app.py

import tensorflow as tf
from PIL import Image
from tensorflow import keras
import numpy as np
import os
import random
import logging
from tensorflow.keras.preprocessing import image as keras_image
from huggingface_hub import from_pretrained_keras
from openai import AzureOpenAI
import gradio as gr
from zipfile import ZipFile

logging.basicConfig(level=logging.INFO)

class DiseaseDetectionApp:
    def __init__(self):
        

        self.class_names =['Normal', 'Tuberculosis']
        self.model =tf.keras.models.load_model("chest_xray_tuberclosis_prediction_model.keras")
        self.client=AzureOpenAI()


    def predict_disease(self, image_path):
        """
        Predict the disease present in the X-Ray image.

        Args:
        - image_data: PIL image data

        Returns:
        - predicted_disease: string
        """
        try:

          # Load the image file, resizing it to the dimensions expected by the model
          img = keras_image.load_img(image_path, target_size=(256, 256))  # Adjust target_size according to your model's expected input dimensions

          # Convert the image to a numpy array
          img_array = keras_image.img_to_array(img)

          # Add an additional dimension to the array: (1, height, width, channels)
          img_array = tf.expand_dims(img_array, 0)  # Model expects a batch of images, but we're only passing a single image
          # print(img_array)
          # Make predictions
          predictions = self.model.predict(img_array)

          # Extract the predicted class and confidence
          predict_class =self.class_names[np.argmax(predictions[0])]
          confidence = round(100 * np.max(predictions[0]), 2)
          return predict_class

        except Exception as e:
            logging.error(f"Error predicting disease: {str(e)}")
            return None

    def classify_disease(self,image_path):

          disease_name=self.predict_disease(image_path)
          print(disease_name)
          if disease_name=="Tuberculosis":
            conversation = [
              {"role": "system", "content": "You are a medical assistant"},
              {"role": "user", "content": f""" your task describe(classify) about the given disease as a summary only in 3 lines.
                        ```{disease_name}```
                """}
                ]
            # Generate completion using ChatGPT model
            response = self.client.chat.completions.create(
                model="ChatGPT",
                messages=conversation,
                temperature=0,
                max_tokens=1000
            )
            # Get the generated topics message

            result = response.choices[0].message.content
            return disease_name,result

          elif disease_name=="Normal":
            result="No problem in your xray image"
            return disease_name,result



    def unzip_image_data(self,filespath):
        """
        Unzips an image dataset into a specified directory.

        Returns:
            str: The path to the directory containing the extracted image files.
        """
        try:
          with ZipFile(filespath,"r") as extract:
            directory_path = random.randrange(100)
            extract.extractall(f"{directory_path}")
            return f"{directory_path}"

        except Exception as e:
                logging.error(f"An error occurred during extraction: {e}")
                return ""

    def example_images(self,filespath):
          """
          Unzips the image dataset and generates a list of paths to the individual image files and use image for showing example

          Returns:
              List[str]: A list of file paths to each image in the dataset.
          """
          image_dataset_folder = self.unzip_image_data(filespath)
          image_extensions = ['.jpg', '.jpeg', '.png', '.gif', '.bmp', '.tiff', '.webp']
          image_count = len([name for name in os.listdir(image_dataset_folder) if os.path.isfile(os.path.join(image_dataset_folder, name)) and os.path.splitext(name)[1].lower() in image_extensions])
          example=[]
          for i in range(image_count):
            for name in os.listdir(image_dataset_folder):
                path=(os.path.join(os.path.dirname(image_dataset_folder),os.path.join(image_dataset_folder,name)))
                example.append(path)

          return example

    def get_example_image(self):
      normal_image="Normal_dataset.zip"
      tuberclosis_image="Tuberculosis_dataset.zip"

      normal_image_unziped=self.example_images(normal_image)
      tuberclosis_image_unziped=self.example_images(tuberclosis_image)

      return normal_image_unziped,tuberclosis_image_unziped

    def gradio_interface(self):

      with gr.Blocks(theme=gr.themes.Soft()) as demo:
        gr.HTML("""<center><h1>Tuberculosis Disease Detection</h1></center>""")

        normal_image,tuberclosis_image=self.get_example_image()

        with gr.Row():
          input_image =gr.Image(type="filepath",sources="upload")
          with gr.Column():
            output=gr.Label(label="Disease Name")
        with gr.Row():
          classify_disease_=gr.Textbox(label="About disease")
        with gr.Row():
          button =gr.Button(value="Detect The Disease")

        button.click(self.classify_disease,[input_image],[output,classify_disease_])

        gr.Examples(
        examples=normal_image,
        label="Normal X-ray Images",
        inputs=[input_image],
        outputs=[output,classify_disease_],
        fn=self.classify_disease,
        examples_per_page=5,
        cache_examples=False)

        gr.Examples(
        examples=tuberclosis_image,
        label="Tuberclosis X-ray Images",
        inputs=[input_image],
        outputs=[output,classify_disease_],
        examples_per_page=5,
        fn=self.classify_disease,
        cache_examples=False)


      demo.launch(debug=True)

if __name__ == "__main__":
    app = DiseaseDetectionApp()
    result=app.gradio_interface()