Glioma subtype classification from histopathological images using in-domain and out-of-domain transfer learning: An experimental study.

We provide in this paper a comprehensive comparison of various transfer learning strategies and deep learning architectures for computer-aided classification of adult-type diffuse gliomas. We evaluate the generalizability of out-of-domain ImageNet representations for a target domain of histopathological images, and study the impact of in-domain adaptation using self-supervised and multi-task learning approaches for pretraining the models using the medium-to-large scale datasets of histopathological images. A semi-supervised learning approach is furthermore proposed, where the fine-tuned models are utilized to predict the labels of unannotated regions of the whole slide images (WSI). The models are subsequently retrained using the ground-truth labels and weak labels determined in the previous step, providing superior performance in comparison to standard in-domain transfer learning with balanced accuracy of 96.91% and F1-score 97.07%, and minimizing the pathologist's efforts for annotation. Finally, we provide a visualization tool working at WSI level which generates heatmaps that highlight tumor areas; thus, providing insights to pathologists concerning the most informative parts of the WSI.

Quantitative SARS-CoV-2 RT-PCR and Bronchoalveolar Cytokine Concentrations Redefine the COVID-19 Phenotypes in Critically Ill Patients.

RATIONALE: Recent studies suggest that both hypo- and hyperinflammatory acute respiratory distress syndrome (ARDS) phenotypes characterize severe COVID-19-related pneumonia. The role of lung Severe Acute Respiratory Syndrome - Coronavirus 2 (SARS-CoV-2) viral load in contributing to these phenotypes remains unknown. OBJECTIVES: To redefine COVID-19 ARDS phenotypes when considering quantitative SARS-CoV-2 RT-PCR in the bronchoalveolar lavage of intubated patients. To compare the relevance of deep respiratory samples versus plasma in linking the immune response and the quantitative viral loads. METHODS: Eligible subjects were adults diagnosed with COVID-19 ARDS who required mechanical ventilation and underwent bronchoscopy. We recorded the immune response in the bronchoalveolar lavage and plasma and the quantitative SARS-CoV-2 RT-PCR in the bronchoalveolar lavage. Hierarchical clustering on principal components was applied separately on the 2 compartments' datasets. Baseline characteristics were compared between clusters. MEASUREMENTS AND RESULTS: Twenty subjects were enrolled between August 2020 and March 2021. Subjects underwent bronchoscopy on average 3.6 days after intubation. All subjects were treated with dexamethasone prior to bronchoscopy, 11 of 20 (55.6%) received remdesivir and 1 of 20 (5%) received tocilizumab. Adding viral load information to the classic 2-cluster model of ARDS revealed a new cluster characterized by hypoinflammatory responses and high viral load in 23.1% of the cohort. Hyperinflammatory ARDS was noted in 15.4% of subjects. Bronchoalveolar lavage clusters were more stable compared to plasma. CONCLUSIONS: We identified a unique group of critically ill subjects with COVID-19 ARDS who exhibit hypoinflammatory responses but high viral loads in the lower airways. These clusters may warrant different treatment approaches to improve clinical outcomes.

Telemedicine Critical Care-Mediated Mortality Reductions in Lower-Performing Patient Diagnosis Groups: A Prospective, Before and After Study.

OBJECTIVES: Studies evaluating telemedicine critical care (TCC) have shown mixed results. We prospectively evaluated the impact of TCC implementation on risk-adjusted mortality among patients stratified by pre-TCC performance. DESIGN: Prospective, observational, before and after study. SETTING: Three adult ICUs at an academic medical center. PATIENTS: A total of 2,429 patients in the pre-TCC (January to June 2016) and 12,479 patients in the post-TCC (January 2017 to June 2019) periods. INTERVENTIONS: TCC implementation which included an acuity-driven workflow targeting an identified "lower-performing" patient group, defined by ICU admission in an Acute Physiology and Chronic Health Evaluation diagnoses category with a pre-TCC standardized mortality ratio (SMR) of greater than 1.5. MEASUREMENTS AND MAIN RESULTS: The primary outcome was risk-adjusted hospital mortality. Risk-adjusted hospital length of stay (HLOS) was also studied. The SMR for the overall ICU population was 0.83 pre-TCC and 0.75 post-TCC, with risk-adjusted mortalities of 10.7% and 9.5% (p = 0.09). In the identified lower-performing patient group, which accounted for 12.6% (n = 307) of pre-TCC and 13.3% (n = 1671) of post-TCC ICU patients, SMR decreased from 1.61 (95% CI, 1.21-2.01) pre-TCC to 1.03 (95% CI, 0.91-1.15) post-TCC, and risk-adjusted mortality decreased from 26.4% to 16.9% (p < 0.001). In the remaining ("higher-performing") patient group, there was no change in pre- versus post-TCC SMR (0.70 [0.59-0.81] vs 0.69 [0.64-0.73]) or risk-adjusted mortality (8.5% vs 8.4%, p = 0.86). There were no pre- to post-TCC differences in standardized HLOS ratio or risk-adjusted HLOS in the overall cohort or either performance group. CONCLUSIONS: In well-staffed and overall higher-performing ICUs in an academic medical center, Acute Physiology and Chronic Health Evaluation granularity allowed identification of a historically lower-performing patient group that experienced a striking TCC-associated reduction in SMR and risk-adjusted mortality. This study provides additional evidence for the relationship between pre-TCC performance and post-TCC improvement.

Use of Inhaled Epoprostenol in Patients With COVID-19 Receiving Humidified, High-Flow Nasal Oxygen Is Associated With Progressive Respiratory Failure.

BACKGROUND: The clinical benefit of using inhaled epoprostenol (iEpo) through a humidified high-flow nasal cannula (HHFNC) remains unknown for patients with COVID-19. RESEARCH QUESTION: Can iEpo prevent respiratory deterioration for patients with positive SARS-CoV-2 findings receiving HHFNC? STUDY DESIGN AND METHODS: This multicenter retrospective cohort analysis included patients aged 18 years or older with COVID-19 pneumonia who required HHFNC treatment. Patients who received iEpo were propensity score matched to patients who did not receive iEpo. The primary outcome was time to mechanical ventilation or death without mechanical ventilation and was assessed using Kaplan-Meier curves and Cox proportional hazard ratios. The effects of residual confounding were assessed using a multilevel analysis, and a secondary analysis adjusted for outcome propensity also was performed in a multivariable model that included the entire (unmatched) patient cohort. RESULTS: Among 954 patients with positive SARS-CoV-2 findings receiving HHFNC therapy, 133 patients (13.9%) received iEpo. After propensity score matching, the median number of days until the composite outcome was similar between treatment groups (iEpo: 5.0 days [interquartile range, 2.0-10.0 days] vs no-iEpo: 6.5 days [interquartile range, 2.0-11.0 days]; P = .26), but patients who received iEpo were more likely to meet the composite outcome in the propensity score-matched, multilevel, and multivariable unmatched analyses (hazard ratio, 2.08 [95% CI, 1.73-2.50]; OR, 4.72 [95% CI, 3.01-7.41]; and OR, 1.35 [95% CI, 1.23-1.49]; respectively). INTERPRETATION: In patients with COVID-19 receiving HHFNC therapy, use of iEpo was associated with the need for invasive mechanical ventilation.

Vocal biomarker predicts fatigue in people with COVID-19: results from the prospective Predi-COVID cohort study.

OBJECTIVE: To develop a vocal biomarker for fatigue monitoring in people with COVID-19. DESIGN: Prospective cohort study. SETTING: Predi-COVID data between May 2020 and May 2021. PARTICIPANTS: A total of 1772 voice recordings were used to train an AI-based algorithm to predict fatigue, stratified by gender and smartphone's operating system (Android/iOS). The recordings were collected from 296 participants tracked for 2 weeks following SARS-CoV-2 infection. PRIMARY AND SECONDARY OUTCOME MEASURES: Four machine learning algorithms (logistic regression, k-nearest neighbours, support vector machine and soft voting classifier) were used to train and derive the fatigue vocal biomarker. The models were evaluated based on the following metrics: area under the curve (AUC), accuracy, F1-score, precision and recall. The Brier score was also used to evaluate the models' calibrations. RESULTS: The final study population included 56% of women and had a mean (±SD) age of 40 (±13) years. Women were more likely to report fatigue (p<0.001). We developed four models for Android female, Android male, iOS female and iOS male users with a weighted AUC of 86%, 82%, 79%, 85% and a mean Brier Score of 0.15, 0.12, 0.17, 0.12, respectively. The vocal biomarker derived from the prediction models successfully discriminated COVID-19 participants with and without fatigue. CONCLUSIONS: This study demonstrates the feasibility of identifying and remotely monitoring fatigue thanks to voice. Vocal biomarkers, digitally integrated into telemedicine technologies, are expected to improve the monitoring of people with COVID-19 or Long-COVID. TRIAL REGISTRATION NUMBER: NCT04380987.

Audio-based Active and Assisted Living: A review of selected applications and future trends.

The development of big data, machine learning, and the Internet of Things has led to rapid advances in the research field of Active and Assisted Living (AAL). A human is placed in the center of such an environment, interacting with different modalities while using the system. Although video still plays a dominant role in AAL technologies, audio, as the most natural means of interaction, is also used commonly, either as a single source of information, or in combination with other modalities. Despite the rapidly increased research efforts in the last decade, there is a lack of systematic overview of audio based technologies and applications in AAL. This review tries to fill this gap, and identifies five major topics where audio is an essential AAL building block: Physiological monitoring, emotion recognition in the context of AAL, human activity recognition, fall detection, and food intake monitoring. We address the data work flow and standard sensing technologies for capturing audio in the AAL environment, provide a comprehensive overview of audio-based AAL applications, and identify datasets available to the research community. Finally, we address the main challenges that should be handled in the upcoming years, and try to identify the potential future trends in audio-based AAL.

A voice-based biomarker for monitoring symptom resolution in adults with COVID-19: Findings from the prospective Predi-COVID cohort study.

People with COVID-19 can experience impairing symptoms that require enhanced surveillance. Our objective was to train an artificial intelligence-based model to predict the presence of COVID-19 symptoms and derive a digital vocal biomarker for easily and quantitatively monitoring symptom resolution. We used data from 272 participants in the prospective Predi-COVID cohort study recruited between May 2020 and May 2021. A total of 6473 voice features were derived from recordings of participants reading a standardized pre-specified text. Models were trained separately for Android devices and iOS devices. A binary outcome (symptomatic versus asymptomatic) was considered, based on a list of 14 frequent COVID-19 related symptoms. A total of 1775 audio recordings were analyzed (6.5 recordings per participant on average), including 1049 corresponding to symptomatic cases and 726 to asymptomatic ones. The best performances were obtained from Support Vector Machine models for both audio formats. We observed an elevated predictive capacity for both Android (AUC = 0.92, balanced accuracy = 0.83) and iOS (AUC = 0.85, balanced accuracy = 0.77) as well as low Brier scores (0.11 and 0.16 respectively for Android and iOS when assessing calibration. The vocal biomarker derived from the predictive models accurately discriminated asymptomatic from symptomatic individuals with COVID-19 (t-test P-values<0.001). In this prospective cohort study, we have demonstrated that using a simple, reproducible task of reading a standardized pre-specified text of 25 seconds enabled us to derive a vocal biomarker for monitoring the resolution of COVID-19 related symptoms with high accuracy and calibration.

Detection of COVID-19 from voice, cough and breathing patterns: Dataset and preliminary results.

COVID-19 heavily affects breathing and voice and causes symptoms that make patients' voices distinctive, creating recognizable audio signatures. Initial studies have already suggested the potential of using voice as a screening solution. In this article we present a dataset of voice, cough and breathing audio recordings collected from individuals infected by SARS-CoV-2 virus, as well as non-infected subjects via large scale crowdsourced campaign. We describe preliminary results for detection of COVID-19 from cough patterns using standard acoustic features sets, wavelet scattering features and deep audio embeddings extracted from low-level feature representations (VGGish and OpenL3). Our models achieve accuracy of 88.52%, sensitivity of 88.75% and specificity of 90.87%, confirming the applicability of audio signatures to identify COVID-19 symptoms. We furthermore provide an in-depth analysis of the most informative acoustic features and try to elucidate the mechanisms that alter the acoustic characteristics of coughs of people with COVID-19.

Design of a 2-Bit Neural Network Quantizer for Laplacian Source.

Achieving real-time inference is one of the major issues in contemporary neural network applications, as complex algorithms are frequently being deployed to mobile devices that have constrained storage and computing power. Moving from a full-precision neural network model to a lower representation by applying quantization techniques is a popular approach to facilitate this issue. Here, we analyze in detail and design a 2-bit uniform quantization model for Laplacian source due to its significance in terms of implementation simplicity, which further leads to a shorter processing time and faster inference. The results show that it is possible to achieve high classification accuracy (more than 96% in the case of MLP and more than 98% in the case of CNN) by implementing the proposed model, which is competitive to the performance of the other quantization solutions with almost optimal precision.

Voice for Health: The Use of Vocal Biomarkers from Research to Clinical Practice.

Diseases can affect organs such as the heart, lungs, brain, muscles, or vocal folds, which can then alter an individual's voice. Therefore, voice analysis using artificial intelligence opens new opportunities for healthcare. From using vocal biomarkers for diagnosis, risk prediction, and remote monitoring of various clinical outcomes and symptoms, we offer in this review an overview of the various applications of voice for health-related purposes. We discuss the potential of this rapidly evolving environment from a research, patient, and clinical perspective. We also discuss the key challenges to overcome in the near future for a substantial and efficient use of voice in healthcare.

Inter-Rater Reliability and Impact of Disagreements on Acute Physiology and Chronic Health Evaluation IV Mortality Predictions.

Acute Physiology and Chronic Health Evaluation is a well-validated method to risk-adjust ICU patient outcomes. However, predictions may be affected by inter-rater reliability for manually entered elements. We evaluated inter-rater reliability for Acute Physiology and Chronic Health Evaluation IV manually entered elements among clinician abstractors and assessed the impacts of disagreements on mortality predictions. DESIGN: Cross-sectional. SETTING: Academic medical center. SUBJECTS: Patients admitted to five adult ICUs. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: Acute Physiology and Chronic Health Evaluation IV manually entered elements were abstracted from a selection of charts (n = 41) by two clinician "raters" trained in Acute Physiology and Chronic Health Evaluation IV methodology. Rater agreement (%) was determined for each manually entered element, including Acute Physiology and Chronic Health Evaluation diagnosis, Glasgow Coma Scale score, admission source, chronic conditions, elective/emergency surgery, and ventilator use. Cohen's kappa (K) or intraclass correlation coefficient was calculated for nominal and continuous manually entered elements, respectively. The impacts of manually entered element choices on Acute Physiology and Chronic Health Evaluation IV mortality predictions were computed using published Acute Physiology and Chronic Health Evaluation IV equations, and observed to expected hospital mortality ratios were compared between rater groups. The majority of manually entered element inconsistency was due to disagreement in choice of Glasgow Coma Scale (63.8% agreement, 0.83 intraclass correlation coefficient), Acute Physiology and Chronic Health Evaluation diagnosis (68.3% agreement, 0.67 kappa), and admission source (90.2% agreement, 0.85 kappa). The difference in predicted mortality between raters related to Glasgow Coma Scale disagreements was significant (observed to expected mortality ratios for Rater 1 [1.009] vs Rater 2 [1.134]; p < 0.05). Differences related to Acute Physiology and Chronic Health Evaluation diagnosis or admission source disagreements were negligible. The new "unable to score" choice for Glasgow Coma Scale was used for 18% of Glasgow Coma Scale measurements but accounted for 63% of "major" Glasgow Coma Scale disagreements, and 50% of the overall difference in Acute Physiology and Chronic Health Evaluation-predicted mortality between raters. CONCLUSIONS: Inconsistent use among raters of the new "unable to score" choice for Glasgow Coma Scale introduced in Acute Physiology and Chronic Health Evaluation IV was responsible for important decreases in both Glasgow Coma Scale and Acute Physiology and Chronic Health Evaluation IV mortality prediction reliability in our study. A Glasgow Coma Scale algorithm we developed after the study to improve reliability related to use of this new "unable to score" choice is presented.

Typhoid Fever Complicated by Hemophagocytic Lymphohistiocytosis and Rhabdomyolysis.

Hemophagocytic lymphohistiocytosis (HLH) and rhabdomyolysis are rare complications of typhoid fever from Salmonella enterica serovar Typhi. Herein, we describe the clinical features in a 21-year-old female from India who presented to the intensive care unit with fever, severe pancytopenia, and rhabdomyolysis.

A 43-year-old man with antisynthetase syndrome presenting with acute worsening of dyspnea.

A 43-year-old man with antisynthetase syndrome was seen in our pulmonary clinic for worsening dyspnea. He was recently diagnosed with antisynthetase syndrome because he had nonspecific interstitial pneumonitis on a surgical lung biopsy and polymyositis associated with anti-Jo-1 and anti-SSA-52 autoantibodies. Along with his worsening dyspnea, he also had a dry cough, lower extremity edema, and abdominal distension. He had gained 11 kg over 1 month. He had been taking prednisone 40 mg daily 2 months prior, which had been recently weaned to 20 mg daily. He had also been on mycophenolate mofetil but had recently discontinued it on his own.