over traditional techniques, such as photobleaching, for studying transport processes in cells.

Position-sensitive scanning fluorescence correlation spectroscopy.

Fluorescence correlation spectroscopy (FCS) uses a stationary laser beam to illuminate a small sample volume and analyze the temporal behavior of the fluorescence fluctuations within the stationary observation volume. In contrast, scanning FCS (SFCS) collects the fluorescence signal from a moving observation volume by scanning the laser beam. The fluctuations now contain both temporal and spatial information about the sample. To access the spatial information we synchronize scanning and data acquisition. Synchronization allows us to evaluate correlations for every position along the scanned trajectory. We use a circular scan trajectory in this study. Because the scan radius is constant, the phase angle is sufficient to characterize the position of the beam. We introduce position-sensitive SFCS (PSFCS), where correlations are calculated as a function of lag time and phase. We present the theory of PSFCS and derive expressions for diffusion, diffusion in the presence of flow, and for immobilization. To test PSFCS we compare experimental data with theory. We determine the direction and speed of a flowing dye solution and the position of an immobilized particle. To demonstrate the feasibility of the technique for applications in living cells we present data of enhanced green fluorescent protein measured in the nucleus of COS cells.

Dual-color photon-counting histogram.

We report on the development of dual-color photon-counting histogram (PCH) analysis. Dual-color PCH is an extension of regular PCH and considers the photon counts received in two detection channels instead of one. Because each detection channel records a different color, dual-color PCH distinguishes fluorescent species not only by differences in their brightness, but also according to their color. The additional discrimination by color increases the sensitivity of PCH in resolving a mixture of species considerably. Most dual-color fluorescence fluctuation experiments are performed on fluorophores with overlapping emission spectra. This overlap results in spectral cross talk between the detector channels, which reduces resolvability. Here, we demonstrate that dual-color PCH is able to resolve binary dye mixtures in the presence of cross talk from a single measurement without any additional information about the sample. We discuss the effect of sampling time on the fit parameters of dual-color PCH. Differences between dual-color fluorescence correlation spectroscopy and dual-color PCH will also be addressed. We quantitatively resolve a mixture of the two fluorescent proteins CFP and YFP, which is challenging because of the strong spectral overlap of their emission spectra. Dichroic mirrors are needed to direct the light into the two detection channels. We quantify the influence of these filters on dual-color PCH analysis and determine the optimal transition wavelength of the dichroic mirror for the CFP-YFP pair.