How AI decodes individual neurons to predict finger forces | Neural Interfacing Research

Can we read individual neurons to precisely control prosthetics and neural interfaces?

In this IEEE NER 2025 presentation, we demonstrate how different machine learning models can predict finger forces from motor neuron spike trains with great accuracy and generalise across different people.

👍 Like this video if you found it interesting! 🔔 Subscribe for more neural engineering research!

⏱️ TIMESTAMPS:
0:00 - Motivation
1:37 - Introduction / Current approaches
4:25 - Methods: Data collection setup
5:13 - Machine learning pipeline
7:00 - Results & models comparison
8:27 - Discussion
10:07 - Q&A

🔬 What we achieved here:
94.4% accuracy (R² = 0.944) in force prediction
First subject- and task-independent motor unit-based model
Standardised pipeline enabling large-scale neural drive databases

🧠 The science behind:
We used high-density electromyography (EMG) to non-invasively record individual motor neuron activity from 25 participants. By tracking 10 carefully selected motor units and training LSTM networks, we achieved performance that matches subject-specific models while generalising across people and fingers.

This opens pathways for:
✓ Advanced prosthetic control with intuitive interfaces
✓ Assistive technologies for motor impairments
✓ Next-generation human-computer interaction
✓ Pooling heterogeneous EMG datasets for more robust AI models

Authors:
Renato Mio¹, Jan Bodenschlägel¹, A. Aldo Faisal¹,²
¹ Chair of Digital Health & Data Science, University of Bayreuth
² Brain & Behaviour Lab, Imperial College London

📄 Citation:
R. Mio, J. Bodenschlägel, and A. A. Faisal, "Finger Force Prediction from Spinal Signals: Machine Learning Pipeline for the Neural Drive," in 2025 12th International IEEE/EMBS Conference on Neural Engineering (NER), 2025, pp. 1194-1199.

Video recording courtesy of IEEE NER 2025 Conference

#NeuralEngineering #MachineLearning #DeepLearning #NeuralInterfaces #BCI #Neuroscience #AI #IEEE #NER2025 #MotorControl #EMG

---

Presented at The 12th International IEEE/EMBS Conference on Neural Engineering (NER 2025)

💬 Questions? Drop them in the comments—I'd love to discuss the research!