🧠📘 Cracking the Neural Code: How Math 📐 Interferes with English Reading Using fNIRS & Deep Learning 🤖

🔍 Introduction: When Numbers Clash with Words

Have you ever read a sentence like:

"Tom had 5 apples 🍎 and gave away 2. How many are left?"

Suddenly, your brain switches gears—from reading mode to calculating mode. This mental tug-of-war is called mathematical interference, and it can disrupt smooth language processing.

🧠 This study uses fNIRS (functional Near-Infrared Spectroscopy) and cutting-edge deep learning models to uncover how the brain handles this conflict between math and English reading.

🎯 Objective

To detect, analyze, and predict the brain's response to mathematical content embedded in English reading, using:

🌈 Real-time fNIRS signals
🔁 Deep learning models (CNNs, RNNs, Transformers)
📊 Behavioral and cognitive metrics

🧪 Methodology

👥 Participants

💡 Adults fluent in English
📚 Reading tasks with varying math content

📖 Task Types

Pure Language (e.g., “She walked to the park.”)
Pure Math (e.g., “6 × 4 = ?”)
Math-Embedded Sentences (e.g., “Anna has 3 pencils ✏️, buys 2 more. How many now?”)

🧠 Data Acquisition

🧴 fNIRS headset records oxygenated/deoxygenated hemoglobin signals in the prefrontal and parietal cortex
🎯 Eye-tracking + reaction times for validation

🧼 Data Preprocessing

🧹 Filter out noise & motion artifacts
🧮 Normalize signals for fair comparison
🧠 Segment based on task transitions (language → math)

🤖 Deep Learning Architecture

We built two main models:

🔗 Model 1: CNN + RNN Hybrid

Captures spatial and temporal patterns in brain activity
Detects real-time switches between reading and calculating

🧠 Model 2: Transformer with Attention

Focuses on key interference points
Identifies when and where math disrupts reading

📈 Increased activity in the DLPFC (executive control center) during interference
🧩 Patterns in fNIRS signals that correlate with comprehension delays
📉 Longer reaction times and more mistakes in math-embedded reading tasks

📈 Preliminary Insights

🧠 Your brain lights up differently when solving "math inside a sentence" vs. plain reading or calculating alone.

Language-only tasks activate classic reading areas 🗣️
Math-only triggers parietal regions 🔢
Mixed tasks show crossover + increased cognitive load ⚖️

🌐 Applications

🎓 Education

Smart reading apps that adapt in real-time to student’s cognitive state
Early detection of dyslexia or dyscalculia

🧠 Neurofeedback

Real-time brain-aware systems for learners or readers under cognitive load

🤖 AI + Neuroscience

Neuroadaptive interfaces in e-learning powered by deep learning & brain data

⚠️ Limitations

fNIRS doesn’t go deep into the brain (limited to cortex)
Needs large datasets for deep learning generalization
Interpretation of neural signals can be noisy or overlapping

🚀 Future Work

Combine fNIRS with EEG for richer neural tracking
Test on children, bilinguals, and individuals with math-related learning challenges
Expand model to real-world tasks (e.g., reading financial documents 💸)

🧠💬 Conclusion

Math isn’t just about numbers—it interacts with language in ways that strain the brain’s circuits. This study reveals how math concepts embedded in English create cognitive interference, using the synergy of neuroscience tools and AI models.