luarAssembly/resnet50Pneumonia

library_name: keras pipeline_tag: image-classification tags:

• medical-imaging
• chest-xray
• pneumonia
• tuberculosis
• keras
• cnn

ChestXray-Keras: Normal / Pneumonia / Tuberculosis (Image Classification)

A Keras/TensorFlow model for tri-class classification of chest X-rays: normal, pneumonia, and tuberculosis.
Built for educational and research use; not a medical device.

Highlights

• Framework: Keras (TensorFlow)
• Task: Image Classification (3 classes)
• Strong ranking ability (ROC-AUC = 0.928) and solid overall performance (Accuracy = 0.753)

Intended Use & Limitations

• Intended use: research, benchmarking, model-building demos.
• Not for clinical use without proper validation and regulatory approval.
• Domain shift (different hospitals/devices/contrast) can degrade performance.
• Ethical note: False negatives for tuberculosis or pneumonia can be high-risk in clinical workflows. Use with human oversight.

Labels

id2label = {0: "normal", 1: "pneumonia", 2: "tuberculosis"}
label2id = {"normal": 0, "pneumonia": 1, "tuberculosis": 2}