PneumoDetect Pro

(Project during AI for Healhcare - Udacity)

Objective:

Quickly and effectively identify positive pneumonia cases from radiological images (DICOM) using a machine learning model. ​

Key Steps:

Data Preprocessing: Verify DICOM data and generate ground truth for 14 common thoracic pathologies using NLP. ​ Data Augmentation: Normalize and split data into validation and training sets. ​ Model Building: Use a pre-trained model, adjust parameters, and build a custom model. ​ Training and Optimization: Compile the model, train with callbacks, and save the best weights. ​ Validation and Deployment: Evaluate performance, make predictions, and save the final model. ​

Key Results:

Classification Threshold: Set at 0.40 to maximize true positives. ​ Performance Metrics: ​

Accuracy: 0.452 ​ Precision: 0.310 ​ Recall: 0.692 ​ F1 Score: 0.429 ​

The model prioritizes high recall to capture as many positive pneumonia cases as possible, sacrificing precision. ​

This project emphasizes detecting pneumonia cases with a focus on minimizing false negatives.

See the report for more details.​