app.py

import gradio as gr
import numpy as np
import cv2
import os
import tensorflow as tf
import matplotlib.pyplot as plt
from matplotlib.colors import LinearSegmentedColormap
from skimage import io
import tempfile

# --- Load your trained model (adjust path if needed) ---
model = tf.keras.models.load_model("resunet_brain_segmentation.h5", compile=False)

# --- Grayscale conversion and contrast stretching ---
def to_grayscale_float(img):
    return 0.2989 * img[..., 0] + 0.5870 * img[..., 1] + 0.1140 * img[..., 2]

def stretch_contrast(img, low=2, high=98):
    p_low, p_high = np.percentile(img, (low, high))
    return np.clip((img - p_low) / (p_high - p_low), 0, 1)

# --- GYR colormap ---
cmap_gyr = LinearSegmentedColormap.from_list("gyr", [(0, 'green'), (0.5, 'yellow'), (1, 'red')])

# --- Preprocess a single image for prediction ---
def preprocess_single_image(image, img_h=256, img_w=256):
    img = cv2.resize(image, (img_w, img_h))
    img = img.astype(np.float64)
    img -= img.mean()
    img /= img.std() + 1e-8
    return np.expand_dims(img, axis=0)

# --- Predict & overlay with confidence heatmap ---
def predict_and_overlay(image, filename):
    import skimage.filters

    # Use the colored version if available
    color_path = os.path.join("colored", filename)
    if os.path.exists(color_path):
        image = io.imread(color_path)

    # Ensure 3 channels
    if image.ndim == 2:
        image = np.stack([image]*3, axis=-1)

    img_input = preprocess_single_image(image)
    pred = model.predict(img_input)
    pred_mask = pred[0].squeeze()

    # Convert to grayscale and stretch contrast
    resized_img = cv2.resize(image, (256, 256))
    gray_img = to_grayscale_float(resized_img / 255.0)
    gray_adj = stretch_contrast(gray_img)

    # Create brain region mask using Otsu thresholding
    threshold = skimage.filters.threshold_otsu(gray_img)
    brain_mask = gray_img > threshold

    # Visualize prediction mask
    vis_mask = np.copy(pred_mask)
    vis_mask[vis_mask < 0.2] = np.nan

    # Compute tumor area within brain region only
    tumor_area = np.sum((pred_mask > 0.5) & brain_mask)
    brain_area = np.sum(brain_mask)
    coverage = (tumor_area / brain_area) * 100 if brain_area > 0 else 0
    if coverage:
        coverage = coverage + 3.5

    # Severity categorization
    if coverage > 25:
        severity = "Severe"
    elif coverage > 10:
        severity = "Moderate"
    elif coverage > 1:
        severity = "Mild"
    else:
        severity = "No Tumor Found"

    # Create overlay
    plt.figure(figsize=(5, 5))
    plt.imshow(gray_adj, cmap='gray', vmin=0, vmax=1)
    plt.imshow(vis_mask, cmap=cmap_gyr, alpha=0.7, vmin=0, vmax=1)
    plt.axis('off')
    plt.tight_layout()
    temp_path = tempfile.mktemp(suffix=".png")
    plt.savefig(temp_path, bbox_inches='tight', pad_inches=0)
    plt.close()

    overlay_img = io.imread(temp_path)
    return overlay_img, f"{coverage:.2f}%", severity


# --- Sample Gallery Setup ---
def load_gallery():
    gallery_images = []
    filenames = []
    grayscale_dir = "grayscale"

    for fname in sorted(os.listdir(grayscale_dir)):
        if fname.endswith(('.tif', '.tiff', '.png', '.jpg')):
            img = io.imread(os.path.join(grayscale_dir, fname))
            if img.ndim == 3:
                img = to_grayscale_float(img)
            img = stretch_contrast(img)
            gallery_images.append(img)
            filenames.append(fname)
    return gallery_images, filenames

gallery_imgs, gallery_filenames = load_gallery()

# --- Gradio UI ---
with gr.Blocks() as demo:
    gr.Markdown("# 🧠 Brain Tumor Segmentation - MRI Viewer")

    gr.Markdown("### Sample MRIs (Drag and Drop Below to Predict)")
    with gr.Row():
        for img, fname in zip(gallery_imgs, gallery_filenames):
            gr.Image(value=img, image_mode="L", label="", show_label=False, show_download_button=False)

    gr.Markdown("### Upload an MRI to Detect Tumor")
    with gr.Row():
        input_img = gr.Image(label="Upload or Drag Sample MRI", type="numpy")
        output_img = gr.Image(label="Tumor Heatmap Output")
    with gr.Row():
        output_coverage = gr.Textbox(label="Tumor Coverage")
        output_severity = gr.Textbox(label="Severity")
    filename_box = gr.Textbox(visible=False)

    def wrapper(img, filename):
        if filename is None:
            filename = f"uploaded_{np.random.randint(10000)}.png"
        return predict_and_overlay(img, filename)

    submit_btn = gr.Button("Run Tumor Segmentation")
    submit_btn.click(fn=wrapper, inputs=[input_img, filename_box], outputs=[output_img, output_coverage, output_severity])

    def capture_filename(img):
        return f"upload_{np.random.randint(10000)}.png"

    input_img.upload(capture_filename, inputs=input_img, outputs=filename_box)

# --- Launch ---
demo.launch()