Scalable deep learning framework for sentiment analysis prediction for online movie reviews.

Sentiment analysis has broad use in diverse real-world contexts, particularly in the online movie industry and other e-commerce platforms. The main objective of our work is to examine the word information order and analyze the content of texts by exploring the hidden meanings of words in online movie text reviews. This study presents an enhanced method of representing text and computationally feasible deep learning models, namely the PEW-MCAB model. The methodology categorizes sentiments by considering the full written text as a unified piece. The feature vector representation is processed using an enhanced text representation called Positional embedding and pretrained Glove Embedding Vector (PEW). The learning of these features is achieved by inculcating a multichannel convolutional neural network (MCNN), which is subsequently integrated into an Attention-based Bidirectional Long Short-Term Memory (AB) model. This experiment examines the positive and negative of online movie textual reviews. Four datasets were used to evaluate the model. When tested on the IMDB, MR (2002), MRC (2004), and MR (2005) datasets, the (PEW-MCAB) algorithm attained accuracy rates of 90.3%, 84.1%, 85.9%, and 87.1%, respectively, in the experimental setting. When implemented in practical settings, the proposed structure shows a great deal of promise for efficacy and competitiveness.

Association Between Self-Reported Medication Adherence and Therapeutic Inertia in Hypertension: A Secondary Analysis of SPRINT (Systolic Blood Pressure Intervention Trial).

BACKGROUND: Therapeutic inertia (TI), failure to intensify antihypertensive medication when blood pressure (BP) is above goal, remains prevalent in hypertension management. The degree to which self-reported antihypertensive adherence is associated with TI with intensive BP goals remains unclear. METHODS AND RESULTS: Cross-sectional analysis was performed of the 12-month visit of participants in the intensive arm of SPRINT (Systolic Blood Pressure Intervention Trial), which randomized adults to intensive (<120 mm Hg) versus standard (<140 mm Hg) systolic BP goals. TI was defined as no increase in antihypertensive regimen intensity score, which incorporates medication number and dose, when systolic BP is ≥120 mm Hg. Self-reported adherence was assessed using the 8-Item Morisky Medication Adherence Scale (MMAS-8) and categorized as low (MMAS-8 score <6), medium (MMAS-8 score 6 to <8), and high (MMAS-8 score 8). Poisson regressions estimated prevalence ratios (PRs) and 95% CIs for TI associated with MMAS-8. Among 1009 intensive arm participants with systolic BP >120 mm Hg at the 12-month visit (mean age, 69.6 years; 35.2% female, 28.8% non-Hispanic Black), TI occurred in 50.8% of participants. Participants with low adherence (versus high) were younger and more likely to be non-Hispanic Black or smokers. The prevalence of TI among patients with low, medium, and high adherence was 45.0%, 53.5%, and 50.4%, respectively. After adjustment, neither low nor medium adherence (versus high) were associated with TI (PR, 1.11 [95% CI, 0.87-1.42]; PR, 1.08 [95% CI, 0.84-1.38], respectively). CONCLUSIONS: Although clinician uncertainty about adherence is often cited as a reason for why antihypertensive intensification is withheld when above BP goals, we observed no evidence of an association between self-reported adherence and TI.

Evidence for Age Bias Contributing to Therapeutic Inertia in Blood Pressure Management: A Secondary Analysis of SPRINT.

BACKGROUND: Despite evidence supporting the cardiovascular and cognitive benefits of intensive blood pressure management, older adults have the lowest rates of blood pressure control. We determined the association between age and therapeutic inertia (TI) in SPRINT (Systolic Blood Pressure Intervention Trial), and whether frailty, cognitive function, or gait speed moderate or mediate these associations. METHODS: We performed a secondary analysis of SPRINT of participant visits with blood pressure above randomized treatment goal. We categorized baseline age as <60, 60 to <70, 70 to <80, and ≥80 years and TI as no antihypertensive medication intensification per participant visit. Generalized estimating equations generated odds ratios for TI associated with age, stratified by treatment group based on nested models adjusted for baseline frailty index score (fit [frailty index, ≤0.10], less fit [0.100.10). CONCLUSIONS: Older age is associated with greater TI independent of physical or cognitive function, implying age bias in hypertension management.

Prevalence of Statin Use for Primary Prevention of Atherosclerotic Cardiovascular Disease by Race, Ethnicity, and 10-Year Disease Risk in the US: National Health and Nutrition Examination Surveys, 2013 to March 2020.

Importance: The burden of atherosclerotic cardiovascular disease (ASCVD) in the US is higher among Black and Hispanic vs White adults. Inclusion of race in guidance for statin indication may lead to decreased disparities in statin use. Objective: To evaluate prevalence of primary prevention statin use by race and ethnicity according to 10-year ASCVD risk. Design, Setting, and Participants: This serial, cross-sectional analysis performed in May 2022 used data from the National Health and Nutrition Examination Survey, a nationally representative sample of health status in the US, from 2013 to March 2020 (limited cycle due to the COVID-19 pandemic), to evaluate statin use for primary prevention of ASCVD and to estimate 10-year ASCVD risk. Participants aged 40 to 75 years without ASCVD, diabetes, low-density lipoprotein cholesterol levels 190 mg/dL or greater, and with data on medication use were included. Exposures: Self-identified race and ethnicity (Asian, Black, Hispanic, and White) and 10-year ASCVD risk category (5%-<7.5%, 7.5%-<20%, ≥20%). Main Outcomes and Measures: Prevalence of statin use, defined as identification of statin use on pill bottle review. Results: A total of 3417 participants representing 39.4 million US adults after applying sampling weights (mean [SD] age, 61.8 [8.0] years; 1289 women [weighted percentage, 37.8%] and 2128 men [weighted percentage, 62.2%]; 329 Asian [weighted percentage, 4.2%], 1032 Black [weighted percentage, 12.7%], 786 Hispanic [weighted percentage, 10.1%], and 1270 White [weighted percentage, 73.0%]) were included. Compared with White participants, statin use was lower in Black and Hispanic participants and comparable among Asian participants in the overall cohort (Asian, 25.5%; Black, 20.0%; Hispanic, 15.4%; White, 27.9%) and within ASCVD risk strata. Within each race and ethnicity group, a graded increase in statin use was observed across increasing ASCVD risk strata. Statin use was low in the highest risk stratum overall with significantly lower rates of use among Black (23.8%; prevalence ratio [PR], 0.90; 95% CI, 0.82-0.98 vs White) and Hispanic participants (23.9%; PR, 0.90; 95% CI, 0.81-0.99 vs White). Among other factors, routine health care access and health insurance were significantly associated with higher statin use in Black, Hispanic, and White adults. Prevalence of statin use did not meaningfully change over time by race and ethnicity or by ASCVD risk stratum. Conclusions and Relevance: In this study, statin use for primary prevention of ASCVD was low among all race and ethnicity groups regardless of ASCVD risk, with the lowest use occurring among Black and Hispanic adults. Improvements in access to care may promote equitable use of primary prevention statins in Black and Hispanic adults.

A hybrid explainable ensemble transformer encoder for pneumonia identification from chest X-ray images.

INTRODUCTION: Pneumonia is a microorganism infection that causes chronic inflammation of the human lung cells. Chest X-ray imaging is the most well-known screening approach used for detecting pneumonia in the early stages. While chest-Xray images are mostly blurry with low illumination, a strong feature extraction approach is required for promising identification performance. OBJECTIVES: A new hybrid explainable deep learning framework is proposed for accurate pneumonia disease identification using chest X-ray images. METHODS: The proposed hybrid workflow is developed by fusing the capabilities of both ensemble convolutional networks and the Transformer Encoder mechanism. The ensemble learning backbone is used to extract strong features from the raw input X-ray images in two different scenarios: ensemble A (i.e., DenseNet201, VGG16, and GoogleNet) and ensemble B (i.e., DenseNet201, InceptionResNetV2, and Xception). Whereas, the Transformer Encoder is built based on the self-attention mechanism with multilayer perceptron (MLP) for accurate disease identification. The visual explainable saliency maps are derived to emphasize the crucial predicted regions on the input X-ray images. The end-to-end training process of the proposed deep learning models over all scenarios is performed for binary and multi-class classification scenarios. RESULTS: The proposed hybrid deep learning model recorded 99.21% classification performance in terms of overall accuracy and F1-score for the binary classification task, while it achieved 98.19% accuracy and 97.29% F1-score for multi-classification task. For the ensemble binary identification scenario, ensemble A recorded 97.22% accuracy and 97.14% F1-score, while ensemble B achieved 96.44% for both accuracy and F1-score. For the ensemble multiclass identification scenario, ensemble A recorded 97.2% accuracy and 95.8% F1-score, while ensemble B recorded 96.4% accuracy and 94.9% F1-score. CONCLUSION: The proposed hybrid deep learning framework could provide promising and encouraging explainable identification performance comparing with the individual, ensemble models, or even the latest AI models in the literature. The code is available here: https://github.com/chiagoziemchima/Pneumonia_Identificaton.

EVAE-Net: An Ensemble Variational Autoencoder Deep Learning Network for COVID-19 Classification Based on Chest X-ray Images.

The COVID-19 pandemic has had a significant impact on many lives and the economies of many countries since late December 2019. Early detection with high accuracy is essential to help break the chain of transmission. Several radiological methodologies, such as CT scan and chest X-ray, have been employed in diagnosing and monitoring COVID-19 disease. Still, these methodologies are time-consuming and require trial and error. Machine learning techniques are currently being applied by several studies to deal with COVID-19. This study exploits the latent embeddings of variational autoencoders combined with ensemble techniques to propose three effective EVAE-Net models to detect COVID-19 disease. Two encoders are trained on chest X-ray images to generate two feature maps. The feature maps are concatenated and passed to either a combined or individual reparameterization phase to generate latent embeddings by sampling from a distribution. The latent embeddings are concatenated and passed to a classification head for classification. The COVID-19 Radiography Dataset from Kaggle is the source of chest X-ray images. The performances of the three models are evaluated. The proposed model shows satisfactory performance, with the best model achieving 99.19% and 98.66% accuracy on four classes and three classes, respectively.

Characterization of GafChromic EBT2 film dose measurements using a tissue-equivalent water phantom for a Theratron® Equinox Cobalt-60 teletherapy machine.

PURPOSE: In vivo dosimetry is a quality assurance tool that provides post-treatment measurement of the absorbed dose as delivered to the patient. This dosimetry compares the prescribed and measured dose delivered to the target volume. In this study, a tissue-equivalent water phantom provided the simulation of the human environment. The skin and entrance doses were measured using GafChromic EBT2 film for a Theratron® Equinox Cobalt-60 teletherapy machine. METHODS: We examined the behaviors of unencapsulated films and custom-made film encapsulation. Films were cut to 1 cm × 1 cm, calibrated, and used to assess skin dose depositions and entrance dose. We examined the response of the film for variations in field size, source to skin distance (SSD), gantry angle and wedge angle. RESULTS: The estimated uncertainty in EBT2 film for absorbed dose measurement in phantom was ±1.72%. Comparison of the measurements of the two film configurations for the various irradiation parameters were field size (p = 0.0193, α = 0.05, n = 11), gantry angle (p = 0.0018, α = 0.05, n = 24), SSD (p = 0.1802, α = 0.05, n = 11) and wedge angle (p = 0.6834, α = 0.05, n = 4). For a prescribed dose of 200 cGy and at reference conditions (open field 10 cm x 10 cm, SSD = 100 cm, and gantry angle = 0º), the measured skin dose using the encapsulation material was 70% while that measured with the unencapsulated film was 24%. At reference irradiation conditions, the measured skin dose using the unencapsulated film was higher for open field configurations (24%) than wedged field configurations (19%). Estimation of the entrance dose using the unencapsulated film was within 3% of the prescribed dose. CONCLUSIONS: GafChromic EBT2 film measurements were significantly affected at larger field sizes and gantry angles. Furthermore, we determined a high accuracy in entrance dose estimations using the film.

Deep inverse reinforcement learning for structural evolution of small molecules.

The size and quality of chemical libraries to the drug discovery pipeline are crucial for developing new drugs or repurposing existing drugs. Existing techniques such as combinatorial organic synthesis and high-throughput screening usually make the process extraordinarily tough and complicated since the search space of synthetically feasible drugs is exorbitantly huge. While reinforcement learning has been mostly exploited in the literature for generating novel compounds, the requirement of designing a reward function that succinctly represents the learning objective could prove daunting in certain complex domains. Generative adversarial network-based methods also mostly discard the discriminator after training and could be hard to train. In this study, we propose a framework for training a compound generator and learn a transferable reward function based on the entropy maximization inverse reinforcement learning (IRL) paradigm. We show from our experiments that the IRL route offers a rational alternative for generating chemical compounds in domains where reward function engineering may be less appealing or impossible while data exhibiting the desired objective is readily available.

Optimizing human pulmonary perfusion measurement using an in silico model of arterial spin labeling magnetic resonance imaging.

Arterial spin labeling (ASL) magnetic resonance imaging (MRI) is an imaging methodology that uses blood as an endogenous contrast agent to quantify flow. One limitation of this method of capillary blood quantification when applied in the lung is the contribution of signals from non-capillary blood. Intensity thresholding is one approach that has been proposed for minimizing the non-capillary blood signal. This method has been tested in previous in silico modeling studies; however, it has only been tested under a restricted set of physiological conditions (supine posture and a cardiac output of 5 L/min). This study presents an in silico approach that extends previous intensity thresholding analysis to estimate the optimal "per-slice" intensity threshold value using the individual components of the simulated ASL signal (signal arising independently from capillary blood as well as pulmonary arterial and pulmonary venous blood). The aim of this study was to assess whether the threshold value should vary with slice location, posture, or cardiac output. We applied an in silico modeling approach to predict the blood flow distribution and the corresponding ASL quantification of pulmonary perfusion in multiple sagittal imaging slices. There was a significant increase in ASL signal and heterogeneity (COV = 0.90 to COV = 1.65) of ASL signals when slice location changed from lateral to medial. Heterogeneity of the ASL signal within a slice was significantly lower (P = 0.03) in prone (COV = 1.08) compared to in the supine posture (COV = 1.17). Increasing stroke volume resulted in an increase in ASL signal and conversely an increase in heart rate resulted in a decrease in ASL signal. However, when cardiac output was increased via an increase in both stroke volume and heart rate, ASL signal remained relatively constant. Despite these differences, we conclude that a threshold value of 35% provides optimal removal of large vessel signal independent of slice location, posture, and cardiac output.