PCOS-ML / app.py
Zane Vijay Falcao
Upload folder using huggingface_hub
d3915f5 verified
raw
history blame
18.3 kB
import gradio as gr
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import pickle
import joblib
import os
# Load the trained model (try both joblib and pickle in case one fails)
def load_model():
try:
model = joblib.load('pcos_model.joblib')
print("Model loaded using joblib")
return model
except:
try:
with open('random_forest_model', 'rb') as file:
model = pickle.load(file)
print("Model loaded using pickle from random_forest_model")
return model
except:
try:
with open('random_forest_model.pkl', 'rb') as file:
model = pickle.load(file)
print("Model loaded using pickle from pcos_model.pkl")
return model
except Exception as e:
print(f"Error loading model: {e}")
# Fallback to a simple model for demo purposes
from sklearn.ensemble import RandomForestClassifier
print("Creating a fallback model for demonstration")
fallback_model = RandomForestClassifier(n_estimators=100, random_state=42)
# Train with dummy data to initialize
X_dummy = np.random.rand(100, 43)
y_dummy = np.random.choice([0, 1], 100)
fallback_model.fit(X_dummy, y_dummy)
return fallback_model
# Load the model
model = load_model()
# Define the features required for prediction
features = [
"Age (yrs)", "Weight (Kg)", "Height(Cm)", "BMI", "Blood Group", "Pulse rate(bpm)",
"RR (breaths/min)", "Hb(g/dl)", "Cycle length(days)", "Cycle(R/I)", "Marraige Status (Yrs)",
"Pregnant(Y/N)", "No. of abortions", "Hip(inch)", "Waist(inch)", "Waist:Hip Ratio",
"Weight gain(Y/N)", "hair growth(Y/N)", "Skin darkening (Y/N)", "Hair loss(Y/N)",
"Pimples(Y/N)", "Fast food (Y/N)", "Reg.Exercise(Y/N)", "BP _Systolic (mmHg)",
"BP _Diastolic (mmHg)", "Follicle No. (L)", "Follicle No. (R)", "Avg. F size (L) (mm)",
"Avg. F size (R) (mm)", "Endometrium (mm)", "FSH(mIU/mL)", "LH(mIU/mL)", "FSH/LH",
"Hip:Waist Ratio", "TSH (mIU/L)", "AMH(ng/mL)", "PRL(ng/mL)", "Vit D3 (ng/mL)",
"PRG(ng/mL)", "RBS(mg/dl)", "Weight gain", "I beta-HCG(mIU/mL)", "II beta-HCG(mIU/mL)"
]
# Create visualizations for the dashboard
def create_visualizations():
# For demo purposes, we'll use sample data similar to what was in your notebook
# In a real application, you would load the actual dataset
# Sample data for visualization (small dataset for demo)
np.random.seed(42)
n_samples = 100
# Create sample data
sample_data = {
"Age (yrs)": np.random.normal(25, 5, n_samples),
"PCOS (Y/N)": np.random.choice([0, 1], n_samples, p=[0.6, 0.4]),
"BMI": np.random.normal(25, 5, n_samples),
"Cycle length(days)": np.random.normal(28, 5, n_samples),
"Follicle No. (L)": np.random.normal(12, 5, n_samples),
"Follicle No. (R)": np.random.normal(12, 5, n_samples),
"Endometrium (mm)": np.random.normal(8, 2, n_samples),
"Cycle(R/I)": np.random.choice([2, 4], n_samples),
"Weight (Kg)": np.random.normal(65, 10, n_samples),
"Hb(g/dl)": np.random.normal(12, 1.5, n_samples)
}
# Create a DataFrame
df = pd.DataFrame(sample_data)
# For PCOS cases, adjust the values to show differences
pcos_indices = df["PCOS (Y/N)"] == 1
df.loc[pcos_indices, "BMI"] += 2
df.loc[pcos_indices, "Cycle length(days)"] += 5
df.loc[pcos_indices, "Follicle No. (L)"] += 8
df.loc[pcos_indices, "Follicle No. (R)"] += 7
df.loc[pcos_indices, "Cycle(R/I)"] = 4
# Create visualizations
visualizations = []
# 1. BMI vs Age scatter plot
fig1, ax1 = plt.subplots(figsize=(8, 6))
sns.scatterplot(x="Age (yrs)", y="BMI", hue="PCOS (Y/N)",
data=df, palette=["teal", "plum"], ax=ax1)
ax1.set_title("BMI vs Age by PCOS Status")
visualizations.append(fig1)
# 2. Cycle length vs Age scatter plot
fig2, ax2 = plt.subplots(figsize=(8, 6))
sns.scatterplot(x="Age (yrs)", y="Cycle length(days)", hue="PCOS (Y/N)",
data=df, palette=["teal", "plum"], ax=ax2)
ax2.set_title("Menstrual Cycle Length vs Age by PCOS Status")
visualizations.append(fig2)
# 3. Follicle distribution scatter plot
fig3, ax3 = plt.subplots(figsize=(8, 6))
sns.scatterplot(x="Follicle No. (L)", y="Follicle No. (R)", hue="PCOS (Y/N)",
data=df, palette=["teal", "plum"], ax=ax3)
ax3.set_title("Follicle Distribution (Left vs Right Ovary)")
visualizations.append(fig3)
# 4. Boxplot for Follicle numbers
fig4, ax4 = plt.subplots(figsize=(10, 6))
sns.boxplot(x="PCOS (Y/N)", y="Follicle No. (L)", data=df, palette=["teal", "plum"], ax=ax4)
ax4.set_title("Follicle Count (Left Ovary) by PCOS Status")
visualizations.append(fig4)
# 5. Endometrium thickness boxplot
fig5, ax5 = plt.subplots(figsize=(10, 6))
sns.boxplot(x="PCOS (Y/N)", y="Endometrium (mm)", data=df, palette=["teal", "plum"], ax=ax5)
ax5.set_title("Endometrium Thickness by PCOS Status")
visualizations.append(fig5)
return visualizations
# Helper function to get numerical value for categorical inputs
def get_numerical_value(value, options):
try:
return options.index(value)
except:
return 0
# Helper function to preprocess inputs
def preprocess_inputs(input_dict):
# Convert checkbox values to 0/1
for key in input_dict:
if isinstance(input_dict[key], bool):
input_dict[key] = 1 if input_dict[key] else 0
# Convert blood group to numeric
blood_groups = ["A+", "A-", "B+", "B-", "AB+", "AB-", "O+", "O-"]
if "Blood Group" in input_dict and input_dict["Blood Group"] in blood_groups:
input_dict["Blood Group"] = blood_groups.index(input_dict["Blood Group"])
return input_dict
# Function to process input and make predictions
def predict_pcos(*args):
if model is None:
return "Model not loaded correctly. Please check if model files are available."
try:
# Convert inputs to a dictionary and then DataFrame
input_dict = {feature: value for feature, value in zip(features, args)}
# Preprocess inputs
input_dict = preprocess_inputs(input_dict)
# Convert to DataFrame
input_df = pd.DataFrame([input_dict])
# Print for debugging
print("Input shape:", input_df.shape)
print("Input data types:", input_df.dtypes)
# Make prediction
try:
prediction = model.predict(input_df)[0]
probability = model.predict_proba(input_df)[0]
result = "Positive for PCOS" if prediction == 1 else "Negative for PCOS"
conf = probability[1] if prediction == 1 else probability[0]
return f"{result} (Confidence: {conf:.2f})"
except AttributeError:
# If model is a numpy array, use a simple threshold-based prediction
# This is a fallback if the loaded model is just coefficients
print("Model is not a classifier object, using fallback prediction")
risk_score = np.mean([
input_df["BMI"].values[0] / 30,
input_df["Follicle No. (L)"].values[0] / 15,
input_df["Follicle No. (R)"].values[0] / 15,
(1 if input_df["Cycle(R/I)"].values[0] > 3 else 0)
])
prediction = 1 if risk_score > 0.6 else 0
result = "Positive for PCOS" if prediction == 1 else "Negative for PCOS"
return f"{result} (Risk Score: {risk_score:.2f})"
except Exception as e:
import traceback
traceback.print_exc()
return f"Error making prediction: {str(e)}"
# Function to display visualizations
def show_visualization(visualization_index):
visualizations = create_visualizations()
if 0 <= visualization_index < len(visualizations):
return visualizations[visualization_index]
return None
# Create the Gradio interface
with gr.Blocks(title="PCOS Detection Tool") as app:
gr.Markdown("# PCOS Detection and Analysis Tool")
gr.Markdown("This application uses machine learning to detect Polycystic Ovary Syndrome (PCOS) based on patient data.")
with gr.Tabs():
with gr.TabItem("Make Prediction"):
with gr.Row():
with gr.Column():
gr.Markdown("### Patient Demographics")
age = gr.Slider(18, 50, value=25, label="Age (yrs)")
weight = gr.Slider(40, 120, value=60, label="Weight (Kg)")
height = gr.Slider(140, 190, value=160, label="Height (cm)")
blood_group = gr.Dropdown(["A+", "A-", "B+", "B-", "AB+", "AB-", "O+", "O-"], value="A+", label="Blood Group")
bmi = gr.Slider(15, 40, value=22, label="BMI")
with gr.Column():
gr.Markdown("### Vital Signs")
pulse = gr.Slider(60, 120, value=80, label="Pulse rate (bpm)")
rr = gr.Slider(12, 25, value=16, label="Respiratory Rate (breaths/min)")
systolic = gr.Slider(90, 180, value=120, label="BP Systolic (mmHg)")
diastolic = gr.Slider(60, 120, value=80, label="BP Diastolic (mmHg)")
hb = gr.Slider(8, 18, value=12, label="Hemoglobin (g/dl)")
with gr.Row():
with gr.Column():
gr.Markdown("### Menstrual History")
cycle_length = gr.Slider(21, 45, value=28, label="Cycle length (days)")
cycle_regularity = gr.Radio([2, 4], value=2, label="Cycle Regularity (2=Regular, 4=Irregular)")
with gr.Column():
gr.Markdown("### Physical Measurements")
hip = gr.Slider(30, 60, value=40, label="Hip (inch)")
waist = gr.Slider(20, 50, value=30, label="Waist (inch)")
waist_hip_ratio = gr.Slider(0.6, 1.2, value=0.75, label="Waist:Hip Ratio")
hip_waist_ratio = gr.Slider(1.0, 2.0, value=1.33, label="Hip:Waist Ratio")
with gr.Row():
with gr.Column():
gr.Markdown("### Symptoms")
weight_gain = gr.Checkbox(label="Weight gain", value=False)
hair_growth = gr.Checkbox(label="Excessive hair growth", value=False)
skin_darkening = gr.Checkbox(label="Skin darkening", value=False)
hair_loss = gr.Checkbox(label="Hair loss", value=False)
pimples = gr.Checkbox(label="Pimples", value=False)
with gr.Column():
gr.Markdown("### Lifestyle")
fast_food = gr.Checkbox(label="Fast food consumption", value=False)
regular_exercise = gr.Checkbox(label="Regular exercise", value=False)
with gr.Row():
with gr.Column():
gr.Markdown("### Ultrasound Findings")
follicle_l = gr.Slider(0, 30, value=10, label="Follicle No. (Left)")
follicle_r = gr.Slider(0, 30, value=10, label="Follicle No. (Right)")
avg_fsize_l = gr.Slider(0, 25, value=5, label="Avg. Follicle size (Left) (mm)")
avg_fsize_r = gr.Slider(0, 25, value=5, label="Avg. Follicle size (Right) (mm)")
endometrium = gr.Slider(1, 20, value=8, label="Endometrium (mm)")
with gr.Column():
gr.Markdown("### Hormone Levels")
fsh = gr.Slider(0, 20, value=6, label="FSH (mIU/mL)")
lh = gr.Slider(0, 20, value=7, label="LH (mIU/mL)")
fsh_lh_ratio = gr.Slider(0, 3, value=0.85, label="FSH/LH Ratio")
tsh = gr.Slider(0, 10, value=2.5, label="TSH (mIU/L)")
amh = gr.Slider(0, 10, value=3, label="AMH (ng/mL)")
prl = gr.Slider(0, 30, value=15, label="Prolactin (ng/mL)")
vit_d3 = gr.Slider(0, 100, value=30, label="Vitamin D3 (ng/mL)")
prg = gr.Slider(0, 20, value=5, label="Progesterone (ng/mL)")
with gr.Row():
with gr.Column():
gr.Markdown("### Other Medical")
married_years = gr.Slider(0, 20, value=0, label="Marriage Status (Years)")
pregnant = gr.Checkbox(label="Currently Pregnant", value=False)
abortions = gr.Slider(0, 5, value=0, label="Number of abortions")
rbs = gr.Slider(70, 200, value=90, label="Random Blood Sugar (mg/dl)")
beta_hcg1 = gr.Slider(0, 100, value=5, label="I beta-HCG (mIU/mL)")
beta_hcg2 = gr.Slider(0, 100, value=5, label="II beta-HCG (mIU/mL)")
predict_btn = gr.Button("Predict PCOS Status")
prediction_output = gr.Textbox(label="Prediction Result")
# Connect inputs to prediction function
input_components = [
age, weight, height, bmi, blood_group, pulse, rr, hb, cycle_length,
cycle_regularity, married_years, pregnant, abortions, hip, waist,
waist_hip_ratio, weight_gain, hair_growth, skin_darkening, hair_loss,
pimples, fast_food, regular_exercise, systolic, diastolic, follicle_l,
follicle_r, avg_fsize_l, avg_fsize_r, endometrium, fsh, lh, fsh_lh_ratio,
hip_waist_ratio, tsh, amh, prl, vit_d3, prg, rbs, weight_gain, beta_hcg1, beta_hcg2
]
predict_btn.click(
predict_pcos,
inputs=input_components,
outputs=prediction_output
)
with gr.TabItem("Visualizations"):
gr.Markdown("### PCOS Data Analysis Visualizations")
visualization_choice = gr.Radio(
["BMI vs Age", "Menstrual Cycle Length vs Age", "Follicle Distribution",
"Follicle Count Boxplot", "Endometrium Thickness"],
value="BMI vs Age",
label="Select Visualization"
)
visualization_output = gr.Plot()
visualization_choice.change(
lambda choice: show_visualization(["BMI vs Age", "Menstrual Cycle Length vs Age",
"Follicle Distribution", "Follicle Count Boxplot",
"Endometrium Thickness"].index(choice)),
inputs=visualization_choice,
outputs=visualization_output
)
with gr.TabItem("About PCOS"):
gr.Markdown("""
# Polycystic Ovary Syndrome (PCOS)
Polycystic ovary syndrome (PCOS) is a hormonal disorder common among women of reproductive age.
Women with PCOS may have infrequent or prolonged menstrual periods or excess male hormone (androgen) levels.
## Common Symptoms
- Irregular periods
- Excess androgen (elevated levels of male hormones)
- Polycystic ovaries
- Weight gain
- Acne
- Excessive hair growth (hirsutism)
- Thinning hair or hair loss
- Infertility
## Risk Factors
- Having a mother or sister with PCOS
- Insulin resistance
- Obesity
## Complications
- Infertility
- Gestational diabetes or pregnancy-induced high blood pressure
- Miscarriage or premature birth
- Type 2 diabetes or prediabetes
- Depression, anxiety, and eating disorders
- Sleep apnea
- Endometrial cancer
- Cardiovascular disease
## Treatment
Treatment focuses on managing your individual concerns, such as infertility, hirsutism, acne or obesity.
Specific treatment might involve lifestyle changes or medication.
""")
with gr.TabItem("Debug Info"):
gr.Markdown("### Model and System Information")
debug_output = gr.Textbox(label="Debug Information", value=f"Model type: {type(model).__name__}")
debug_btn = gr.Button("Check Model Status")
def check_model():
try:
if model is None:
return "Model not loaded"
model_info = f"Model type: {type(model).__name__}\n"
# Try to get additional info based on model type
if hasattr(model, 'n_estimators'):
model_info += f"Number of estimators: {model.n_estimators}\n"
if hasattr(model, 'feature_importances_'):
top_features = np.argsort(model.feature_importances_)[-5:]
model_info += "Top 5 important features (indices): " + str(top_features.tolist()) + "\n"
# Check if the model has predict and predict_proba methods
has_predict = hasattr(model, 'predict') and callable(getattr(model, 'predict'))
has_proba = hasattr(model, 'predict_proba') and callable(getattr(model, 'predict_proba'))
model_info += f"Has predict method: {has_predict}\n"
model_info += f"Has predict_proba method: {has_proba}\n"
return model_info
except Exception as e:
return f"Error checking model: {str(e)}"
debug_btn.click(check_model, outputs=debug_output)
# Launch the app
if __name__ == "__main__":
app.launch(share=True, debug=True)