graido config

app.py

@@ -0,0 +1,81 @@
+import gradio as gr
+import numpy as np
+
+from tensorflow.keras.models import load_model
+
+model_p01 = load_model("notebook/models_artifacts/lstm_model_patient01.keras")
+model_p02 = load_model("notebook/models_artifacts/lstm_model_patient02.keras")
+model_p03 = load_model("notebook/models_artifacts/lstm_model_patient03.keras")
+model_p04 = load_model("notebook/models_artifacts/lstm_model_patient04.keras")
+
+
+
+def predict_glucose(p_num,time_of_day,BG2h45, BG2h30, BG2h15, BG2h, BG1h45, BG1h30, BG1h15, BG1h, BG45m, BG30m, BG15m,BG0m):
+    
+    day = {"Morning":1, "Afternoon":2, "Evening":3, "Night":0}
+    
+    predicted = 0.0
+    alert = ""
+    inputs = np.expand_dims([day[time_of_day],
+                             BG2h45, BG2h30, BG2h15, 
+                             BG2h, BG1h45, BG1h30, BG1h15, BG1h,
+                             BG45m, BG30m, BG15m, BG0m
+                             ,],axis=0)
+    if p_num == "Patient 1":
+        predicted = float(model_p01.predict(inputs)[0][0])
+    elif p_num == "Patient 2":
+        predicted = float(model_p02.predict(inputs)[0][0])
+    elif p_num == "Patient 3":
+        predicted = float(model_p03.predict(inputs)[0][0])
+    elif p_num == "Patient 4":
+        predicted = float(model_p04.predict(inputs)[0][0])
+
+    # Determine alert level
+    if predicted <= 3.0:
+        alert = "Severe hypoglycemia: urgent alert!"
+    elif predicted <= 4.0:
+        alert = "Mild hypoglycemia: consume 15g fast-acting carbs"
+    elif predicted >= 13.9:
+        alert = "Severe hyperglycemia: urgent alert!"
+    elif predicted >= 10.0:
+        alert = "Mild hyperglycemia: hydrate and adjust treatment"
+    else:
+        alert = "Normal range"
+
+    # Trend-based alert
+    last_bg = BG0m  # most recent measurement (15m ago)
+    slope = (predicted - last_bg) / 60.0  # mmol/L per minute
+    if slope > 0.1:
+        alert += " | Rapid rise"
+    elif slope < -0.1:
+        alert += " | Rapid drop"
+
+    return float(np.round(predicted, 2)), alert
+
+with gr.Blocks() as demo:
+    gr.Markdown("# Blood Glucose Prediction Interface")
+    with gr.Row():
+        with gr.Column():
+            BG2h45 = gr.Number(label="BG 2h45m ago (mmol/L)",value=8.0)
+            BG2h30 = gr.Number(label="BG 2h30m ago (mmol/L)",value=11.0)
+            BG2h15 = gr.Number(label="BG 2h15m ago (mmol/L)",value=13.0)
+            BG2h   = gr.Number(label="BG 2h ago (mmol/L)",value=10.0)
+            BG1h45 = gr.Number(label="BG 1h45m ago (mmol/L)",value=9.0)
+            BG1h30 = gr.Number(label="BG 1h30m ago (mmol/L)",value=8.7)
+        with gr.Column():
+            BG1h15 = gr.Number(label="BG 1h15m ago (mmol/L)",value=7.9)
+            BG1h   = gr.Number(label="BG 1h ago (mmol/L)",value=11.8)
+            BG45m  = gr.Number(label="BG 45m ago (mmol/L)",value=7.4)
+            BG30m  = gr.Number(label="BG 30m ago (mmol/L)",value=18.7)
+            BG15m  = gr.Number(label="BG 15m ago (mmol/L)",value=13.8)
+            BG0m  = gr.Number(label="BG now (mmol/L) (mmol/L)",value=15.0)
+    time_of_day = gr.Radio(["Morning", "Afternoon", "Evening", "Night"], label="Time of Day")
+    p_num = gr.Radio(["Patient 1", "Patient 2", "Patient 3", "Patient 4"], label="Patient")
+    predict_btn = gr.Button("Predict")
+    output_pred = gr.Number(label="Predicted BG in 1h (mmol/L)")
+    output_alert = gr.Textbox(label="Alerts")
+    predict_btn.click(predict_glucose, 
+                      inputs=[p_num,time_of_day,BG2h45, BG2h30, BG2h15, BG2h, BG1h45, BG1h30, BG1h15, BG1h, BG45m, BG30m, BG15m,BG0m],
+                      outputs=[output_pred, output_alert])
+
+demo.launch()