Morteza S

- Research Program Mentor

Bio

Project ideas

A Multimodal AI System for Early Dementia Classification Combining Brain Imaging, Medical Records, and Genetics

Diagnosing dementia early is one of the hardest challenges in neurology — no single test tells the full story. Clinicians must piece together information from brain scans, patient history, lab results, and increasingly, genetic data. In this project, we will build a multimodal predictive model that learns from all of these data sources together, rather than in isolation, to classify dementia risk at an early stage. We will work with paired patient datasets containing MRI scans, electronic medical records, and genetic information, learning how to align and fuse different data types into a single AI pipeline. Along the way, we will grapple with real research challenges — missing data, small sample sizes, model interpretability, and the ethical handling of sensitive patient information. Final deliverable: A trained and evaluated multimodal classification model, along with a research paper reporting model performance, key predictive features across data modalities, limitations, and recommendations for clinical translation.

Developing machine learning models for automated classification of healthy and cancerous tissues based on MRI images

The goals of this project is to develop a Python script for machine learning and write a research paper aimed at developing machine learning models for classification of MRI images. We will go over the following points: 1- Learn how to perform data analysis, image preprocessing, visualization, correlation study, data mining and exploration in Python. How to find high quality datasets online. 2- Learn how to develop and train machine learning algorithms in Python that can identify unhealthy individuals or people with high risk of cancer by looking at their MRI images. How to optimize and compare the performance of these models. 3- Write the results of findings as a research paper to be published in reputable research journals, also showcase your work on public sites such as GitHub and Kaggle. Many of such projects also were proceeded to science fairs. Image Reference: https://www.healthtravellersworldwide.com/cancer-detection-techniques/

Development of a machine-learning guided assistive device for helping disabled people

The goal of this project is to develop a prototype device using 3D printing, microcontrollers and robotics which can be presented in a science fair and published as a research paper. Throughout the project we will cover three skillsets: 1- Prototyping robotic devices with healthcare applications, using 3D printing (optional) 2 -Programming the device using Python to interact the the environment and process the inputs received from the user 3- Test the prototype and evaluate its performance to report findings as a science fair application and/or scientific paper Reference for the image: https://www.selectmedical.com/why-choose-us/advanced-robotics-and-rehabilitation-technology/

Machine learning modeling and app development for early prediction of dementia based on patient's medical records

In this project, we aim to use machine learning algorithms to predict the factors that can contribute to dementia, one of the most serious mental diseases, by analyzing the patient's medical records and history. Through this project, we will go over programming and data analysis in Python, focused on machine learning models. We look into best ways to do research and find datasets online. We will also aim to develop an online app using Python libraries such as Streamlit and Flask to run the trained machine learning models online. Finally, we'll write a research paper to summarize the findings for submission to a research journal.

Smart Health App Development: Integrating Machine Learning for Personalized Healthcare

For our research project at the intersection of app development, healthcare, and machine learning, we will design and prototype a mobile application that leverages machine learning to analyze user-inputted health data and provide personalized health insights. We will use Python as our primary programming language, utilizing libraries such as Pandas and NumPy for data manipulation and numerical analysis, Scikit-learn for building and evaluating machine learning models, and TensorFlow or Keras for deep learning tasks. For data visualization and presenting trends to users, we will incorporate Matplotlib and Seaborn. To develop the app’s backend and APIs, we will consider frameworks like Flask or Django, which are well-suited for healthcare applications due to their scalability, security features, and ease of integration with machine learning workflows. By combining these tools, we aim to create an accessible and intelligent digital health solution that empowers individuals to monitor and manage their health more effectively. Result of this project can also be presented to a Science Fair, Competition or as a research article.

A Multimodal AI System for Tracking ALS Disease Progression Turning Heterogeneous Patient Data into Actionable Clinical Insights

ALS progresses differently in every patient, making it extraordinarily difficult to predict how the disease will evolve and when to adjust care. In this project, we will build a multimodal AI system that integrates data from multiple sources — clinical assessments, speech recordings, wearable sensor data, and electronic health records — to model and predict individual ALS disease trajectories over time. We will work through the real challenges of longitudinal patient data, including irregular measurement intervals, missing data, and the need for models that clinicians can actually interpret and trust. Final deliverable: A multimodal disease progression model evaluated on longitudinal ALS patient data, along with a research paper reporting predictive accuracy, clinically meaningful insights, model interpretability, and recommendations for integration into palliative and supportive care planning.

Generative AI for Synthetic Patient Data in Parkinson's Disease Research Solving the Data Scarcity Problem in Rare Neurological Conditions

One of the biggest barriers to AI research in neurodegenerative diseases is the lack of large, diverse, and well-annotated patient datasets. In this project, we will investigate how generative AI models can be used to synthesize realistic patient data — including motor assessments, clinical notes, and imaging features — for Parkinson's disease research. We will design and evaluate a generative pipeline, rigorously test whether synthetic data can augment real datasets to improve downstream model performance, and critically examine the risks of training clinical AI systems on data that was itself AI-generated. Final deliverable: A generative data pipeline producing synthetic Parkinson's patient data, validated against real-world distributions, along with a research paper evaluating data quality, downstream model impact, and the ethical boundaries of synthetic data use in clinical AI.

LLM-Assisted Early Detection of Cognitive Decline Through Speech and Language Analysis Listening for What the Brain is Trying to Tell Us

Changes in the way people speak and write are among the earliest detectable signs of cognitive decline — yet these subtle shifts are easy to miss in a standard clinical visit. In this project, we will build an AI pipeline that analyzes speech patterns, word choice, sentence complexity, and fluency from recorded patient conversations to flag early signs of cognitive decline associated with Alzheimer's and related dementias. We will explore how LLMs can be combined with audio processing tools to extract meaningful linguistic features, and evaluate the system's ability to distinguish between healthy aging and early-stage neurodegeneration across diverse patient populations. Final deliverable: A multimodal speech and language analysis tool evaluated on clinical or publicly available datasets, along with a research paper reporting predictive performance, key linguistic biomarkers identified, and implications for scalable, non-invasive cognitive screening.

Reinforcement Learning-Based Autonomous Navigation for a Raspberry Pi Robot Building a Low-Cost, Real-World Autonomous Driving Platform

Autonomous navigation is one of the most exciting and demanding challenges in robotics — requiring a system to perceive its environment, make decisions in real time, and recover gracefully from mistakes. In this project, we will design and build a small-scale autonomous driving robot on a Raspberry Pi platform, using reinforcement learning to teach the robot to navigate real-world environments without explicit programming. Starting from scratch, we will assemble the hardware, set up the sensing pipeline using onboard cameras and sensors, and train an RL agent that learns through trial and error to follow lanes, avoid obstacles, and make navigation decisions on the fly. A key focus of this project is bridging the gap between simulation-based training and real-world deployment — one of the hardest open problems in autonomous systems research. Final deliverable: A fully functional Raspberry Pi-based autonomous robot trained using reinforcement learning and tested in a real physical environment, along with a research paper documenting the hardware setup, RL training methodology, sim-to-real transfer strategies, performance results, and lessons learned for scaling to more complex autonomous systems.

Coding skills

Languages I know

Teaching experience

Credentials

Completed Projects

Developing Optimized Machine Learning Models For Timely Prediction and Prevention of College Dropout

Arko's Project

with Morteza

Development of a Machine Learning Model for Early Detection of Amyotrophic Lateral Sclerosis (ALS) Using Speech and Stability Analysis

Arnav's Project

with Morteza