🧠 Sleep Stage Analysis using EEG Data

This project analyzes EEG data from 10 subjects—including 5 normal, 3 insomniacs, and 2 narcoleptic—sourced from the CAP Sleep Database. We focus on extracting, visualizing, and comparing sleep stage patterns using EEG signal processing and various analysis techniques.

---

📂 Dataset

• Source: CAP Sleep Database
• Subjects:
  • 5 Normal
  • 3 Insomniacs
  • 2 Narcoleptics
• Files Used:
  • .edf - EEG signal data
  • .txt - Sleep stage annotations
  • .edf.st - MCAP events

---

🎯 Project Goals

• Extract sleep stage annotations
• Create hypnograms to visualize stage transitions
• Perform band power analysis (Delta, Theta, Alpha, Beta)
• Plot Power Spectral Density (PSD)
• Generate EEG frequency heatmaps
• Compare subject types: Normal vs Insomniac vs Narcoleptic

---

🛠️ Preprocessing & Annotation Extraction

• Parse .txt files to extract labeled sleep stages
• Convert annotations into DataFrames
• Align annotations with EEG samples for stage-wise analysis

---

🛌 Hypnograms

✅ Normal Subject

• Clear cyclic progression through sleep stages
• Prominent deep sleep (S3) early in the night
• REM sleep appears more in later hours
• Sparse and even MCAP distribution

❗ Insomniac Subject

• Long wakeful periods
• Fragmented transitions and reduced REM
• Clustered MCAP events (micro-arousals)

⚡ Narcoleptic Subject

• Abrupt transitions between Wake, REM, and light sleep
• Early REM onset, fragmented cycles
• Abnormal Delta wave presence during wake stages

---

📊 Band Power Analysis

Band	Frequency Range	Significance
Delta	0.5–4 Hz	Deep sleep (S3)
Theta	4–8 Hz	Light sleep (S1, S2)
Alpha	8–13 Hz	Relaxed wakefulness
Beta	13–30 Hz	Alertness, stress, focus

Normal Subject

• High Delta and Theta during deep and light sleep
• Minimal Alpha and Beta → restful brain state

Insomniac Subject

• Elevated Alpha and Theta during sleep → wake-like activity
• Reduced Delta → shallow, fragmented sleep

Narcoleptic Subject

• Strong Delta during wake and REM
• Frequent Theta/Alpha in unusual stages → unstable transitions

---

📈 Power Spectral Density (PSD)

Used to observe how power is distributed across frequency bands.

Normal Subject

• Peak at Delta (0–5 Hz), smooth drop off
• Minimal high-frequency activity

Insomniac Subject

• Peak at Delta, but with an Alpha spike
• Indicates disrupted sleep patterns

Narcoleptic Subject

• Sustained Delta power even during wakefulness
• Irregular transitions and REM intrusions

---

🌡️ EEG Frequency Heatmaps

Normal Subjects

• High Delta wave power across all brain regions
• Balanced activity with minimal Beta

Insomniac Subjects

• Delta dominant but scattered
• Low Alpha/Beta → shallow sleep, less depth

Narcoleptic Subjects

• Delta appears where it shouldn’t (e.g., wake)
• Weak high-frequency activity → irregular REM activity

---

🔥 Time Domain & Channel Heatmaps

Normal Subject

• Smooth Delta dominance
• Balanced channel activity across regions

Insomniac Subject

• Fragmented Delta
• Uneven distribution → sleep instability

Narcoleptic Subject

• Delta present in wake or S1
• P4-O2 shows irregular REM-related activity

---

✅ Conclusion

This study highlights distinct sleep characteristics for:

• Normal subjects: Stable and restorative cycles
• Insomniacs: Fragmented and shallow sleep
• Narcoleptics: Rapid REM onset, irregular transitions

By analyzing EEG signals through hypnograms, band power, PSD, and heatmaps, we gain deep insights into the neurophysiology of sleep disorders.

---

🧰 Tools & Libraries

• Python
• NumPy, Pandas
• Matplotlib, Seaborn
• MNE, SciPy, pyEDFlib