Current Research Projects

• Prediction of decompensation in COVID-19 patients using physiologic data analysis (Co-PI) IRB: 20-07294-XP

Coronavirus disease 2019 (COVID-19) has emerged as a pandemic and a public health crisis. COVID-19 causes an overwhelming inflammatory response that leads to immunosuppression and patients with severe chronic health conditions can decompensate quickly and require early supportive therapy. Recent work using state-of-the-art machine learning and deep learning techniques have suggested that ‘physiomarkers’ exist several hours in continuous physiological data-streams (CPD) prior to the onset of cardiorespiratory events and systemic inflammatory response. It is also showed that the rapid deterioration of COVID-19 patients is driven by immunopathogenesis, whereby the release of inflammatory markers leads to acute respiratory distress syndrome, multiorgan failure, and death. Our primary objective is to develop artificial intelligence capable of predicting decompensation earlier using a minimal set of streaming physiological data in real-time. In this aim we hypothesize that ‘physiomarkers’ can be identified in COVID-19 patients well before the onset of decompensation using continuous physiological data-streams, leading to earlier interventions and reduced mortality. Predictive algorithms having high sensitivity and specificity could be used to generate smart alarms to provide early warning to clinicians prior to decompensation and organ failure, allowing time for pre-emptive or early treatment. This approach may also guide decision-making about monitoring and ventilation support requirement.

• Machine learning predicts early onset acute organ failure in ICU patients with sickle cell disease (Research Support Staff) IRB:18-05881-XP

Sickle cell disease (SCD) is a genetic disorder of the red blood cells, resulting in multiple acute and chronic complications. Patients with SCD develop chronic organ dysfunction, which may progress to organ failure during disease exacerbations. Early detection of acute physiological deterioration leading to organ failure is not always attainable. The aim of this study is to test the hypothesis that machine learning physiomarkers can predict the development of organ dysfunction in a sample of adult patients with SCD admitted to intensive care units (ICUs). The discovery of salient physiomarkers through machine learning techniques has the potential to further accelerate the development and implementation of innovative care delivery protocols and strategies for medically vulnerable patients.

Mohammed A, Podila P, Davis R, Ataga K, Hankins J, Kamaleswaran R. (2020) Using Machine Learning to Predict Early Onset Acute Organ Failure in Critically Ill Intensive Care Unit Patients With Sickle Cell Disease: Retrospective Study. Journal of Medical Internet Research

• Physiomarkers from continuous physiological data streams predicts early onset of sepsis in critically ill adults (Research Support Staff) IRB: 16-04985-XP

Sepsis is a life-threatening condition with high mortality rates. Early detection and treatment are critical to improving outcomes. Our primary objective is to analyze continuous physiological data streams and characterize unique temporal profiles of these physiomarkers for predicting sepsis earlier. Physiomarkers, a set of signal processing features, were derived from five physiological data streams including heart rate, respiratory rate and blood pressure (systolic, diastolic, and mean), captured every minute from the bedside monitors. This discovery of salient physiomarkers derived from continuous bedside monitoring may shown to be temporally and differentially expressed in septic patients. By the use of this information, artificial intelligence models can be developed to predict sepsis earlier in critically ill patients.

Mohammed A, Van Wyk F, Chinthala LK, Khojandi A, Davis RL, Blum JM, Coopersmith CM, Kamaleswaran R. Temporal and Differential Expression of Physiomarkers Predicts Sepsis in Critically Ill Adults. SHOCK (In Press)

• Applying Artificial Intelligence for Early Identification of Parkinson’s Disease (Research Support Staff) IRB: 19-06751-XP

Parkinson’s disease (PD) is a progressive disabling neurodegenerative disorder that is understood to have three developmental phases, referred to as preclinical, premotor (or prodromal), and motor phases. This project will discover novel noninvasive electrocardiogram (EKG) features of PD onset that can be processed within an artificial intelligence framework to identify patients at the prodromal PD phase. As suggested by our preliminary data, artificial intelligence and machine learning can be utilized to uncover hidden, novel EKG markers that enable accurate classification of subjects with prodromal PD. Cardiac electrical activity provides important information about the likelihood of future PD not captured by classical heart rate variability metrics. Machine learning applied to standard 10-second ECGs may help identify subjects at high risk of having prodromal PD. The research also seeks to understand the relationship among age, gender, ethnicity, race, and neighborhood information with the development of Parkinson’s Disease.

Akbilgic O, Kamaleswaran R, Mohammed A, Ross W, Masaki K, Petrovitch H, Tanner CM, Davis RL, Goldman SM. Electrocardiographic Changes Predate Parkinson’s Disease Onset. Scientific Reports (Nature Publishing Group) (In Revision)

• Predicting intracranial hypertension episodes in PICU patients (Research Support Staff) IRB:19-06976-XP

When intracranial pressure (ICP) becomes sufficiently elevated, brain perfusion is compromised and cerebral ischemia can result; if allowed to progress, the inevitable outcome is herniation and death.  Absent invasive monitoring, characteristic vital sign changes comprising the Cushing Response (hypertension, bradycardia and irregular respirations) provide insight into the very late stage impending herniation only.  Clinical conditions are not always ideal for invasive ICP monitor placement, yet, having very early insight into ICP elevation is needed to minimize brain injury.  In contrast, invasive arterial blood pressure monitoring is a mainstay of ICU care, and it is becoming increasingly recognized that high resolution arterial waveform data can provide insights into various physiologic states. Our hypothesis is that machine-/deep-learning analyses of arterial waveforms can predict intracranial hypertension using patient systemic physiologic and ICP waveform data streams that have been collected from past patients. Additionally, we will use this retrospective data to understand shortcomings and data processing steps necessary to stage a prospective study. We will design an optimal prospective study to use machine-/deep-learning methods to identify signals within systemic physiologic data streams that predict intracranial hypertension.

• TweenVax: A comprehensive practice-, provider-, and parent/patient-level intervention to improve adolescent HPV vaccination. In collaboration with researchers at Emory and Johns Hopkins universities, we conduct a multi-phase study to utilize formative research to assess the efficacy of the TweenVax package to improve the provider recommendation and administration of HPV vaccine among pediatric and family physician practices in Georgia and Tennessee. This project is funded by ational Cancer Institute, National Institutes of Health, US Department of Health and Human Services.
  
  
• SIEMA: The Semantics, Interoperability, and Evolution for Malaria Analytics (SIEMA) is an analytic framework to help integrate dynamic surveillance data from multiple sources and health systems to support decision making for malaria control and elimination in Sub-Saharan Africa. We used ontologies and semantic web services to increase the interoperability of the various sources. We also designed a tool allowing researchers to detect and identify updates and changes in malaria data sources and to update the services so that the whole surveillance systems becomes more resilient. SIEMA has been funded by Melinda and Bill Gates Foundation, and Microsoft Research Inc.
  
  
• SPACES: Semantic Platform for Adverse Childhood Experience (ACEs) Surveillance (SPACES) is a clinical mental health counseling and surveillance platform that facilitates the access to the relevant integrated information related to adverse childhood experiences, enables discovering the causality pathways and assists researchers, clinicians, public health practitioners, social workers, and health organization in studying the ACEs, identifying the trends, as well as planning and implementing preventive and therapeutic strategies. This project is supported by Memphis Research Consortium (MRC).
  
  
• SODHS: Social Determinants of Health Surveillance (SODHS) provides a framework to integrate social, economic, and environmental determinants of health (Sociomarkers), and biological markers (Biomarkers) with multiple health related data sets to investigate the association between socio-economic, racial and ethnical disparities with multiple adverse health outcomes, such as Asthma, Mental Health, Sepsis, Adverse Surgical Outcome, Comorbidity of Chronic Diseases, Sickle Cell Disease (in collaboration with researchers from St. Jude Children's Research Hospital and University of Memphis), and COVID-19. The funding for this project is partially provided through grants from Memphis Research Consortium (MRC), and the Children's Foundation Research Institute (CFRI) at Le Bonheur Children’s Hospital.
  
  
• UPHO: Memphis Urban Public Health Observatory (UPHO) uses advanced AI, Knowledge Representation and Semantic Web tools and techniques to assist epidemiologists, public health authorities and researchers collect data from several resources, foster the collection of surveillance data in a consistent and comparable way across jurisdictions to estimate the incidence and prevalence of different health conditions, as well as related risk factors. 
  
  
• The PHL: Personal health libraries (PHL) are intended to help consumers manage their health information and play a more active role in their healthcare, health, and wellness. We propose implementing an integrated PHL solution that takes patients' perspectives, social and behavioral characteristics into account using a Hybrid Recommender Digital Librarian system (HRDL). The PHL system provides a single point of secure access to patients' digital health information and improves their health information seeking skills through personalized and tailored alerts and recommendations based on the unique attributes of an individual user or group.
  
  
• POLE.VAULT: The POLicy EVAlUation & Logical Testing (POLE.VAULT) Framework assists different stakeholders and decision-makers in making informed decisions about different health-related interventions, programs and ultimately policies, based on the contextual knowledge and the best available evidence at both individual and aggregate levels. POLE.VAULT uses the theory of change (TOC) approach along with logic models that define the intervention under consideration to generate a causal diagram and an ontology-based inference model for causal description. The resulting causal diagram will then be compared to existing knowledge and data to determine whether the intervention is coherent, internally consistent and its goals are achievable in the allotted time with the resources provided. The contextual knowledge and semantics provided by the ontology will generate a more explainable, understandable, and trustworthy approach to compare and assess different interventions based on their shared goals.