Internal Medicine 2023

Artificial intelligence-based analysis of multimodal retinal images in patients with Alzheimer’s Disease and related dementia


Dr. Jinghua Chen

Email
jinghuachen@ufl.edu

Phone
(857) 389-2728

Faculty Department/Division
General Internal Medicine

This project is primarily:
Clinical

Research Project Description:
Background
Alzheimer’s disease is the most common type of dementia. It is a progressive disease beginning with mild memory loss and possibly leading to loss of the ability to carry on a conversation and respond to the environment. It can seriously affect a person’s ability to carry out daily activities. In 2020, as many as 5.8 million Americans were living with Alzheimer’s disease (AD).1 This number is projected to nearly triple to 14 million people by 2060.1 Diagnosis of Alzheimer’s disease is complex and typically involves expensive and sometimes invasive tests not commonly available outside of highly specialized clinical settings. For example, biomarkers of amyloid β and phosphorylated tau measured through cerebrospinal fluid assessments, PET scans, MRI and plasma assays are helpful for Alzheimer’s disease diagnosis, but these tests are invasive, time-consuming, expensive, and impractical for mass screening for AD.2 It is necessary to develop a rapid, inexpensive, and straightforward tool to identify AD. The retina might be an important site of inquiry because it shares the same embryonic precursor as the brain and displays several structural and functional similarities with the organ.3
Retinal changes in Alzheimer’s disease have been shown in post-mortem histopathological studies.4,5 Clinical studies using OCT/OCTA and color fundus photo show a range of retinal changes in patients with Alzheimer’s disease, such as changes in the retinal vasculature (e.g., vessel calibre and retinopathy signs), the optic nerve, and the retinal nerve fiber layer.6,7 Artificial intelligence (AI), particularly machine learning, has recently emerged as a powerful tool in the field of neuroscience. AI has been used to analyze OCT and retinal images.8-13
Fundus fluorescein angiography (FA) is the gold standard test to illustrate the abnormal vasculature of retina, help to identify vascular biomarkers for AD. There is no published data to use FA to assist the diagnosis of AD.
Purpose
To develop an AI algorithm to have high diagnostic performance to identify biomarkers for Alzheimer’s Disease and related dementia on optical coherence tomography (OCT), color fundus photo and fluorescein angiography (FA).

Method
We are going do a retrospective study to review all the available color fundus photos, FA images, OCT/OCTA images and clinical data of the patients with and without AD at the University of Florida (UF) Health Hospital between 2012 and 2022. We develop our model with consideration of having two types of inputs. The first input type is images. It has OCT/OCTA, angiogram and fundus images as image input, and the second input type is medical data as textual input including name, age, diagnosis.
We will build deep neural networks (DNN) to take one or multiple retinal image modalities as input. The DNNs will automatically extract low- to high-level features from retinal images from multiple modalities. The output feature maps from the convolutional layers will be concatenated with clinical data including demographics, medical history, and other medical measurements available in the electronic health record. The concatenated feature vector will pass through a number of fully connected layers to predict the AD diagnosis.

Data analysis
We will split the data from UF health as training and test set with a ratio of 70% to 30% and perform cross-validation in the training data to search for the optimal hyper-parameters of the deep neural networks. The trained model will be evaluated on the test data to summarize its performance metrics, including accuracy, sensitivity, specificity, precision, recall, F1 score, and ROC curve with Area Under the Curve (AUC). We will also compare with state-of-the-art method to diagnose AD from retinal images.

The role of medical student
To submit IRB and review and collect the patients’ OCT/OCTA, color fundus photos, auto fluorescein and FA results and medical history, eye examination and diagnosis.

Significance
This project may provide a new diagnostic tool to help clinicians to early detect and treat AD or identify patients at risk and reduce the burden of AD patients on societies and their families. The biomarkers identified by AI may also help to monitor the progression of the disease or response to treatments.

Does this project have an international component or travel?
No

Using AI to Assess Medical Student Communication

Name:
Dr. Heather Harrell

Email
harrellh@ufl.edu

Phone
(352) 273-7925

Faculty Department/Division
General Internal Medicine

This project is primarily:
Clinical

Research Project Description:
This is a medical education project but there was no med ed category. These are really 2 separate projects.

Specific Aim: Develop an assessment of verbal and written communication by medical students in standardized patient encounters using AI technology.
Background: Effective communication skills are a core competency for clinicians. UF College of Medicine has the following institutional learning outcome, “Graduates must be able to communicate effectively, respectfully, in a culturally sensitive manner with patients, their families, and with other members of the healthcare team.” Assessment of these skills can be labor intensive as it requires direct observation and/or reading medical notes. Standardized patient (SP) encounters with checklists are used to assess a variety of clinical skills that rely on direct observation. However, they are limited by the SPs recall. Consequently, checklists often focus more on the data gathered rather than the communication skills used when obtaining the data.
All SP examinations are video recorded for quality control, teaching, and feedback. Many strategies have been tried to evaluate student notes generated during the SP encounters but they are not always evaluated and have even been placed on hold due to the significant resources required to provide useful feedback. Thus, we have a large repository of verbal, nonverbal, and written student communications from standardized clinical encounters. The rapidly developing filed of machine learning and augmented intelligence present a potential solution to assessing communication skills in a standardized setting.
Methods: Recordings from stored medical student SP encounters will be transcribed and de-identified. Written notes also will be de-identified. The machine will be taught how to assess the appropriateness of the communication through the use of tools and lists that identify communication components (e.g. language level, jargon, empathetic statements) which will be supplemented by the research team. For written communications, key features of acceptable and ideal notes will be developed by the research team. Actual notes will be used to teach the machine how to discriminate between levels of performance. Categories of strengths and weaknesses will be constructed using performance patterns and linked to feedback suggestions, such that the machine will not only provide a score but narrative feedback.

Medical student role: project design, data interpretation, manuscript preparation (depends on how far we are by time the MSRP begins

Does this project have an international component or travel?
No

“Being” and “Becoming” a Physician: A longitudinal study of professional identity development among medical students


Dr. Melanie Hagen

Email
melanie.hagen@medicine.ufl.edu

Phone
(352) 222-4895

Faculty Department/Division
General Internal Medicine

This project is primarily:
Clinical

Research Project Description:
Rebecca Henderson PHD and UF COM Class of 2024 , and Dr. Melanie Hagen, along with a team from Florida Atlantic University, are following a cohort of 28 medical students from the UF COM and the FAU COM to understand their longitudinal professional identify formation using periodic interviews. We are analyzing the qualitative data from these interviews. We are using the conceptual framework of Jarvis-Selinger et al to describe how three key elements of student experience–focus, catalyst and context–serve to help medical students process and navigate their emerging professional identity. So far, we have submitted a paper on the effect of the Covid 19 epidemic on the students.

The summer medical student’s role would be to analyze the transcripts from the interviews using QSR Nvivo software and help with thematic qualitative analysis to identify themes.

Jarvis-Selinger S, MacNeil KA, Costello GRL, Lee K, Holmes CL. Understanding Professional Identity Formation in Early Clerkship: A Novel Framework. Acad Med. 2019 Oct;94(10):1574-1580.

Does this project have an international component or travel?
No