IT's Artificial Intelligence Projects at the U

Accordion Group

Open All Tabs

• AI Projects for Academics

  Type of AI Project	Status	Project Name	Description
  Generative AI	Completed - In Production	Blackboard Design Assistant	A generative AI system built into Blackboard Ultra that can help faculty design syllabi, assignments, and test questions.
  Generative AI	Completed - In Production	Improve Degraded Videos	Improve the quality of degraded videos with Topaz Video so the videos can be used in student and faculty projects.
  Generative AI	Completed - In Production	Improve Image Resolution	Enhance old footage and images to today's standards of resolution.
  Generative AI	In Flight	Obstetrics Teaching Dialogue	Using Open AI gpt to generate teaching dialogue for obstetric emergency simulations.
  Generative AI	Completed - In Production	AI Faculty Guidance	Guidance for faculty on ways they should and should not use generative AI in their courses, including examples, sample syllabus language, the use of AI detection software, and other considerations.
  Generative AI	In Flight	Donor Interaction	Leveraging Generative AI to analyze donor reports, generate thank you letters, provide recommendations, and generate custom news feeds based on donor interests.
  Generative AI	In Flight	Language Learning Bots	Using generative AI to create language learning bots to assist with acquisition of foreign languages.
  Machine Learning (ML)	Completed - In Production	Grading Assistance	Assists faculty in grading assignments with computer code, equations, essays, and other long-form answers through the use of Gradescope, a commercially available platform that integrates with Blackboard.
  Machine Learning (ML)	In Flight	Student Retention Model	Utilizing machine learning to identify students at-risk of early attrition.
  Machine Learning (ML)	In Flight	Admission Essay Scoring	Using AI to score personal essays for undergraduate admissions. Scores can be generated against a rubric.
  Natural Language Processing (NLP)	Completed - In Production	Faculty Letter Entitlement	NLP models that extract important information about faculty entitlements within faculty offer letters, such as entitlement for lab space, equipment, renovations, and research funding.

Accordion Group

Open All Tabs

• AI Projects for Clinical Care

  Type of AI Project	Status	Project Name	Description
  Computer Vision	Completed - In Production	Laparoscopic Video Review	Video recording of the laparoscopic procedures that uses AI to review the procedure steps and evaluate patient outcomes.
  Computer Vision	Completed - In Production	Augmented Reality Surgical Guidance	Augmented reality surgical guidance.
  Computer Vision	Completed - In Production	Cardiology Computer Vision Analysis	Using computer vision to support blood flow analysis, 3D modeling without invasive procedure, and coronary artery disease diagnosis.
  Computer Vision	Completed - Not In Use	Enhance Patient Arrival with Computer Vision	Deploying Microsoft facial and license plate recognition to identify patients with scheduled appointments as they approach the facility to provide a low friction arrival experience. License plate recognition for visitor management and parking integration.
  Computer Vision	Completed - Not In Use	Food Service Automation - Just Walk Out	Deploying Amazon Just Walk Out (JWO) technology on Medical Campus (Calder Library) to pilot the use of automatous food service option in areas with limited availability for staff and patients.
  Computer Vision	Completed - In Production	Reduce Radiology Error with Image Analysis	Reduce radiologist errors by analyzing post processed images, providing confirmation of findings and/or identify findings for possible inclusion in the final report.
  Computer Vision	Completed - In Production	Mammo Post Processing	Mammo post processing - density secondary capture scorecard.
  Computer Vision	Completed - In Production	Calcium Scoring	Calcium Scoring of non-typical CT valve contrast studies.
  Computer Vision	Completed - In Production	Neurosurgical Planning	Neurosurgical planning software which incorporates connectomics; utilizing cloud computing for large-volume data processing, and intuitive browser-based interfaces.
  Computer Vision	Completed - In Production	Brain Volume Calculations	MS PLex and Morphometric brain volume calculations.
  Computer Vision	Completed - In Production	Nuclear Medicine Processing	Nuclear Medicine processing, display and analysis for SPECT-CT, PET-CT, PET-MR including dosimetry; image quantification and advanced image reconstruction.
  Computer Vision	Completed - Not In Use	Emotional Recognition Monitoring	Real-time detection of patient experience in the inpatient setting through facial expressions allows for the opportunity to intervene to improve patient satisfaction prior to discharge.
  Generative AI	Completed - In Production	Patient Message Draft	Draft responses to patient messages.
  Generative AI	In Flight	Pre-Visit Note Summarization	Summarize recent notes before a visit.
  Generative AI	Completed - Not In Use	Efficacy Study - AI Patient Intake	Partnering with SOAP AI and the Mayo Clinic on a NCI-funded research study to study the efficacy of AI patient intake for primary care new patient visits replacing the manual questionnaires and summarizing key findings for the provider.
  Generative AI	Completed - In Production	Quantitative Liver Analysis	Quantitative liver analysis with software that provides clinicians with an interpretable report of liver health.
  Machine Learning (ML)	Completed - In Production	Predictive Models - Order Search	The cloud-based order search model learns from the sequences of search terms used across your whole organization, as well as user-level and department-level statistics that are calculated in Chronicles. Brings the most common order up when user searches for order.
  Machine Learning (ML)	Completed - In Production	Predictive Models - Deterioration Index Score	Model identifies hospitalized patients who may be clinically deteriorating, which allows clinicians to proactively intervene and treat patients earlier, which may help reduce adverse events, ICU admissions, and cardiac arrests. Model updates dynamically based on information entered into the chart, providing clinicians with the most up-to-date risk assessment information.
  Machine Learning (ML)	Completed - In Production	Predictive Models - Forecasted Hospital Census	Model predicts expected capacity for the upcoming hour, day, week, month, or year on operational dashboards to better identify and mitigate potential overutilization or capacity problems before they arise. This information can help plan discharges, schedules, and staffing to keep up with expected capacity.
  Machine Learning (ML)	Completed - In Production	Predictive Models - Inbasket language Detection	The language detection model uses cloud computing to label incoming patient medical advice request messages with the language they were written in. Organizations can set up rules based on the message’s language to allow users to forward the message to an appropriate translation pool.
  Machine Learning (ML)	Completed - In Production	Predictive Models - Inbasket Categorization	The system reviews patient medical advice request messages to identify a category to apply to the message. Clinical and support staff can use these categories to sort and triage messages sent to a pool. Categories include such things as clinical actionable messages and FYI messages.
  Machine Learning (ML)	Completed - In Production	Predictive Models - Readmissions Risk Score	This model helps identify patients who are most at risk of an unplanned readmission in the 30 days after discharge and can be fully integrated within clinical workflows, allowing clinicians and hospital staff to focus more resources on at-risk patients. The model uses dozens of data points such as diagnoses, medications, and lab results.
  Machine Learning (ML)	Completed - Not In Use	Predictive Model - OCR Image to Text Transcription	The model transcribes the text of scanned documents, such as notes, results, and referrals, depending on your organization's configuration options. After the model runs on a set of image files, the transcribed text from those documents is available in your system and can appear in your clinicians' search results. This model requires the Cognitive Computing license but doesn't count toward the ten models included with that license.
  Machine Learning (ML)	Completed - Not In Use	Predictive Models - ED Patient Likelihood to Occupy an Inpatient Bed	Using this model, hospital staff can take a proactive approach to bed planning. Instead of constantly checking the ED census, staff can review the model predictions for patients who will likely need a bed in an inpatient unit and start preparing for these arrivals before an admit order is placed. This increased lead time enables capacity planning staff to proactively estimate the number of beds and staff needed at a given time. The model is integrated within the capacity dashboards already used for capacity management.
  Machine Learning (ML)	Completed - In Production	Predictive Model: Amb Risk of HTN	Care managers and clinicians can use this predictive model to help determine which patients have the highest risk of developing hypertension within the next two years. Due to a shift in the definition of hypertension by the American Heart Association and American College of Cardiology, the model is effectively predicting the risk of stage 2 hypertension. Refer to the Additional Considerations topic for more information.
  Machine Learning (ML)	In Flight	Predictive Models - Risk of Surgical Site Infection	This model identifies patients prior to surgery who are most at risk of developing a surgical site infection (SSI).
  Machine Learning (ML)	In Flight	Predictive Model - Sepsis (Version 2)	Cloud based version 2 for early detection of sepsis in adult patients.
  Machine Learning (ML)	In Flight	Predictive Model: Amb Risk of Initial Myocardial Infarction	Care managers and other clinicians can use this model to see an adult patient's risk of an initial myocardial infarction in the next two years so they can take preventive measures.
  Machine Learning (ML)	In Flight	Predictive Model - Risk of Hospital Admission or ED Visit (Version 2)	Version 2 of the Risk of Hospital Admission or ED Visit model is a cloud-based logistic regression model with 55 features. This version offers the potential for improved performance due to the expanded feature set and its ability to localize on your patient data.
  Machine Learning (ML)	In Flight	Predictive Model -  Risk of ICU Readmission or Mortality	Use this model to identify patients who are at risk of being readmitted to the ICU or dying. The model uses dozens of data points such as code status, vital signs, ventilation, and LDAs to help ICU clinicians make better informed decisions during rounding and discharge planning workflows.
  Machine Learning (ML)	In Flight	Predictive Model: Likelihood of Discharge by End of Day Today or by End of Day Tomorrow	Using this model, your staff can more accurately predict whether admitted patients will be discharged today or tomorrow.
  Machine Learning (ML)	In Flight	Patient Affordability	Analyzing affordability of UHealth to patients in our market.
  Machine Learning (ML)	In Flight	Predictive Models - DVT/PE	Identifies patients at-risk of developing deep vein thrombosis (DVT) and/or pulmonary embolism (PE), which can be life-threatening if not identified earlier.
  Machine Learning (ML)	Completed - Not In Use	Predictive Model - Risk of Patient No-Show (Version 2)	Version 2 of the Risk of Patient No-Show model is a cloud-based, random forest model. The model is localized on your patient population and predicts a patient's likelihood to no-show. By using Nebula, Epic's cloud-based platform, version 2 of the model has better performance because it uses a different model type and expanded feature set, and it includes patient-initiated late cancels in the definition of a no-show.
  Machine Learning (ML)	Completed - Not In Use	Predictive Models - Remaining Length of Stay	Model predicts the remaining of length of stay of all adult patients with an inpatient admission (inpatient, observation status) at all locations.
  Machine Learning (ML)	Completed - Not In Use	Predictive Model - Risk of Negative Diabetes Outcomes	Using this model, care managers and other clinicians can review diabetes-related complications a patient is expected to develop over the next two years. Then they can drill down into the contributing factors and determine what to do to lower that risk. Care managers can also perform outreach or bulk enroll patients in Compass Rose episodes based on the risk of negative outcomes of type 2 diabetes score.
  Machine Learning (ML)	Completed - Not In Use	Predictive Model - Hospital Acquired Acute Kidney Injury	This model identifies admitted patients who are at risk of experiencing acute kidney injury (AKI) so clinicians can determine whether it's possible to limit their use of potentially harmful treatments and closely monitor serum creatinine and urine output to detect AKI more quickly.
  Machine Learning (ML)	Completed - Not In Use	Predictive Model - Sepsis (version 1)	Model helps clinicians identify patients with sepsis before their condition worsens with real-time, point-of-care warnings to allow for rapid and risk-adjusted interventions. The model uses data points such as medications, lab results, vital signs, and LDAs to identify patients prior to sepsis progression and organ damage or death occurs.
  Machine Learning (ML)	Completed - Not In Use	Predictive Model - Risk of Opioid Use Disorder or Overdose	The model can be used a number of ways in clinician-facing workflows, but two primary workflows are recommended for this model. The first is showing the score in Storyboard or a similarly visible location in the chart so clinicians can reference a patient’s score before ordering. This workflow is ideal for settings where opioids might be prescribed on a regular basis, such as in the emergency department, in a pain clinic, or during inpatient discharge planning. The second primary workflow is in a BestPractice Advisory (BPA) related to opioid ordering so users placing an opioid order are alerted when a patient has high risk. This workflow is ideal for outpatient and emergency department ordering. The model could also be used as additional filtering for your organization’s existing opioid-related BPAs.
  Machine Learning (ML)	Completed - Not In Use	Predictive Model - Early detection of pediatric sepsis	Use a set of three models created by Nationwide Children's Hospital to detect and treat sepsis in pediatric patients earlier in both inpatient and emergency departments. One model scores patients in the ED and the other two models score admitted patients. One inpatient model scores the general pediatric population and one model scores pediatric patients with high-risk conditions who are predisposed to developing sepsis. These models uses data points such as temperature, BP, behavioral assessments, length of stay, and diagnoses to help clinicians identify septic pediatric patients (under 20 years old) before they worsen. The inpatient model that scores patients with high-risk conditions is not available to organizations in the EU or the UK due to locale-based differences in the regulation of decision support.
  Machine Learning (ML)	Completed - Not In Use	Predictive Model - Amb Risk of Hospital Admission or ED Visit for Asthma	Ambulatory - if implementing or has an existing population health workflow around asthma patients. Care managers and other clinicians can use this model to see an adult asthma patient's risk of a hospital admission or ED visit related to asthma in the next year. The Hospital Admissions and ED Visits for Asthma model helps clinicians and care managers identify the asthma patients who are most at risk of ending up in the ED due to asthma-related conditions. The statistically derived model uses a proprietary algorithm and 17 variables to stratify your adult asthma population into risk buckets. Identifying and stratifying these patients makes it possible for clinicians to deliver proactive, targeted interventions that are designed to prevent dangerous flare-ups. The goal of the model is to better outcomes for patients and better utilization of your resources, such as your ED.
  Machine Learning (ML)	Completed - Not In Use	Predictive Model: Amb Risk of Pediatric Asthma exacerbation	Ambulatory - if pop health workflow around this population. Clinicians can see the risk of a pediatric asthma patient under the age of 18 having a severe asthma attack in the next 90 days. They can review the factors contributing to that risk and allocate care and resources accordingly.
  Machine Learning (ML)	Completed - Not In Use	Predictive Model: ICU In-Hospital Mortality Risk (Retrospective)	Epic's statistically derived in-hospital mortality risk model for critical care patients uses a proprietary algorithm and 49 variables to calculate a patient's expected mortality risk for retrospective risk-adjusted benchmarking of ICU mortality rates, enabling clinicians, unit managers, and hospital administrators to see outcomes and quality of care in their ICUs. This model allows you to benchmark your outcomes without the need for time-intensive manual data review and abstraction
  Machine Learning (ML)	Completed - Not In Use	Predictive Model:  Inpatient Risk of Falls	Use this model to identify patients who are at risk of falling in the ED or an inpatient unit. The model uses dozens of data points such as financial class, vitals, assessments, medications, and lab results to help your organization identify patients who are most at risk of falling. This information enables nurses to focus interventions on the patients who are most likely to fall without them. The model can improve on the performance of manual fall risk assessments, while requiring minimal additional input from clinicians or operational staff. After it is implemented, the model can run continuously, saving up to 3 minutes per patient per day.
  Machine Learning (ML)	Completed - Not In Use	Predictive Model: Amb Risk of Admission for Heart Failure	Care managers and other clinicians can use this model to see an adult patient's risk of a hospital admission related to heart failure in the next year.
  Machine Learning (ML)	Completed - Not In Use	Predictive Model: Amb Pediatric Risk of Hospital Admission or ED Visit	Care managers and other clinicians can see the percent chance that a pediatric patient will end up admitted or in the emergency department in the next six months. They can review the factors contributing to that risk, reach out to the patient or the patient's guardian, and reduce the likelihood that an adverse health event occurs.
  Machine Learning (ML)	Completed - Not In Use	Predictive Model - Likely Surgical Admit Destination	This model is standardly included in the Capacity Management dashboard for organizations that automatically create bed requests for surgical admit patients.
  Machine Learning (ML)	Completed - Not In Use	Predictive Model: ICU Length of Stay (Retrospective)	Epic's statistically derived ICU length of stay model uses a proprietary algorithm and 60 variables to calculate a patient's expected length of stay in the ICU for retrospective risk-adjusted benchmarking, enabling clinicians, unit managers, and hospital administrators to see outcomes and quality of care in their ICUs. This model allows you to benchmark your outcomes without the need for time-intensive manual data review and abstraction.
  Machine Learning (ML)	In Flight	Predictive Model: End of Life Index	The End of Life Care Index is a logistic regression that predicts the risk of one-year mortality. This model is designed to be used on all adult patients using frequently documented features.
  Machine Learning (ML)	In Flight	Predictive Models - Patient No Show (Epic)	Epic provided model to predict patient no shows.
  Machine Learning (ML)	Completed - In Production	Predictive Models - No Show (Custom)	Identifies the likelihood of an outpatient appointment being a no show. The model predictions are aggregated at the physician, clinic, and department levels to better assist in capacity management and planning, while limiting the potential impact on model biases and limiting access to care to a single patient.
  Machine Learning (ML)	In Flight	Predictive Models - Hospital Ambulatory Surgery Admissions (Custom)	Some hospital ambulatory surgeries end up requiring a hospital bed following surgery, putting strain on capacity planning. This model will help predict the number of patients requiring a hospital bed based on scheduled procedures.
  Machine Learning (ML)	In Flight	Predictive Models - Likelihood to Recommend	Identifies the likelihood of patient experience survey resources and correlate key variables about patients or their visit.
  Natural Language Processing (NLP)	Completed - In Production	ICD10 Recommendation	Free text indication interpretation and ICD10 recommendation. Aims to improve ordering provider efficiency and reimbursements.
  Natural Language Processing (NLP)	In Flight	Clinical Entity Recognition - Thrombolytic/Bleeding Events in Cancer patients (Custom)	Using a set of custom natural language processing (NLP) models to identify thrombolytic or bleeding events in cancer patients to aid with quality improvement and potentially reporting of adverse events of patients on clinical trials. All clinical notes of patients within defined cohorts are processed by the NLP engine, which outputs annotated notes for easier review by clinicians. This process is estimated to reduce the time to review by over 20 times (2000%). Currently, models are running for 3 patient cohorts (oncology patients initiating chemotherapy, patients who received cancer-related surgery, and pediatric oncology patients), and is being trialed in patients with multiple myeloma.
  Natural Language Processing (NLP)	Completed - In Production	Patient Experience Word Cloud	Taking all comments from patients within the Patient Experience CAHPS surveys, an NLP model is used highlight topics and keywords while removing any low impact words. The results are visualized as a word cloud.
  Natural Language Processing (NLP)	In Flight	Changes in Medication Treatment for Glaucoma Patients	Treatment plans for glaucoma patients are documented within clinical notes, making it difficult to study. NLP models are being developed to identify if there were any changes to treatment plans indicated within the clinical note, and to extract the medication information (medication name and dosage). A pre-trained model processes all relevant clinical notes and outputs notes annotated with medications, allowing clinicians to review for accuracy.
  Natural Language Processing (NLP)	Completed - In Production	Ambient Note Writing	Use ambient voice technology to write visit notes based on the conversation during office visits.
  Robotics Process Automation (RPA)	In Flight	Electronic Medication Prior Authorization	Automates medication prior authorization workflows with a fully integrated solution that proactively initiates prior auth requests and receives quick responses with the EHR.
  Robotics Process Automation (RPA)	In Flight	Automated Exam Room Allocation	Working with Epic R&D to pilot new module for staff and exam room allocation that will provide operational lead with actionable data on the allocation and assignment of exam rooms to providers.
  Robotics Process Automation (RPA)	In Flight	Automated Physician Leave Requests	Streamlining the leave request process for faculty by automate the requests for approval and directly sending the approved leave into the scheduling system eliminating manual reconciliation work performed by master scheduling team.

Accordion Group

Open All Tabs

• AI Projects for Corporate Services

  Type of AI Project	Status	Project Name	Description
  Computer Vision	Completed - In Production	Enhance Tower Visitor Management	Enhancement to visitor management system in the Tower to identify visitors, as well as identify individuals on a security watchlist and alert public safety.
  Generative AI	Completed - In Production	Adobe Text to Image Generation	Generative AI that Allows for text to image.
  Generative AI	In Flight	Chrome Help Me Write	An experimental feature in Google Chrome that uses generative AI to assist users in writing more effectively online. It's designed to provide suggestions for short-form content like reviews and surveys. This AI project has launched for the University community's use; however, we are currently discovering additional features within the tool, so the project continues in flight until futher notice.
  Generative AI	In Flight	Google Gemini	A multimodal suite of AI models that power a variety of Google products and services, including the Bard chatbot. This AI project has launched for the University community's use; however, we are currently discovering additional features within the tool, so the project continues in flight until futher notice.
  Generative AI	In Flight	Microsoft Copilot	AI-powered language model capable of generating human-like text based on context and past conversations. Formerly known as Bing Chat. This AI project has launched for the University community's use; however, we are currently discovering additional features within the tool, so the project continues in flight until futher notice.
  Generative AI	Completed - In Production	Cyber Threat Prevention	Protects against internet threats, by automatically preventing unknown/malicious cyber-attacks.
  Generative AI	Completed - In Production	Malware Protection	Protect against malware with next-gen antivirus.
  Generative AI	In Flight	Azure OpenAI (AOAI) Chatbots	Azure OpenAI (AOAI) is similar to Copilot, but can be accessed programmatically in a secure manner, enabling OpenAI’s LLM to power chatbots. Specifically, AOAI can be used in various ways including Retrieval-Augmented Generation (RAG), which optimizes the output of a LLM so it references internal documents as the knowledge base outside of its initial training data. Use cases include creating a chatbot to look up UM’s Investigator Manual or a chatbot to help train radiology residents in use of various radiology software.
  Machine Learning (ML)	In Flight	Expense Forecast	Forecasting expenses using past trend of expenses and forecasted future volume.
  Machine Learning (ML)	Completed - In Production	Cyber Theat Blocking	Monitor, identify, prevent, correlate anomaly behaviors, alert, and block malicious threats.
  Machine Learning (ML)	Completed - In Production	Cyber Threat Monitoring & Prevention	Monitor, identify, correlate anomaly behaviors, and predict cyber threats.
  Machine Learning (ML)	Completed - In Production	Email Protection	Email security solution to identify, protect and prevent against email driven cyber-attacks, such as phishing.
  Machine Learning (ML)	Completed - In Production	Network Device Discovery	Discovery of on-network devices.
  Machine Learning (ML)	Completed - In Production	Data Integrity Check	Data integrity checker within the cybervault for airgap back-ups.
  Machine Learning (ML)	Completed - In Production	Revenue Plan Development	Model for creating clinical volume and gross charges plan, preserving the correlation between different types of clinical volumes and related gross charges.
  Machine Learning (ML)	Completed - In Production	Leading Indicator Revenue Forecast	Forecasting Net Patient Service Revenue using leading indicator of scheduled ambulatory appointments and the past trend of downstream revenue generated from completed ambulatory visits.
  Machine Learning (ML)	Completed - In Production	Lagging Indicator Revenue Forecast	Forecasting Net Patient Service Revenue using lagging indicator of past gross charges.
  Machine Learning (ML)	Completed - In Production	Data Architecture Modernization	Modernize Epic data architecture.
  Natural Language Processing (NLP)	Completed - In Production	Cybersecurity Incident Response	Generative AI security layer that enhances security team's visibility to ability to act.
  Robotics Process Automation (RPA)	Completed - In Production	Workday Chatbot Assistant	Chatbot to assist with HR, finance, and supply chain questions. Assists with such tasks as recommending support material and flu vaccine status submissions.
  Robotics Process Automation (RPA)	Completed - In Production	Cybersecurity	Tool leveraging ML, NLP, and RPA to enhance cyber security.
  Speech	Completed - Not In Use	IT Helpdesk Call Triage	Leveraging conversational AI in the UHIT Help desk for call triage, routing and escalations.  The tool will learn and evolve with the intent of providing optimal assistance to our internal users.
  Machine Learning (ML)	In Flight	Scan360	Application to enhance clinical research.
  Machine Learning (ML)	Completed - In Production	Block Malicious Emails	Software to automatically detect and block malicious inbound emails based on IPs.

Accordion Group

Open All Tabs

• AI Projects for Research

  Type of AI Project	Status	Project Name	Description
  Computer Vision	In Flight	Virgo Endoscopy Reading	Virgo is seeking FDA approval for their device using AI to read endoscopy procedures and detect issues. UHealth is participating in this study for follow up endoscopy procedures for patients who underwent HBOT for UC.
  Computer Vision	In Flight	CT Ablation Tumor Site Identification	Development of a novel AI technique for identifying the tumor site within a CT ablation procedure. The tumor site will be segmented and quantified in terms of size, shape, location, microvascular invasion, and the imaging markers of degree of cirrhosis over the time of the cancer progression to reveal the prognostic pathway. Successful development of a model can provide a personized prognostic indication for hepatocellular carcinoma (HCC) patients following ablation.
  Computer Vision	Completed - In Production	DICOM Image De-identification	Image de-identification for utilization in research studies and for quality improvement. De-identification includes data within the meta-data (DICOM headers) and within the image at the pixel level. Without these models, a custom solution will have to be built for each imaging vendor and system, as well as each software version upgrade.
  Generative AI	Completed - In Production	Social Media Sentiment Analysis	Academic Research in the School of Communication - using generative AI to create a sentiment analysis of social media posts.
  Generative AI	Completed - In Production	Text to 3D Model Generation	Generating 3D models based on text prompts.
  Generative AI	Completed - In Production	Legal Document Summarization	Using generative AI to analyze and summarize public legal documents.
  Generative AI	In Flight	Incidental Pulmonary Embolism Detection	FDA-cleared AI algorithm that estimates the likelihood of a patient having incidental pulmonary embolism and then prioritizes these cases to be read by the radiologists.
  Generative AI	In Flight	Intracranial Hemorrhage Detection	FDA-cleared AI algorithm developed by Aidoc, which estimates the likelihood of a patient having intracranial hemorrhage and then prioritizes these cases to be read by the radiologists.
  Generative AI	In Flight	Pulmonary Embolism Detection	FDA-cleared AI algorithm developed by Aidoc, which estimates the likelihood of a patient having pulmonary embolism and then prioritizes these cases to be read by the radiologists.
  Machine Learning (ML)	In Flight	D3 for Dementia	Patient Digital Marker (PDM) model was developed by Indiana University to identify patients at risk of developing dementia. The model is being validated here at University of Miami and will be deployed as part of a NIH-sponsored clinical trial.
  Natural Language Processing (NLP)	Completed - In Production	Audio Sentiment Analysis for OCD Therapy	Use of a tool that uses AI to analyze the emotional content of audio recordings.  This is used for analysis of exposure therapy in people with OCD.
  Natural Language Processing (NLP)	In Flight	Custom Clinical Notes LLM for UHealth	Investigating the possibility of developing a custom large language model based on clinical notes from UHealth. There are several approaches including training a model from scratch to fine-tuning an existing LLM. The latter is the most performant approach and also the most cost effective. Different existing LLMs are being evaluated and testing is ongoing.

Accordion Group

Open All Tabs

• AI Projects for Revenue Cycle

  Type of AI Project	Status	Project Name	Description
  Robotics Process Automation (RPA)	In Flight	Automated Infusion Charging Calculator	Leveraging charge automation with Epic for CTU Infusion hospital charging
  Robotics Process Automation (RPA)	Completed - In Production	Automated Soft-Coded ED Procedure Charging	Leveraging charge automation with Epic to eliminating a manual step for emergency room procedures, eliminating a manual step for every account with soft-coded procedures.
  Robotics Process Automation (RPA)	Completed - Not In Use	Apple Health Submission for International and Healthcare Tourists	Investigating the ability to digitalize our intake process for international and domestic healthcare tourists to simplify the process for patients and our staff. This would provide an option for patients to submit data directly from their Apple Health into UHealth and would potentially support abstractions of PDFs and language translation services.
  Natural Language Processing (NLP)	Completed - In Production	SVC Coding Automation - Ambulatory Services (Retail)	Epic Simple Visit Coding (SVC) logic to automate outpatient coding on hospital visits
  Natural Language Processing (NLP)	Completed - In Production	SVC Coding Automation - - Ambulatory Services (Retail)	Epic Simple Visit Coding (SVC) logic to automate outpatient coding on hospital visits
  Natural Language Processing (NLP)	Completed - In Production	SVC Coding Automation- Executive Medicine	Epic Simple Visit Coding (SVC) logic to automate outpatient coding on hospital visits
  Natural Language Processing (NLP)	Completed - In Production	SVC Coding Automation - Family Medicine	Epic Simple Visit Coding (SVC) logic to automate outpatient coding on hospital visits
  Natural Language Processing (NLP)	Completed - In Production	SVC Coding Automation - Genetics	Epic Simple Visit Coding (SVC) logic to automate outpatient coding on hospital visits
  Natural Language Processing (NLP)	Completed - In Production	SVC Coding Automation- Neurological Surgery	Epic Simple Visit Coding (SVC) logic to automate outpatient coding on hospital visits
  Natural Language Processing (NLP)	Completed - In Production	SVC Coding Automation - Neurology	Epic Simple Visit Coding (SVC) logic to automate outpatient coding on hospital visits
  Natural Language Processing (NLP)	Completed - In Production	SVC Coding Automation - Obstetrics and Gynecology	Epic Simple Visit Coding (SVC) logic to automate outpatient coding on hospital visits
  Natural Language Processing (NLP)	Completed - In Production	SVC Coding Automation - Ophthalmology	Epic Simple Visit Coding (SVC) logic to automate outpatient coding on hospital visits
  Natural Language Processing (NLP)	Completed - In Production	SVC Coding Automation - Orthopedics	Epic Simple Visit Coding (SVC) logic to automate outpatient coding on hospital visits
  Natural Language Processing (NLP)	Completed - In Production	SVC Coding Automation - Otolaryngology	Epic Simple Visit Coding (SVC) logic to automate outpatient coding on hospital visits
  Natural Language Processing (NLP)	Completed - In Production	SVC Coding Automation - Pathology	Epic Simple Visit Coding (SVC) logic to automate outpatient coding on hospital visits
  Natural Language Processing (NLP)	Completed - In Production	SVC Coding Automation - Pediatrics	Epic Simple Visit Coding (SVC) logic to automate outpatient coding on hospital visits
  Natural Language Processing (NLP)	Completed - In Production	SVC Coding Automation - Physical Therapy - Physical Medicine & Rehabilitation	Epic Simple Visit Coding (SVC) logic to automate outpatient coding on hospital visits
  Natural Language Processing (NLP)	Completed - In Production	SVC Coding Automation - Psychiatry & Behavioral Science	Epic Simple Visit Coding (SVC) logic to automate outpatient coding on hospital visits
  Natural Language Processing (NLP)	Completed - In Production	SVC Coding Automation - Urology	Epic Simple Visit Coding (SVC) logic to automate outpatient coding on hospital visits
  Robotics Process Automation (RPA)	Completed - In Production	Account Creation & Merger Automation - Hospital Account Auto Create	Hospital and Guarantor account auto-create and merge in tandem.
  Robotics Process Automation (RPA)	Completed - In Production	Account Creation & Merger Automation - Guarantor Auto Create	Hospital and Guarantor account auto-create and merge in tandem.
  Robotics Process Automation (RPA)	Completed - In Production	Account Creation & Merger Automation - Hosipital Account and Guarantor Merge/Link	Hospital and Guarantor account auto-create and merge in tandem.
  Robotics Process Automation (RPA)	Completed - In Production	Payment Posting Automation - Automated Payment Posting	Automated payment posting, reconciliation, and transfers in tandem.
  Robotics Process Automation (RPA)	Completed - In Production	Payment Posting Automation - Automated Reconciliation (Including Adjustment Threshold)	Automated payment posting, reconciliation, and transfers in tandem.
  Robotics Process Automation (RPA)	Completed - In Production	Payment Posting Automation - Automated HB/PB Transfer for Relevant Balances	Automated payment posting, reconciliation, and transfers in tandem.
  Robotics Process Automation (RPA)	Completed - In Production	Patient Payment Notification - Instant MyChart Balance Notification on Post	Complete MyChart balance notifications and self-service payment plans in tandem.
  Robotics Process Automation (RPA)	Completed - In Production	Patient Payment Notification - Open Self-Service Payment Plan Enrollment	Complete MyChart balance notifications and self-service payment plans in tandem.
  Robotics Process Automation (RPA)	Completed - In Production	Patient Payment Notification - Auto-Apply New Balance to Open Payment Plan	Complete MyChart balance notifications and self-service payment plans in tandem.
  Robotics Process Automation (RPA)	Completed - In Production	Authorization Determination & Collection - Authorization Determiniation (Across Services)	Authorization determination and auth collections in tandem.
  Robotics Process Automation (RPA)	Completed - In Production	Authorization Determination & Collection - Authorization Collection and Status Update	Authorization determination and auth collections in tandem.
  Robotics Process Automation (RPA)	Completed - In Production	Claim Submission & Attachments - Automated Claim Outbound Submisson	Complete claim submission and attachments in tandem.
  Robotics Process Automation (RPA)	Completed - In Production	Claim Submission & Attachments - Automated Claim Attachments to avoid preventable denials	Complete claim submission and attachments in tandem.
  Robotics Process Automation (RPA)	Completed - In Production	Visit Payment & Debt Transfer - Automated Patient Payment for Visit Balance	Complete visit auto pay and bad debt transfer in tandem.
  Robotics Process Automation (RPA)	Completed - In Production	Visit Payment & Debt Transfer - Automated Bad Debt Transfer to Partners and Finance Reconciliation	Complete visit auto pay and bad debt transfer in tandem.
  Generative AI	In Flight	Coding Assistant	Working with Epic R&D to pilot and develop AI applications within Epic to assist professional coders in their review of clinical documentation and suggesting procedure and diagnosis codes for approval. Project focus is on specialties without established coding automation tools that are currently highly manual, full review coding practices.
  Speech	In Flight	Scheduling Voice Bot	Leveraging conversational AI to provide an optimal experience for patient scheduling management, including schedule, reschedule, confirm, cancel, recall.
  Machine Learning (ML)	In Flight	Likelihood to Pay	Identifies the likelihood of payment for insured and self-pay patients.
  Robotics Process Automation (RPA)	Completed - In Production	Automate Pulling Insurance Card Information	Using optical character recognition to pull insurance information from scanned patient insurance cards into the visit registration to reduce manual error and accelerate the financial clearance process.
  Machine Learning (ML)	Completed - In Production	History Note Nudges	Delivers real time clinical insights ("nudges") to help clinicians document a complete patient history before a note is saved to the EHR.  Reduces retrospective queries.  Utilizes AI powered worklist prioritization to drive coder and CDI team efficiency.
  Robotics Process Automation (RPA)	Completed - In Production	Soft-Coded Procedure Automation	Leveraging charge automation with Epic to eliminating a manual step for every account with soft-coded procedures.  This expands on the soft-coded emergency room procedure charging.
  Generative AI	Completed - In Production	Radiology Professional Coding Automation	Automates professional coding through translation of the electronic health record into accurate set of medical codes and triages unautomated cases back to coding specialists.
  Machine Learning (ML)	Completed - In Production	PB Account Prioritization	Utilizing machine learning algorithms to assign account prioritization and return the scoring back into Epic for PB billing workqueues.
  Machine Learning (ML)	Completed - In Production	HB Account Prioritization	Utilizing machine learning algorithms to assign account prioritization and return the scoring back into Epic for HB billing workqueues.
  Computer Vision	Completed - In Production	Welcome Facial Recognition	Enhance Welcome kiosks experience with facial recognition.
  Computer Vision	Completed - In Production	MyChart Account Recovery	Use facial recognition for MyChart account recovery.
  Machine Learning (ML)	Completed - In Production	Hospital Coding Prioritization	Prioritize work within hospital coding workqueues.
  Machine Learning (ML)	Completed - In Production	LOS Calculator Ambulatory Clinics	Doctor visit Level of Service calculators for ambulatory clinics.
  Machine Learning (ML)	Completed - In Production	LOS Calculator - Surgery Center	Doctor visit level of service calculators for surgery center visits.
  Machine Learning (ML)	Completed - In Production	Cost Center Automation	Bill Area/Cost Center auto-determination.
  Machine Learning (ML)	Completed - In Production	Coding Workqueue Prioritization	Prioritize work within coding workqueues.
  Machine Learning (ML)	Completed - In Production	Procedure Authorization - Neuro & ENT	Enhance procedure authorization intake in Neurology and ENT.
  Machine Learning (ML)	Completed - In Production	Scheduling Decision Support	Decision Support for scheduling requirements.
  Machine Learning (ML)	Completed - In Production	Automated Wait List Entries	Automate patient wait list entries.
  Natural Language Processing (NLP)	Completed - In Production	Coding Bots - Interventional Radiology	Autonomous coding bots to review results and determine procedure and diagnosis code for interventional radiology procedures.
  Natural Language Processing (NLP)	Completed - In Production	Coding Bots - Cardiology	Autonomous coding bots to review results and determine procedure and diagnosis codes for cardiology procedures.
  Natural Language Processing (NLP)	Completed - In Production	Coding Bots - Surgery	Autonomous coding bots to review results and determine procedure and diagnosis codes for surgical procedures.
  Natural Language Processing (NLP)	Completed - In Production	Scheduling Guidance - Neuro & ENT	Guided scheduling in Neurology and ENT.
  Robotics Process Automation (RPA)	Completed - In Production	External Orders	Improve scheduling with connection to external system for orders.
  Natural Language Processing (NLP)	Completed - In Production	Scheduling Bot	Patient self-service scheduling bot.
  Natural Language Processing (NLP)	Completed - In Production	Registration Bot	Patient self-service registration bot.
  Natural Language Processing (NLP)	Completed - In Production	Insurance Eligibility Verification	Insurance eligibility verification
  Robotics Process Automation (RPA)	Completed - In Production	Self-Pay Estimation Automation	Automatic patient self-pay estimates
  Robotics Process Automation (RPA)	Completed - In Production	Corporate Guarantor Creation & Assignment	Generate and assign corporate guarantor accounts for GameChanger.
  Robotics Process Automation (RPA)	Completed - In Production	Charge Automation - Lab Venipuncture	Automate charging for lab venipuncture.
  Robotics Process Automation (RPA)	Completed - In Production	Account Creation - Lab Venipuncture	Lab venipuncture outreach account creation.
  Robotics Process Automation (RPA)	Completed - In Production	Retrigger Late Charges	Late charge coding retrigging
  Robotics Process Automation (RPA)	Completed - In Production	Automated Pull in and Copy - Late Charge Coding	Automate pulling in and copying for late charge coding.
  Robotics Process Automation (RPA)	Completed - In Production	Paper Modifier Automation	Automated payer modifier attachments
  Robotics Process Automation (RPA)	Completed - In Production	External Charging	Enhance medical coding with charge interface to external locations.
  Robotics Process Automation (RPA)	Completed - In Production	Account Merger Automation	Automated account coding merger.
  Robotics Process Automation (RPA)	Completed - In Production	Automated Pull in and Copy - Account Coding	Automate pulling in and copying for account coding.
  Robotics Process Automation (RPA)	Completed - In Production	Nightly Denial Automation	Automate nightly denial actioning.
  Robotics Process Automation (RPA)	Completed - In Production	Credit & Write-Off Automation	Automate credit resolution and small balance write-offs.
  Robotics Process Automation (RPA)	Completed - In Production	PCP Referral Follow-Up	Enhance patient scheduling with PCP referral follow-up.
  Robotics Process Automation (RPA)	Completed - In Production	Automated Visit Reminders	Automate patient visit reminders.