Gene replacement therapy in Bietti crystalline corneoretinal dystrophy: an open-label, single-arm, exploratory trial.

Bietti crystalline corneoretinal dystrophy is an inherited retinal disease caused by mutations in CYP4V2, which results in blindness in the working-age population, and there is currently no available treatment. Here, we report the results of the first-in-human clinical trial (NCT04722107) of gene therapy for Bietti crystalline corneoretinal dystrophy, including 12 participants who were followed up for 180-365 days. This open-label, single-arm exploratory trial aimed to assess the safety and efficacy of a recombinant adeno-associated-virus-serotype-2/8 vector encoding the human CYP4V2 protein (rAAV2/8-hCYP4V2). Participants received a single unilateral subretinal injection of 7.5 × 1010 vector genomes of rAAV2/8-hCYP4V2. Overall, 73 treatment-emergent adverse events were reported, with the majority (98.6%) being of mild or moderate intensity and considered to be procedure- or corticosteroid-related; no treatment-related serious adverse events or local/systemic immune toxicities were observed. Compared with that measured at baseline, 77.8% of the treated eyes showed improvement in best-corrected visual acuity (BCVA) on day 180, with a mean ± standard deviation increase of 9.0 ± 10.8 letters in the 9 eyes analyzed (p = 0.021). By day 365, 80% of the treated eyes showed an increase in BCVA, with a mean increase of 11.0 ± 10.6 letters in the 5 eyes assessed (p = 0.125). Importantly, the patients' improvement observed using multifocal electroretinogram, microperimetry, and Visual Function Questionnaire-25 further supported the beneficial effects of the treatment. We conclude that the favorable safety profile and visual improvements identified in this trial encourage the continued development of rAAV2/8-hCYP4V2 (named ZVS101e).

Exploring the Diagnostic Potential of EPB41L3 Methylation in Cervical Cancer and Precancerous Lesions: A Systematic Review and Meta-Analysis.

OBJECTIVE: This meta-analysis aimed to comprehensively evaluate the diagnostic use of erythrocyte membrane protein band 4.1like3 (EPB41L3) methylation detection in cervical cancer (CC) and its precancerous lesions. METHODS: CNKI, Wanfang, Cochrane Library, PubMed, and Ovid databases were searched using a combination of subject headings and free words. Pertinent data were retrieved after screening for inclusion and exclusion criteria, and the quality of the included studies was evaluated using QUADAS-2 criteria. The appropriate software was used for heterogeneity analysis and combined effect size calculation. Additionally, sensitivity analysis was used to evaluate the robustness of the combined results, and meta-regression and subgroup analysis were conducted to investigate the origins of heterogeneity. RESULTS: This meta-analysis included six studies, including 525 healthy individuals, 182 cervical intraepithelial neoplasia 1 (CIN1) samples, 182 CIN2 samples, 281 CIN3 samples, and 226 CC samples. EPB41L3 methylation detection for CIN2 and above lesions demonstrated combined sensitivity, specificity, positive likelihood ratio, negative likelihood ratio, diagnostic odds ratio (DOR), and the area under the curve of the comprehensive receiver operating characteristic curve of 0.67, 0.76, 3.19, 0.41, 7.60, and 0.80, respectively; CIN3 and above lesions demonstrated these evaluations at 0.73, 0.84, 4.35, 0.33, 23.94, and 0.90, respectively. Meta-regression analysis revealed that the population, time, sample type, detection method, literature quality, and sample size were not significant sources of heterogeneity affecting the combined diagnostic efficacy of CIN2 and above lesions (p > 0.05). Subgroup analysis revealed higher combined diagnostic values of CIN2 and above lesions in retrospective studies, tissue samples, and Chinese populations, with DORs of 41.03, 14.59, and 13.70, respectively. CONCLUSION: EPB41L3 methylation demonstrated a relatively low diagnostic performance in CC and precancerous lesions. However, it merits further investigation as a potential biomarker. Integrating it with multiple gene detection, human papillomavirus testing, and ThinPrep liquid-based cytology test examination is recommended to explore improved diagnostic strategies for CC and its precancerous lesions.

Development of an Immediate-Release Prototype Tablet Formulation of Hydroxychloroquine Sulfate with an Interwoven Taste-Masking System.

Hydroxychloroquine sulfate (HCQS) was granted US-FDA approval in 1955 for the prevention and treatment of malaria. Since then, its uses have expanded to treat systemic lupus erythematosus and rheumatoid arthritis.  For each indication, HCQS is a crucial option for the treatment of pediatric, juvenile, adult, and elderly populations. Existing currently on the market are only 200-mg strength tablets exclusively for adult administration. To facilitate weight-based administration for pediatric and juvenile patients, an HCQS suspension is made by compounding a 200-mg HCQS tablet and suspending the crushed granules into water and Ora-Plus®. The Ora-Plus® suspension does not alter the extreme bitterness of HCQS such that it facilitates oral administration. Additional research has been executed to affirm that a slightly buffered, ion-pairing system, reduces the bitterness of HCQS. The buffered, ion-pairing system can be interwoven into an immediate-release tablet formulation likely without compromising tablet performance. With the taste-masking system embedded, the tablet could be more easily be compounded and suspended in water to generate a palatable oral suspension. Such a novel HCQS 200-mg tablet would be tailored for adult usages wherein the interwoven task-masking system could be utilized to facilitate weight-based administration for pediatric and juvenile patients. The dual quality target product profile of the tablet and the tablet compounded for suspension in water would make the tablet formulation applicable to a wide patient population ranging from pediatric to elder adults to facilitate in improving compliance and overall health outcomes.

Explainable Artificial Intelligence Methods in Combating Pandemics: A Systematic Review.

Despite the myriad peer-reviewed papers demonstrating novel Artificial Intelligence (AI)-based solutions to COVID-19 challenges during the pandemic, few have made a significant clinical impact, especially in diagnosis and disease precision staging. One major cause for such low impact is the lack of model transparency, significantly limiting the AI adoption in real clinical practice. To solve this problem, AI models need to be explained to users. Thus, we have conducted a comprehensive study of Explainable Artificial Intelligence (XAI) using PRISMA technology. Our findings suggest that XAI can improve model performance, instill trust in the users, and assist users in decision-making. In this systematic review, we introduce common XAI techniques and their utility with specific examples of their application. We discuss the evaluation of XAI results because it is an important step for maximizing the value of AI-based clinical decision support systems. Additionally, we present the traditional, modern, and advanced XAI models to demonstrate the evolution of novel techniques. Finally, we provide a best practice guideline that developers can refer to during the model experimentation. We also offer potential solutions with specific examples for common challenges in AI model experimentation. This comprehensive review, hopefully, can promote AI adoption in biomedicine and healthcare.

Proposing Causal Sequence of Death by Neural Machine Translation in Public Health Informatics.

Each year there are nearly 57 million deaths worldwide, with over 2.7 million in the United States. Timely, accurate and complete death reporting is critical for public health, especially during the COVID-19 pandemic, as institutions and government agencies rely on death reports to formulate responses to communicable diseases. Unfortunately, determining the causes of death is challenging even for experienced physicians. The novel coronavirus and its variants may further complicate the task, as physicians and experts are still investigating COVID-related complications. To assist physicians in accurately reporting causes of death, an advanced Artificial Intelligence (AI) approach is presented to determine a chronically ordered sequence of conditions that lead to death (named as the causal sequence of death), based on decedent's last hospital discharge record. The key design is to learn the causal relationship among clinical codes and to identify death-related conditions. There exist three challenges: different clinical coding systems, medical domain knowledge constraint, and data interoperability. First, we apply neural machine translation models with various attention mechanisms to generate sequences of causes of death. We use the BLEU (BiLingual Evaluation Understudy) score with three accuracy metrics to evaluate the quality of generated sequences. Second, we incorporate expert-verified medical domain knowledge as constraints when generating the causal sequences of death. Lastly, we develop a Fast Healthcare Interoperability Resources (FHIR) interface that demonstrates the usability of this work in clinical practice. Our results match the state-of-art reporting and can assist physicians and experts in public health crisis such as the COVID-19 pandemic.

Embedding Fe3C and Fe3N on a Nitrogen-Doped Carbon Nanotube as a Catalytic and Anchoring Center for a High-Areal-Capacity Li-S Battery.

The biggest obstacles of putting lithium-sulfur batteries into practice are the sluggish redox kinetics of polysulfides and serious "shuttle effect" under high sulfur mass loading and lean-electrolyte conditions. Herein, Fe3C/Fe3N@nitrogen-doped carbon nanotubes (NCNTs) as multifunctional sulfur hosts are designed to realize high-areal-capacity Li-S batteries. The Fe3N and Fe3C particles attached to NCNT can promote the conversion of polysulfides. Besides, NCNT can not only enhance the chemisorption of polysulfides but also increase the special surface area and electrical conductivity by constructing a three-dimensional skeleton network. Integrating the merits of high electrical conductivity, high catalytic activity, and strong chemical binding interaction with lithium polysulfides (LiPSs) to achieve in situ anchoring conversion, the Fe3C/Fe3N@NCNT multifunctional hosts realize high sulfur mass loading and accelerate redox kinetics. The novel Fe3C/Fe3N@NCNT/S composite cathode exhibits steady cycle ability and a high areal capacity of 9.10 mAh cm-2 with a sulfur loading of 13.12 mg cm-2 at 2.20 mA cm-2 after 50 cycles.

Effectiveness of intraoperative radiotherapy vs hypofractionated postmastectomy radiotherapy for early stage breast cancer: A protocol for systematic review and meta-analysis.

BACKGROUND: Radiotherapy is one of the essential components of breast cancer treatment. It destroys the remaining cells in the chest area after breast cancer surgery and is useful for reducing the necessity of mastectomies. As a single dose of radiation at the time of breast conserving surgery, intraoperative radiotherapy delivers radiotherapy directly and accurately to the tumor itself or the tumor bed whilst delivering minimal dose to the surrounding normal tissues. Hypofractionated postmastectomy radiotherapy with shorter and more convenient hypofractionated dose schedules might help to treat more patients and reduce cost. We will conduct a comprehensive systematic review and meta-analysis to compare the effectiveness of these 2 therapies in the management of early stage breast cancer. METHODS: Four English databases (PubMed, Embase, Cochrane Library, and Web of Science) and 3 Chinese databases (China National Knowledge Infrastructure, China Science and Technology Journal Database, and Chinese Biomedical Literature Database) will be searched from inception of databases to December 2020 without language limitation. Two reviewers will independently conduct selection of studies, data extraction and management, and assessment of risk of bias. Any disagreement will be resolved by the third reviewer. Review Manager 5.3 (The Cochrane Collaboration, Software Update, Oxford, UK) will be used for data synthesis. Cochrane risk of bias assessment tool will be used to assess the risk of bias. RESULTS: This study will provide a systematic synthesis of current published data to compare the effectiveness of intraoperative radiotherapy vs hypofractionated postmastectomy radiotherapy for early stage breast cancer. CONCLUSIONS: This systematic review and meta-analysis will provide clinical evidence for the effectiveness of intraoperative radiotherapy vs hypofractionated postmastectomy radiotherapy for early stage breast cancer, and inform our understanding of the value of intraoperative radiotherapy and hypofractionated postmastectomy radiotherapy for early stage breast cancer. STUDY REGISTRATION NUMBER: INPLASY2020110115.

Polymeric Nanoscale Drug Carriers Mediate the Delivery of Methotrexate for Developing Therapeutic Interventions Against Cancer and Rheumatoid Arthritis.

Methotrexate (MTX) is widely used as an anticancer and anti-inflammtory drug for treating various types of cancer and autoimmune diseases. The optimal dose of MTX is known to inhibit the dihydrofolatereductase that hinders the replication of purines. The nanobiomedicine has been extensively explored in the past decade to develop myriad functional nanostructures to facilitate the delivery of therapeutic agents for various medical applications. This review is focused on understanding the design and development of MTX-loaded nanoparticles alongside the inclusion of recent findings for the treatment of cancers. In this paper, we have made a coordinated effort to show the potential of novel drug delivery systems by achieving effective and target-specific delivery of methotrexate.

Platinum Atomic Clusters Embedded in Defects of Anatase/Graphene for Efficient Electro- and Photocatalytic Hydrogen Evolution.

Electro- and photocatalytic hydrogen evolution reaction (e-HER and p-HER) are two promising strategies to produce green hydrogen fuel from water. High intrinsic activity, sufficient active sites, fast charge-transfer capacity, and good optoelectronic properties must be taken into consideration simultaneously in pursuit of an ideal bifunctional catalyst. Here, platinum atomic clusters embedded in defects of TiO2 nanocrystals/graphene nanosheets (Pt-T/G) are reported as a bifunctional catalyst for electro- and photocatalytic hydrogen evolution reaction (e-HER and p-HER). High activity is delivered due to the charge transfer from the other part of the catalyst to the active center (Pt2-O4-Tix), decreasing the activation energy of the rate-limiting step, which is revealed by synchrotron X-ray absorption spectroscopy, photoelectrochemical measurements, and simulated calculations. In regard to e-HER, it outperforms the commercial 20 wt % Pt/C catalyst by a factor of 17.5 on Pt mass basis, allowing for a 93% reduction in Pt loadings. In regard to p-HER, it achieves photocatalytic efficiency (686.8 µmol h-1) without any attenuation in 9 h.

Selective Adsorption and Electrocatalysis of Polysulfides through Hexatomic Nickel Clusters Embedded in N-Doped Graphene toward High-Performance Li-S Batteries.

The shuttle effect hinders the practical application of lithium-sulfur (Li-S) batteries due to the poor affinity between a substrate and Li polysulfides (LiPSs) and the sluggish transition of soluble LiPSs to insoluble Li2S or elemental S. Here, we report that Ni hexatomic clusters embedded in a nitrogen-doped three-dimensional (3D) graphene framework (Ni-N/G) possess stronger interaction with soluble polysulfides than that with insoluble polysulfides. The synthetic electrocatalyst deployed in the sulfur cathode plays a multifunctional role: (i) selectively adsorbing the polysulfides dissolved in the electrolyte, (ii) expediting the sluggish liquid-solid phase transformations at the active sites as electrocatalysts, and (iii) accelerating the kinetics of the electrochemical reaction of multielectron sulfur, thereby inhibiting the dissolution of LiPSs. The constructed S@Ni-N/G cathode delivers an areal capacity of 9.43 mAh cm-2 at 0.1 C at S loading of 6.8 mg cm-2, and it exhibits a gravimetric capacity of 1104 mAh g-1 with a capacity fading rate of 0.045% per cycle over 50 cycles at 0.2 C at S loading of 2.0 mg cm-2. This work opens a rational approach to achieve the selective adsorption and expediting of polysulfide transition for the performance enhancement of Li-S batteries.

Wnt/ß-catenin signaling as a useful therapeutic target in hepatoblastoma.

Hepatoblastoma is a malignant tumor in the liver of children that generally occurs at the age of 2-3 years. There have been ample evidence from the preclinical as well as clinical studies suggesting the activation of Wnt/ß-catenin signaling in hepatoblastoma, which is mainly attributed to the somatic mutations in the exon 3 of ß-catenin gene. There is increased translocation of ß-catenin protein from the cell surface to cytoplasm and nucleus and intracellular accumulation is directly linked to the severity of the cancer. Accordingly, the alterations in ß-catenin and its target genes may be used as markers in the diagnosis and prognosis of pediatric live tumors. Furthermore, scientists have reported the therapeutic usefulness of inhibition of Wnt/ß-catenin signaling in hepatoblastoma and this inhibition of signaling has been done using different methods including short interfering RNA (siRNA), miRNA and pharmacological agents. Wnt/ß-catenin works in association with other signaling pathways to induce the development of hepatoblastoma including Yes-associated protein (YAP)1 (YAP-1), mammalian target of rapamycin (mTOR) 1 (mTOR-1), SLC38A1, glypican 3 (GPC3), nuclear factor κ-light-chain-enhancer of activated B cells (NF-kB), epidermal growth factor receptor, ERK1/2, tumor necrosis factor-α (TNF-α), regenerating islet-derived 1 and 3 α (REG1A and 3A), substance P (SP)/neurokinin-1 receptor and PARP-1. The present review describes the key role of Wnt/ß-catenin signaling in the development of hepatoblastoma. Moreover, the role of other signaling pathways in hepatoblastoma in association with Wnt/ß-catenin has also been described.

Sulfur-/Nitrogen-Rich Albumen Derived "Self-Doping" Graphene for Sodium-Ion Storage.

The development of sodium-ion batteries (SIBs) is hindered by the rapid reduction in reversible capacity of carbon-based anode materials. Outside-in doping of carbon-based anodes has been extensively explored. Nickel and NiS2 particles embedded in nitrogen and sulfur codoped porous graphene can significantly improve the electrochemical performance. Herein a built-in heteroatom "self-doping" of albumen-derived graphene for sodium storage is reported. The built-in sulfur and nitrogen in albumen act as the doping source during the carbonization of proteins. The sulfur-rich proteins in albumen can also guide the doping and nucleation of nickel sulfide nanoparticles. Additionally, the porous architecture of the carbonized proteins is achieved through removable KCl/NaCl salts (medium) under high-temperature melting conditions. During the carbonization process, nitrogen can also reduce the carbonization temperature of thermally stable carbon materials. In this work, the NS-graphene delivered a specific capacity of 108.3 mAh g-1 after 800 cycles under a constant current density of 500 mA g-1 . In contrast, the Ni/NiS2 /NS-graphene maintained a specific capacity of 134.4 mAh g-1 ; thus the presence of Ni/NiS2 particles improved the electrochemical performance of the whole composite.

Improving Classification of Breast Cancer by Utilizing the Image Pyramids of Whole-Slide Imaging and Multi-Scale Convolutional Neural Networks.

Whole-slide imaging (WSI) is the digitization of conventional glass slides. Automatic computer-aided diagnosis (CAD) based on WSI enables digital pathology and the integration of pathology with other data like genomic biomarkers. Numerous computational algorithms have been developed for WSI, with most of them taking the image patches cropped from the highest resolution as the input. However, these models exploit only the local information within each patch and lost the connections between the neighboring patches, which may contain important context information. In this paper, we propose a novel multi-scale convolutional network (ConvNet) to utilize the built-in image pyramids of WSI. For the concentric image patches cropped at the same location of different resolution levels, we hypothesize the extra input images from lower magnifications will provide context information to enhance the prediction of patch images. We build corresponding ConvNets for feature representation and then combine the extracted features by 1) late fusion: concatenation or averaging the feature vectors before performing classification, 2) early fusion: merge the ConvNet feature maps. We have applied the multi-scale networks to a benchmark breast cancer WSI dataset. Extensive experiments have demonstrated that our multiscale networks utilizing the WSI image pyramids can achieve higher accuracy for the classification of breast cancer. The late fusion method by taking the average of feature vectors reaches the highest accuracy (81.50%), which is promising for the application of multi-scale analysis of WSI.

Early Prediction of Sepsis in EMR Records Using Traditional ML Techniques and Deep Learning LSTM Networks.

Sepsis is a life-threatening condition caused by infection and subsequent overreaction by the immune system. Physicians effectively treat sepsis with early administration of antibiotics. However, excessive use of antibiotics on false positive cases cultivates antibiotic resistant bacterial strains and can waste resources while false negative cases result in unacceptable mortality rates. Accurate early prediction ensures correct, early antibiotic treatment; unfortunately, prediction remains daunting due to error-ridden electronic medical records (EMRs) and the inherent complexity of sepsis. We aimed to predict sepsis using only the first 24 and 36 hours of lab results and vital signs for a patient. We used the Medical Information Mart for Intensive Care III (MIMIC3) dataset to test machine learning (ML) techniques including traditional methods (i.e., random forest (RF) and logistic regression (LR)) as well as deep learning techniques (i.e., long short-term memory (LSTM) neural networks). We successfully created a data pipeline to process and clean data, identified important predictive features using RF and LR techniques, and trained LSTM networks. We found that our best performing traditional classifier, RF, had an Area Under the Curve (AUC-ROC) score of 0.696, and our LSTM networks did not outperform RF.

Super-resolution imaging by anticorrelation of optical intensities.

Photon bunching, a feature of classical thermal fields, has been widely exploited to implement ghost imaging. Here we show that spatial photon antibunching can be experimentally observed via low-pass filtering of the intensities of the two thermal light beams from a beamsplitter correlation system. Through suitable choice of the filter thresholds, the minimum of the measured normalized anti-correlation function, i.e., antibunching dip, can be lower than 0.2, while its full-width-at-half-maximum can be much narrower than that of the corresponding positive correlation peak. Based on this anti-correlation effect, a super-resolution negative ghost image is achieved in a lensless scheme, in which the spatial resolution can exceed the Rayleigh diffraction limit by more than a factor of two. The setup is quite simple and easy to implement, which is an advantage for practical applications.

Multi-omics profiling of younger Asian breast cancers reveals distinctive molecular signatures.

Breast cancer (BC) in the Asia Pacific regions is enriched in younger patients and rapidly rising in incidence yet its molecular bases remain poorly characterized. Here we analyze the whole exomes and transcriptomes of 187 primary tumors from a Korean BC cohort (SMC) enriched in pre-menopausal patients and perform systematic comparison with a primarily Caucasian and post-menopausal BC cohort (TCGA). SMC harbors higher proportions of HER2+ and Luminal B subtypes, lower proportion of Luminal A with decreased ESR1 expression compared to TCGA. We also observe increased mutation prevalence affecting BRCA1, BRCA2, and TP53 in SMC with an enrichment of a mutation signature linked to homologous recombination repair deficiency in TNBC. Finally, virtual microdissection and multivariate analyses reveal that Korean BC status is independently associated with increased TIL and decreased TGF-ß signaling expression signatures, suggesting that younger Asian BCs harbor more immune-active microenvironment than western BCs.

Hyperbaric oxygen treatment for Parkinson's disease with severe depression and anxiety: A case report.

RATIONALE: Patients with Parkinson's disease (PD) frequently suffer from psychiatric disorders, and treating these symptom whereas managing the motor symptoms associated with PD can be a therapeutic challenge. PATIENT CONCERNS: We report a case of PD patient with severe depression and anxiety that refused to be treated with dopaminagonists or SSRIs, the most common treatments for PD patients suffering from psychiatric symptoms. DIAGNOSES: Parkinson's disease with severe depression and anxiety. INTERVENTIONS: This man was treated with hyperbaric oxygen treatment for 30 days. OUTCOMES: Clinical assessment scores for depression and anxiety, including Unified Parkinson's Disease Rating ScaleI (UPDRS I), UPDRS II, Hanmilton Depression Rating Scale, and Hamiliton Anxiety Rating Scale, were improved following the hyperbaric oxygen treatment. LESSONS: Hyperbaric oxygen treatment may be a potential therapeutic method for PD patient suffering from depression and anxiety. Further research is needed to validate this finding and explore a potential mechanism.

Low-grade slightly elevated and polypoid colorectal adenomas display differential ß-catenin-TCF/LEF activity, c-Myc, and cyclin D1 expression.

AIM: To comparatively investigate the cellular and molecular characteristics of low-grade slightly elevated adenomas and polypoid adenomas. METHODS: Colorectal tumors were collected from 24 patients with slightly elevated adenomas and 23 patients with polypoid adenomas. Five commonly mutated genes (APC, BRAF, KRAS, NRAS, and PIK3CA) were selected for mutational analysis. Paraffin-embedded tumor sections were used to calculate the apoptotic index (AI) and Ki-67 labeling index (KLI). Two pure colorectal epithelial cell lines were created by pooling the slightly elevated and polypoid tumors. Western blots, luciferase assays for ß-catenin-T-cell factor protein/ß-catenin-lymphoid enhancer factor (ß-catenin-TCF/LEF)-driven transcriptional activity, and caspase activity assays were conducted on the two cell lines. RESULTS: Slightly elevated lesions showed a significantly lower APC mutational frequency and a significantly higher KRAS mutational frequency (both P < 0.05). Slightly elevated lesions showed a significantly lower AI (P < 0.05). ß-catenin and ß-catenin-TCF/LEF-driven transcriptional activity was significantly upregulated in slightly elevated lesions (both P < 0.05). In slightly elevated lesions, c-Myc was significantly downregulated, while cyclin D1 was significantly upregulated (both P < 0.05). ß-catenin-TCF/LEF-driven transcriptional activity was negatively correlated with c-Myc (ρ = -0.78). Slightly elevated lesions displayed significant Bcl-2 and Bcl-xL upregulation (both P < 0.05) along with significant decreases in caspase-9 and caspase-3 activity (both P < 0.05). c-Myc was negatively correlated with Bcl-2 and Bcl-xL (ρ = -0.74 and -0.78, respectively). CONCLUSION: The lower apoptotic activity of low-grade slightly elevated adenomas can be partly attributed to upregulated ß-catenin pathway activity and downregulated c-Myc expression.

Interpretable Predictions of Clinical Outcomes with An Attention-based Recurrent Neural Network.

The increasing accumulation of healthcare data provides researchers with ample opportunities to build machine learning approaches for clinical decision support and to improve the quality of health care. Several studies have developed conventional machine learning approaches that rely heavily on manual feature engineering and result in task-specific models for health care. In contrast, healthcare researchers have begun to use deep learning, which has emerged as a revolutionary machine learning technique that obviates manual feature engineering but still achieves impressive results in research fields such as image classification. However, few of them have addressed the lack of the interpretability of deep learning models although interpretability is essential for the successful adoption of machine learning approaches by healthcare communities. In addition, the unique characteristics of healthcare data such as high dimensionality and temporal dependencies pose challenges for building models on healthcare data. To address these challenges, we develop a gated recurrent unit-based recurrent neural network with hierarchical attention for mortality prediction, and then, using the diagnostic codes from the Medical Information Mart for Intensive Care, we evaluate the model. We find that the prediction accuracy of the model outperforms baseline models and demonstrate the interpretability of the model in visualizations.

A Novel Temporal Similarity Measure for Patients Based on Irregularly Measured Data in Electronic Health Records.

Patient similarity measurement is an important tool for cohort identification in clinical decision support applications. A reliable similarity metric can be used for deriving diagnostic or prognostic information about a target patient using other patients with similar trajectories of health-care events. However, the measure of similar care trajectories is challenged by the irregularity of measurements, inherent in health care. To address this challenge, we propose a novel temporal similarity measure for patients based on irregularly measured laboratory test data from the Multiparameter Intelligent Monitoring in Intensive Care database and the pediatric Intensive Care Unit (ICU) database of Children's Healthcare of Atlanta. This similarity measure, which is modified from the Smith Waterman algorithm, identifies patients that share sequentially similar laboratory results separated by time intervals of similar length. We demonstrate the predictive power of our method; that is, patients with higher similarity in their previous histories will most likely have higher similarity in their later histories. In addition, compared with other non-temporal measures, our method is stronger at predicting mortality in ICU patients diagnosed with acute kidney injury and sepsis. CATEGORIES AND SUBJECT DESCRIPTORS: H.3.3 [Information Storage and Retrieval]: Retrieval models and rankings - similarity measures; J.3 [Applied Computing]: Life and medical sciences - health and medical information systems. GENERAL TERM: Algorithm.