Tracking Cognitive Decline Through Speech Patterns

Inspiration

Alzheimer’s disease affects language early, often before severe memory loss is noticeable. My grandma suffers from Alzheimer’s, and one of her first symptoms was a decline in her speech and difficulty holding conversations. We only discovered much later what was causing it, which inspired me to create a project exploring how speech decline is a key early indicator of Alzheimer’s. By paying attention to small changes in language, we might be able to detect Alzheimer’s earlier.

What it does

This project analyzes text samples to identify linguistic patterns associated with Alzheimer’s disease. It extracts simple features such as word count, vocabulary diversity, and repetition, then compares these features between Alzheimer’s and control groups to observe trends. The goal is to explore whether basic language statistics can reveal meaningful differences in speech behaviour.

How we built it

The analysis was built using Python in Google Colab. Text samples were processed using basic tokenization, and linguistic features were computed for each sample. Group-wise averages and visualizations were generated using pandas and matplotlib to compare patterns between diagnostic groups. The project emphasizes interpretability and reproducibility rather than predictive modelling.

Challenges we ran into

One challenge was ensuring that feature definitions were both simple and meaningful. It was important to avoid overly complex methods while still capturing relevant linguistic behaviour. Debugging data processing steps and ensuring consistency across samples also required careful attention.

Accomplishments that we're proud of

• Successfully extracting interpretable linguistic features from real-world text data
• Producing clear visual comparisons between Alzheimer’s and control groups
• Building a fully reproducible analysis pipeline with clear documentation
• Completing a research-focused project independently within a short timeframe ## What we learned This project reinforced the importance of interpretability in biomedical applications. Even simple features can provide insight when chosen thoughtfully. I also gained experience working with real datasets, structuring analyses for reproducibility, and communicating results clearly. ## What's next for Tracking Cognitive Decline Through Speech Patterns
• Expanding feature extraction to sentence structure and syntax
• Training a lightweight classifier to quantify separability
• Validating results on external datasets
• Incorporating audio-based features such as pauses and speech rate

Built With

• colab
• collections.counter
• google
• matplotlib
• numpy
• pandas
• python