Thyroid Disease Detection

Overview

Thyroid disease is a prevalent medical condition characterized by abnormalities in thyroid hormone levels, which regulate the body's metabolism. This project aims to predict whether a patient is at risk for thyroid disease using a Random Forest Classifier, with the goal of improving early diagnosis and treatment.

Key Features:

• Predicts the likelihood of thyroid disease (Normal, Hyperthyroidism, or Hypothyroidism).
• Built using a Kaggle dataset.
• High-performance model with metrics demonstrating excellent accuracy.
• Deployed as a web application using Streamlit.

Live App Link

Dataset

• Source: Kaggle Thyroid Disease Dataset.
• Features:
  • Various medical attributes related to thyroid function and patient history.
  • Target: Classification into one of three classes: Normal (0), Hyperthyroidism (1), Hypothyroidism (2).

Model Implementation

The project uses the Random Forest Classifier to predict thyroid conditions. Key steps include:

1. Data Preprocessing:
   • Handling missing values.
   • Encoding categorical variables.
   • Scaling numerical features.
2. Feature Selection:
   • Selected features based on importance to the model.
3. Model Training:
   • Used a Random Forest Classifier with hyperparameter tuning.
4. Evaluation:
   • Evaluated model using various performance metrics.

Performance Metrics

The model achieved exceptional results:

• Accuracy: 98.44%
• Precision (weighted): 98.43%
• Recall (weighted): 98.44%
• F1 Score (weighted): 98.39%
• ROC-AUC Score: 99.84%

Classification Report:

Class	Precision	Recall	F1-Score	Support
Normal (0)	99%	99%	99%	1350
Hyperthyroidism (1)	96%	78%	86%	63
Hypothyroidism (2)	95%	98%	96%	124

Confusion Matrix:

Deployment

The model is deployed as a Streamlit web application. Users can upload patient data in CSV format and get predictions directly on the web.

How to Use:

1. Visit the deployed app.
2. Upload a CSV file containing patient data.
3. View predictions and insights instantly.

Future Improvements

1. Enhancing Recall for Hyperthyroidism:
   • Employ techniques like SMOTE or ADASYN to balance the dataset.
   • Fine-tune hyperparameters for better recall.
2. Model Explainability:
   • Integrate SHAP or LIME for better interpretability of feature importance.
3. Additional Features:
   • Incorporate more patient demographics and medical history.

Getting Started Locally

Prerequisites:

• Python 3.7+
• Required libraries (install using requirements.txt):

  pip install -r requirements.txt
  

Run Locally:

1. Clone the repository:

   git clone https://github.com/vivekd16/Thyroid-Disease-Detection.git
   

2. Navigate to the project directory:

   cd Thyroid-Disease-Detection
   

3. Start the Streamlit app:

    streamlit run app.py
   

Repository Structure

└── Thyroid-Disease-Detection/
    ├── README.md
    ├── LICENSE
    ├── app.py
    ├── requirements.txt
    ├── dataset/
    │   ├── processed/
    │   │   └── ProcessedthyroidDF.csv
    │   └── raw/
    │       └── thyroidDF.csv
    ├── model/
    │   └── thyroid_disease_model.pkl
    └── sources/
        └── thyroid_disease_detection.ipynb

Project Members

-Vivek Deshmukh | Github | Linkedin

-Maviya Mahagami | Github | Linkedin

Acknowledgments

• Kaggle Thyroid Disease Dataset