Date: Tuesday, November 11, 2025
Time: 8:00 AM – 12:00 PM local time
Location: Pacific C Ballroom
Registration: neuro.embs.org/2025/registr...

📖 Read the full paper here → www.nature.com/articles/s44...

🧠 We are now adding more cognitive domains to the assessment and testing the algorithm in a large group of individuals.

Interested in building with us? Please reach out!

📈 We used our Flow2 measurement system to record brain data during language and memory tasks, and built a machine learning classifier to assign each participant a single score encapsulating their likelihood of having a current MCI diagnosis.

Stay tuned for more exciting news from @kernelneuro.bsky.social!

Special thanks to co-authors: J. Dubois, J. Duffy, R. Field, E. Koch, Z. Aghajan, N. Miller, K. Perdue, G. Sahagian, M. Taylor

Moving forward, we hope this approach can be used to help patients with MCI get the care they need to preserve their cognition by providing quantitative and timely assessments of brain functioning. Early identification of MCI will support the fight to end Alzheimer's! #ENDALZ

Our machine learning model for classifying if an individual is likely to have MCI had excellent performance with 85% accuracy, 80% sensitivity and 90% specificity. This is better performance than current standard screening tools used in clinical practice.

We combine patterns of brain activation with the task performance and a simple questionnaire into a machine learning model that classifies subjects as either being healthy or having MCI.

However, to be clinically relevant, we need to be looking at individuals and not groups. This is where machine learning comes in!

We also found that there were differences in brain activity between the two groups during both the memory (a) and language (b) challenges. Red regions indicate more brain activity in the HC group while blue indicates more in the MCI group.