import os import subprocess import sys # import tensorflow as tf import numpy as np from PIL import Image import cv2 import gradio as gr # from numpy import asarray from transformers import pipeline from tensorflow.keras.layers import Dense, Flatten, GlobalAveragePooling2D, BatchNormalization, Dropout,AveragePooling2D import tensorflow as tf from tensorflow.keras.applications import DenseNet201 from keras.models import Model from keras.models import Sequential from keras.regularizers import * from tensorflow import keras from tensorflow.keras import layers import tensorflow as tf import matplotlib.pyplot as plt from PIL import Image import cv2 from transformers import pipeline from weasyprint import HTML from pathlib import Path # تابع پیش‌بینی def predict_demo(image, model_name): if model_name == "how dense is": image = np.asarray(image) # مدل اول def load_model(): model = tf.keras.models.load_model("model.h5", compile=False) model.compile(optimizer=tf.keras.optimizers.legacy.Adam(learning_rate=0.00001, decay=0.0001), metrics=["accuracy"], loss=tf.keras.losses.CategoricalCrossentropy(label_smoothing=0.1)) model.load_weights("modeldense1.h5") return model model = load_model() def preprocess(image): image = cv2.resize(image, (224, 224)) kernel = np.array([[0, -1, 0], [-1, 5, -1], [0, -1, 0]]) im = cv2.filter2D(image, -1, kernel) if im.ndim == 3: # اضافه کردن بعد جدید برای ورودی مدل im = np.expand_dims(im, axis=0) elif im.ndim == 2: # اگر تصویر سیاه و سفید باشد im = np.expand_dims(im, axis=-1) im = np.repeat(im, 3, axis=-1) im = np.expand_dims(im, axis=0) return im class_name = ['Benign with Density=1', 'Malignant with Density=1', 'Benign with Density=2', 'Malignant with Density=2', 'Benign with Density=3', 'Malignant with Density=3', 'Benign with Density=4', 'Malignant with Density=4'] def predict_img(img): img = preprocess(img) img = img / 255.0 pred = model.predict(img)[0] return {class_name[i]: float(pred[i]) for i in range(8)} predict_mamo= predict_img(image) return predict_mamo elif model_name == "what kind is": image = cv2.cvtColor(np.array(image), cv2.COLOR_BGR2RGB) im_pil = Image.fromarray(image) pipe = pipeline("image-classification", model="DHEIVER/finetuned-BreastCancer-Classification") def predict(image): result = pipe(image) return {result[i]['label']: float(result[i]['score']) for i in range(2)} return predict(im_pil) def generate_fixed_size_chart(predictions, image_file, chart_width=6, chart_height=5): # بارگذاری تصویر ماموگرافی mammo_image = plt.imread(image_file) # تعداد مدل‌ها num_models = len(predictions) # ایجاد figure با تنظیم عرض و ارتفاع هر زیرنمودار fig, axes = plt.subplots(1, num_models + 1, figsize=(chart_width * (num_models + 1), chart_height), constrained_layout=True) # fig.subplots_adjust(wspace=0.7) # فاصله ثابت بین نمودارها # نمایش تصویر ماموگرافی در subplot اول axes[0].imshow(mammo_image, cmap='gray') axes[0].axis('off') axes[0].set_title("Mammogram") # ایجاد نمودارهای پیش‌بینی برای هر مدل در subplots بعدی for i, (model_name, prediction) in enumerate(predictions.items(), start=1): labels, values = zip(*prediction.items()) axes[i].barh(labels, values, color='skyblue') axes[i].set_xlabel('Probability (%)') axes[i].set_title(f'{model_name}') # ذخیره‌ی نمودار در فایل chart_path = f"{os.getcwd()}/{os.path.basename(image_file)}_combined_chart.png" plt.savefig(chart_path, bbox_inches='tight') plt.close(fig) return chart_path # استفاده از pdfkit برای تولید PDF def generate_pdf(patient_info, predictions): all_charts = [] for image_file, prediction in predictions: chart = generate_fixed_size_chart(prediction, image_file) all_charts.append(chart) # تبدیل مسیر نسبی به مسیر مطلق image_path = str(Path(patient_info.get('ImagePath', '')).resolve()) # تولید محتوای HTML برای PDF html_content = f"""

Patient Report

Patient Image:

Patient Image

Patient Information:

{"".join(f"
{key}: {value if value else '-'}
" for key, value in patient_info.items() if key != "ImagePath")}

Predictions:

{"".join(f"
" for chart in all_charts)} """ # # تنظیمات PDF pdf_path = "patient_report.pdf" html = HTML(string= html_content, base_url="/app") html.write_pdf(pdf_path) return pdf_path # تابع نمایش گزارش و تولید PDF def display_report(patient_info, predictions): pdf_path = generate_pdf(patient_info, predictions) report_content = f"

Patient Report

{patient_info}

Predictions

{predictions}" return report_content, pdf_path # پردازش تصاویر ماموگرافی با مدل‌های انتخابی def process_images(patient_info, selected_models, images): all_predictions = [] total_images = len(images) # حلقه برای پردازش تصاویر for idx, image_file in enumerate(images): # به‌روزرسانی وضعیت پردازش status = f"Processing Image {idx + 1} of {total_images}..." yield all_predictions, status # خروجی موقت image = Image.open(image_file) image_predictions = {model: predict_demo(image, model) for model in selected_models} all_predictions.append((image_file, image_predictions)) # پس از اتمام پردازش final_status = "Processing Complete!" yield all_predictions, final_status # خروجی نهایی # رابط Gradio with gr.Blocks() as demo: gr.Markdown("## Breast Cancer Detection - Multi-Model Interface") # صفحه اول - اطلاعات بیمار with gr.Tab("Patient Info"): patient_image = gr.Image(label="Upload Patient Profile Image", type="pil") name = gr.Textbox(label="Name") height = gr.Number(label="Height (cm)") weight = gr.Number(label="Weight (kg)") age = gr.Number(label="Age") gender = gr.Radio(["Male", "Female", "Other"], label="Gender") residence = gr.Textbox(label="Residence") birth_place = gr.Textbox(label="Birth Place") occupation = gr.Textbox(label="Occupation") medical_history = gr.Textbox(label="Medical History") patient_info = gr.State() snackbar = gr.HTML("
Patient information was saved
", visible=False) # پیام snackbar patient_info_submit = gr.Button("Save") # صفحه دوم - انتخاب مدل‌ها و آپلود تصاویر ماموگرافی with gr.Tab("Model & Image Selection"): model_choice = gr.CheckboxGroup(["how dense is", "what kind is"], label="Select Model(s)", interactive=True) mammography_images = gr.File(label="Upload Mammography Image(s)", file_count="multiple", type="filepath") predictions = gr.State() process_button = gr.Button("Process Images") # المان برای نمایش وضعیت پردازش status_box = gr.Textbox(label="Processing Status", interactive=False) # صفحه سوم - نمایش اطلاعات و پیش‌بینی with gr.Tab("Results"): report_display = gr.HTML(label="Patient Report") download_button = gr.Button("Download Report") # جمع‌آوری اطلاعات بیمار و انتقال به مرحله بعدی def collect_patient_info(image, name, height, weight, age, gender, residence, birth_place, occupation, medical_history): # ذخیره تصویر بیمار و اضافه کردن مسیر به اطلاعات بیمار image_path = "patient_image.jpg" image.save(image_path) return { "Name": name, "Gender": gender, "Height": height, "Weight": weight, "Age": age, "Residence": residence, "Birth Place": birth_place, "Occupation": occupation, "Medical History": medical_history, "ImagePath": image_path # اضافه کردن مسیر تصویر }, gr.update(visible=True) # نمایش snackbar و مخفی شدن آن پس از 3 ثانیه patient_info_submit.click( collect_patient_info, inputs=[patient_image, name, height, weight, age, gender, residence, birth_place, occupation, medical_history], outputs=[patient_info, snackbar] # نمایش اطلاعات و snackbar ) process_button.click( process_images, inputs=[patient_info, model_choice, mammography_images], outputs=[predictions, status_box] # دو خروجی تنظیم شده است ) # نمایش گزارش بیمار و پیش‌بینی‌ها در صفحه سوم download_button.click( display_report, inputs=[patient_info, predictions], outputs=[report_display, gr.File(label="Download PDF Report")] # اصلاح خروجی برای Gradio ) demo.launch(debug=True)