SEPO-WBC

Automated White Blood Cell Analysis Software for Veterinary Hematology

---

📌 Introduction

SEPO-WBC is a deep learning-based software that automatically detects and classifies white blood cells in canine blood smear images to calculate differential counts.

• Two-stage AI pipeline: YOLOv8 (detection) + DenseNet-201 (classification)
• Supports 6 cell types: Band, Segment, Lymphocyte, Monocyte, Eosinophil, nRBC
• 55x faster than manual analysis with expert-level accuracy

---

🔬 Architecture

---

📊 Performance

Stage	Metric	Score
Detection	mAP50	0.935
Classification	Accuracy	0.878
Clinical Agreement	ICC	0.969

---

🚀 How to Run

1. Executable File

1. Download SEPO-WBC.exe from Releases
2. Run the executable

2. Python Environment

# Clone
git clone https://github.com/ISW-LAB/Blood-Cell-Detection-Classification.git
cd Blood-Cell-Detection-Classification

# Install
pip install -r requirements.txt

# Run
python main.py

Requirements: Python 3.8+, NVIDIA GPU (optional)

---

📁 Project Structure

SEPO-WBC/
├── main.py                    
├── requirements.txt
├── models/
│   ├── yolov8_best.pt        # Detection model
│   └── densenet201_best.pth  # Classification model
├── images/                    # Sample images
├── config/                    # Settings
├── core/                      # AI services
└── ui/                        # GUI components

---

📖 How to Use

1️⃣ Load Images

Click [Load Images] to upload blood smear images.

2️⃣ Run Detection & Classification

Click [Selected Image Detection & Classification] to analyze.

3️⃣ Export Differential Counts

Click [Export Class Counts] and enter total WBC count to save differential count results.

📋 Reference Intervals (Canine)

The differential count results are compared against the reference intervals below and visualized as Low, Normal, or High status.

Cell Type	Reference Interval (cells/µL)
Total WBC	5,050 – 16,760
Segmented Neutrophil	2,950 – 11,640
Lymphocyte	1,050 – 5,100
Monocyte	260 – 1,120
Eosinophil	60 – 1,230

Reference: IDEXX ProCyte Dx Hematology Analyzer

📧 Contact

If you have any questions or provide your cell images, please contact us by email

• Sohui Shin: sohi8451@cbnu.ac.kr
• Kyungchang Jeong:kc.jeong-isw@cbnu.ac.kr
• Euijong Lee^†: kongjjagae@cbnu.ac.kr