from pathlib import Path import numpy as np import gradio as gr import requests import json from transformers import ViTImageProcessor, ViTModel from PIL import Image # Store the server's URL SERVER_URL = "https://affordable-prot-bind-clarke.trycloudflare.com/" CURRENT_DIR = Path(__file__).parent DEPLOYMENT_DIR = CURRENT_DIR / "deployment_files" KEYS_DIR = DEPLOYMENT_DIR / ".fhe_keys" CLIENT_DIR = DEPLOYMENT_DIR / "client_dir" SERVER_DIR = DEPLOYMENT_DIR / "server_dir" USER_ID = "user_id" EXAMPLE_CLINICAL_TRIAL_LINK = "https://www.trials4us.co.uk/ongoing-clinical-trials/recruiting-healthy-adults-c23026?_gl=1*1ysp815*_up*MQ..&gclid=Cj0KCQjwr9m3BhDHARIsANut04bHqi5zE3sjS3f8JK2WRN3YEgY4bTfWbvTdZTxkUTSISxXX5ZWL7qEaAowwEALw_wcB&gbraid=0AAAAAD3Qci2k_3IERmM6U1FGDuYVayZWH" # Define possible categories for fields without predefined categories additional_categories = { "Gender": ["Male", "Female", "Other"], "Ethnicity": ["White", "Black or African American", "Asian", "American Indian or Alaska Native", "Native Hawaiian or Other Pacific Islander", "Other"], "Geographic_Location": ["North America", "South America", "Europe", "Asia", "Africa", "Australia", "Antarctica"], "Smoking_Status": ["Never", "Former", "Current"], "Diagnoses_ICD10": ["Actinic keratosis", "Melanoma", "Dermatofibroma", "Vascular lesion","None"], "Medications": ["Metformin", "Lisinopril", "Atorvastatin", "Amlodipine", "Omeprazole", "Simvastatin", "Levothyroxine", "None"], "Allergies": ["Penicillin", "Peanuts", "Shellfish", "Latex", "Bee stings", "None"], "Previous_Treatments": ["Chemotherapy", "Radiation Therapy", "Surgery", "Physical Therapy", "Immunotherapy", "None"], "Alcohol_Consumption": ["None", "Occasionally", "Regularly", "Heavy"], "Exercise_Habits": ["Sedentary", "Light", "Moderate", "Active", "Very Active"], "Diet": ["Omnivore", "Vegetarian", "Vegan", "Pescatarian", "Keto", "Mediterranean"], "Functional_Status": ["Independent", "Assisted", "Dependent"], "Previous_Trial_Participation": ["Yes", "No"] } # Define the input components for the form age_input = gr.Slider(minimum=18, maximum=100, label="Age ", step=1, value=30) gender_input = gr.Radio(choices=additional_categories["Gender"], label="Gender", value="Male") ethnicity_input = gr.Radio(choices=additional_categories["Ethnicity"], label="Ethnicity", value="White") geographic_location_input = gr.Radio(choices=additional_categories["Geographic_Location"], label="Geographic Location", value="North America") medications_input = gr.CheckboxGroup(choices=additional_categories["Medications"], label="Medications", value=["Metformin"]) allergies_input = gr.CheckboxGroup(choices=additional_categories["Allergies"], label="Allergies", value=["Peanuts"]) previous_treatments_input = gr.CheckboxGroup(choices=additional_categories["Previous_Treatments"], label="Previous Treatments", value=["None"]) blood_glucose_level_input = gr.Slider(minimum=0, maximum=300, label="Blood Glucose Level", step=1, value=100) blood_pressure_systolic_input = gr.Slider(minimum=80, maximum=200, label="Blood Pressure (Systolic)", step=1, value=120) blood_pressure_diastolic_input = gr.Slider(minimum=40, maximum=120, label="Blood Pressure (Diastolic)", step=1, value=80) bmi_input = gr.Slider(minimum=10, maximum=50, label="BMI ", step=1, value=20) smoking_status_input = gr.Radio(choices=additional_categories["Smoking_Status"], label="Smoking Status", value="Never") alcohol_consumption_input = gr.Radio(choices=additional_categories["Alcohol_Consumption"], label="Alcohol Consumption", value="None") exercise_habits_input = gr.Radio(choices=additional_categories["Exercise_Habits"], label="Exercise Habits", value="Sedentary") diet_input = gr.Radio(choices=additional_categories["Diet"], label="Diet", value="Omnivore") condition_severity_input = gr.Slider(minimum=1, maximum=10, label="Condition Severity", step=1, value=5) functional_status_input = gr.Radio(choices=additional_categories["Functional_Status"], label="Functional Status", value="Independent") previous_trial_participation_input = gr.Radio(choices=additional_categories["Previous_Trial_Participation"], label="Previous Trial Participation", value="Yes") # def encrypt_array(user_symptoms: np.ndarray, user_id: str) -> bytes: # """ # Encrypt the user symptoms vector. # Args: # user_symptoms (np.ndarray): The vector of symptoms provided by the user. # user_id (str): The current user's ID. # Returns: # bytes: Encrypted and serialized symptoms. # """ # # Retrieve the client API # client = FHEModelClient(path_dir=DEPLOYMENT_DIR, key_dir=KEYS_DIR / f"{user_id}") # client.load() # # Ensure the symptoms are properly formatted as an array # user_symptoms = np.array(user_symptoms).reshape(1, -1) # # Encrypt and serialize the symptoms # encrypted_quantized_user_symptoms = client.quantize_encrypt_serialize(user_symptoms) # # Ensure the encryption process returned bytes # assert isinstance(encrypted_quantized_user_symptoms, bytes) # # Save the encrypted data to a file (optional) # encrypted_input_path = KEYS_DIR / f"{user_id}/encrypted_input" # with encrypted_input_path.open("wb") as f: # f.write(encrypted_quantized_user_symptoms) # # Return the encrypted data # return encrypted_quantized_user_symptoms # def decrypt_result(encrypted_answer: bytes, user_id: str) -> bool: """ Decrypt the encrypted result. Args: encrypted_answer (bytes): The encrypted result. user_id (str): The current user's ID. Returns: bool: The decrypted result. """ # Retrieve the client API # client = FHEModelClient(path_dir=DEPLOYMENT_DIR, key_dir=KEYS_DIR / f"{user_id}") # client.load() # Decrypt the result # decrypted_result = client.decrypt_deserialize(encrypted_answer) # # Return the decrypted result # return decrypted_result def encode_categorical_data(data): categories = ["Gender", "Ethnicity", "Geographic_Location", "Diagnoses_ICD10", "Medications", "Allergies", "Previous_Treatments", "Smoking_Status", "Alcohol_Consumption", "Exercise_Habits", "Diet", "Functional_Status", "Previous_Trial_Participation"] encoded_data = [] for i in range(len(categories)): sub_cats = additional_categories[categories[i]] if data[i] in sub_cats: encoded_data.append(sub_cats.index(data[i]) + 1) else: encoded_data.append(0) return encoded_data def clear_data_to_json(data): print(data) patient_data = { "model_names": ["second_model"], "patient": { "Age": data.get("age", 30), "Blood_Glucose_Level": data.get("blood_glucose_level", 0), "Blood_Pressure_Systolic": data.get("blood_pressure_systolic", 0), "Blood_Pressure_Diastolic": data.get("blood_pressure_diastolic", 0), "BMI": data.get("bmi", 0), "Condition_Severity": data.get("condition_severity", 0), "Gender": data.get("Gender", 0), "Ethnicity": data.get("Ethnicity", 0), "Geographic_Location": data.get("Geographic_Location", 0), "Smoking_Status": data.get("Smoking_Status", 0), "Diagnoses_ICD10": data.get("Diagnoses_ICD10", 0), "Medications": data.get("Medications", 0), "Allergies": data.get("Allergies", 0), "Previous_Treatments": data.get("Previous_Treatments", 0), "Alcohol_Consumption": data.get("Alcohol_Consumption", 0), "Exercise_Habits": data.get("Exercise_Habits", 0), "Diet": data.get("Diet", 0), "Functional_Status": data.get("Functional_Status", 0), "Previous_Trial_Participation": data.get("Previous_Trial_Participation", 0) } } return patient_data def process_patient_data(age, gender, ethnicity, geographic_location, diagnoses_icd10, medications, allergies, previous_treatments, blood_glucose_level, blood_pressure_systolic, blood_pressure_diastolic, bmi, smoking_status, alcohol_consumption, exercise_habits, diet, condition_severity, functional_status, previous_trial_participation): # Encode the data categorical_data = [gender, ethnicity, geographic_location, diagnoses_icd10, medications, allergies, previous_treatments, smoking_status, alcohol_consumption, exercise_habits, diet, functional_status, previous_trial_participation] print(f"Categorical data: {categorical_data}") encoded_categorical_data = encode_categorical_data(categorical_data) numerical_data = np.array([age, blood_glucose_level, blood_pressure_systolic, blood_pressure_diastolic, bmi, condition_severity]) print(f"Numerical data: {numerical_data}") print(f"One-hot encoded data: {encoded_categorical_data}") combined_data = np.hstack((numerical_data, encoded_categorical_data)) ordered_categories = ["Gender", "Ethnicity", "Geographic_Location", "Diagnoses_ICD10", "Medications", "Allergies", "Previous_Treatments", "Smoking_Status", "Alcohol_Consumption", "Exercise_Habits", "Diet", "Functional_Status", "Previous_Trial_Participation"] zipped_data = zip(ordered_categories, encoded_categorical_data) # Convert the zipped data to a dictionary encoded_categorical_dict = {category: value for category, value in zipped_data} # Convert the data to JSON json_data = clear_data_to_json({ "age": age, "blood_glucose_level": blood_glucose_level, "blood_pressure_systolic": blood_pressure_systolic, "blood_pressure_diastolic": blood_pressure_diastolic, "bmi": bmi, "condition_severity": condition_severity, **encoded_categorical_dict }) print(f"JSON data: {json_data}") print(f"Combined data: {combined_data}") # encrypted_array = encrypt_array(combined_data, "user_id") # Send the data to the server url = SERVER_URL + "inference/clear-match" headers = {"Content-Type": "application/json", "X-API-KEY": "secret"} response = requests.post(url, data=json_data, headers=headers) # Check if the data was sent successfully if response.status_code == 200: print("Data sent successfully.") else: print(f"Error sending data. Status code: {response.status_code}") # return f"**[Possible Trial Link]({EXAMPLE_CLINICAL_TRIAL_LINK})**" # Decrypt the result # decrypted_result = decrypt_result(response.content, USER_ID) # decrypted_result = None try: decrypted_result = response.json() print(f"Decrypted result: {decrypted_result}") except json.JSONDecodeError as e: print(f"Error decrypting result: Invalid JSON. {e}") decrypted_result = False return "There was an error processing the data." except Exception as e: print(f"An error occurred: {e}") decrypted_result = False return "There was an error processing the data." else: print("Json parsed successfully.") # If the answer is True, return the link if decrypted_result: return f"**[Possible Trial Link]({EXAMPLE_CLINICAL_TRIAL_LINK})**" else: return "There was an error processing the data." # Define the function to handle image upload def handle_image_upload(image): url = 'http://images.cocodataset.org/val2017/000000039769.jpg' image = Image.open(requests.get(url, stream=True).raw) processor = ViTImageProcessor.from_pretrained('google/vit-base-patch16-224-in21k') model = ViTModel.from_pretrained('google/vit-base-patch16-224-in21k') inputs = processor(images=image, return_tensors="pt") outputs = model(**inputs) pooler_output = outputs.pooler_output[0] sclaed_output = 127 + 127 * pooler_output / pooler_output.abs().max() sclaed_output = sclaed_output.to(int) url = "/inference/clear-diagnosis" return ["Melanoma", "Vascular lesion"] # Create the Gradio interface with gr.Blocks() as demo: gr.Markdown("# Patient Data Criteria Form\nPlease fill in the criteria for the type of patients you are looking for.") with gr.Column(): with gr.Group(): age_input.render() gender_input.render() ethnicity_input.render() geographic_location_input.render() medications_input.render() allergies_input.render() previous_treatments_input.render() blood_glucose_level_input.render() blood_pressure_systolic_input.render() blood_pressure_diastolic_input.render() bmi_input.render() smoking_status_input.render() alcohol_consumption_input.render() exercise_habits_input.render() diet_input.render() condition_severity_input.render() functional_status_input.render() previous_trial_participation_input.render() with gr.Group(): diagnoses_icd10_input = gr.CheckboxGroup(choices=additional_categories["Diagnoses_ICD10"], label="Skin Diagnosis", interactive=False) image_input = gr.Image(label="Upload an Image") gr.Button("Upload").click(handle_image_upload, inputs=image_input, outputs=diagnoses_icd10_input) with gr.Group(): output = gr.Markdown("**Server response**") gr.Button("Submit").click(process_patient_data, inputs=[ age_input, gender_input, ethnicity_input, geographic_location_input, diagnoses_icd10_input, medications_input, allergies_input, previous_treatments_input, blood_glucose_level_input, blood_pressure_systolic_input, blood_pressure_diastolic_input, bmi_input, smoking_status_input, alcohol_consumption_input, exercise_habits_input, diet_input, condition_severity_input, functional_status_input, previous_trial_participation_input ], outputs=output) # Launch the app demo.launch()