🎯 Real-Time Optical Flow for Probe Trajectory Scoring

Extracting Motion Patterns from Fetal Ultrasound (US)

---

📌 Overview

This project focuses on analyzing probe movement in fetal ultrasound videos using Optical Flow and Deep Learning techniques.

The system computes motion-based stability scores and uses them to automatically generate labels, followed by training a 3D Convolutional Neural Network (3D CNN) for classification of probe trajectory quality.

---

🚀 Key Features

• 🎥 Video preprocessing and frame extraction
• 🌊 Optical Flow-based motion analysis
• 📊 Stability score computation
• 🏷️ Automatic dataset labeling
• 🗂️ Dataset organization (stable vs unstable)
• 🧠 3D CNN model for classification
• 📈 Model training and evaluation

---

🧠 Project Pipeline

Raw Ultrasound Videos
        ↓
Optical Flow Analysis (main.py)
        ↓
Motion Features + Stability Score
        ↓
Auto Label Generation (sort_dataset.py)
        ↓
Structured Dataset
   ├── stable/
   └── unstable/
        ↓
3D CNN Training (train.py)
        ↓
Classification Accuracy

---

📁 Project Structure

PBL-Project/
│
├── src/
│   ├── main.py              # Optical flow + feature extraction
│   ├── sort_dataset.py     # Label generation + sorting
│   ├── train.py            # Model training
│   ├── data_loader.py      # Dataset loading
│   ├── model_3dcnn.py      # 3D CNN model
│
├── dataset/
│   ├── stable/
│   └── unstable/
│
├── data/
│   └── labels.xlsx
│
├── Results/
│   └── final_results.xlsx
│
├── requirements.txt
└── README.md

---

⚙️ Installation

git clone https://github.com/CodeVoyager7777/PBL-Project.git
cd PBL-Project
python -m venv venv
venv\Scripts\activate
pip install -r requirements.txt

---

▶️ Usage

1️⃣ Optical Flow Analysis

python src/main.py

2️⃣ Dataset Preparation

python src/sort_dataset.py

3️⃣ Model Training

python src/train.py

---

🧪 Model Details

• Model: 3D Convolutional Neural Network (3D CNN)

• Input: Video clips

• Frame size: 64 × 64

• Frames per video: 8

• Classes:

  • Stable Probe Movement
  • Unstable Probe Movement

---

💡 Core Idea

Instead of manual labeling, this system:

Automatically generates labels using motion-based stability scores derived from optical flow analysis

This enables scalable and intelligent dataset creation.

---

📈 Future Scope

• Real-time deployment in ultrasound systems
• Integration with medical decision support
• Improved motion feature extraction
• Higher accuracy deep learning models

---

👨‍💻 Author

Laksh Makkar

---

⭐ Note

This project demonstrates a complete pipeline combining Computer Vision (Optical Flow) and Deep Learning (3D CNN) for real-world medical video analysis.