Add demos

.gitignore

@@ -63,6 +63,8 @@ venv.bak/
 /data
 /assets
 /test/data
+*.zip
+/raw
 
 #WandB
 /notebooks/wandb

data081/HCC-TACE-Seg/raw/1.2.276.0.7230010.3.1.3.8323329.41.1604860085.518229/LICENSE

@@ -0,0 +1,14 @@
+License Information. The HCC-TACE-Seg collection is distributed under the CC BY 4.0 at https://creativecommons.org/licenses/by/4.0/ By downloading the data, you agree to abide by terms of this license.
+Data Usage Policy
+
+Any user accessing TCIA data must agree to:
+- Not use the requested datasets, either alone or in concert with any other information, to identify or contact individual participants from whom data and/or samples were collected and follow all other conditions specified in the TCIA Site Disclaimer. Approved Users also agree not to generate and use information (e.g., facial images or comparable representations) in a manner that could allow the identities of research participants to be readily ascertained. These provisions do not apply to research investigators operating with specific IRB approval, pursuant to 45 CFR 46, to contact individuals within datasets or to obtain and use identifying information under an IRB-approved research protocol. All investigators including any Approved User conducting “human subjects research” within the scope of 45 CFR 46 must comply with the requirements contained therein.
+
+- Acknowledge in all oral or written presentations, disclosures, or publications the specific dataset(s) or applicable accession number(s) and the NIH-designated data repositories through which the investigator accessed any data. Citation guidelines for doing this are outlined below.
+
+- If you are considering mirroring a copy of our publicly available datasets or providing direct access to any of the TCIA data via another tool or website using the REST API (https://wiki.cancerimagingarchive.net/x/NIIiAQ) please review our Data Analysis Centers (DACs) page (https://wiki.cancerimagingarchive.net/x/x49XAQ) for more information. DACs must provide attribution and links back to this TCIA data use policy and must require downstream users to do the same.
+
+The summary page for every TCIA dataset includes a Citations & Data Usage Policy tab.  Please consult the Citation & Data Usage Policy for each Collection before using them.  
+- Most data are freely available to browse, download, and use for commercial, scientific and educational purposes as outlined in the Creative Commons Attribution 3.0 Unported License or the Creative Commons Attribution 4.0 International License.  In rare circumstances commercial use may be prohibited using Attribution-NonCommercial 3.0 Unported (CC BY-NC 3.0) or Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0). 
+
+- Most data are immediately accessible and do not require account registration.  A small subset of collections do require registration and special permission to gain access. Refer to the "Access" column on https://www.cancerimagingarchive.net/collections/ for more details.

src/app.py

@@ -6,59 +6,44 @@
 import streamlit as st
 import torch
 
-# Now you can import from the submodule
-from diffdrr.data import read
-from diffdrr.visualization import plot_drr
-from drr import create_drr
-from model import TACEnet
-from training import loadData, sampleVolume
+from demo import demonstration, get_demo_data, load_model
 
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 
 # Set the page configuration
-st.set_page_config(page_title="DRR Enhancement Model Demo", layout="wide")
+st.set_page_config(page_title="Tacetastic", layout="wide")
 
 
-# Caching the model loading to avoid reloading the model on each interaction
 @st.cache_resource(show_spinner="Loading the model...")
-def load_model():
-    model = TACEnet()
-    model.load_state_dict(
-        torch.load("../models/TACEnet_vessel_enhancement_deformations_30052024.pth")
-    )
-    model.eval()
-    return model
+def cached_load_model(deformation, device):
+    return load_model(deformation, device)
 
 
-# Caching the DRR generation to avoid redundant computations
-@st.cache_data(show_spinner="Generating DRR...")
-def cached_create_drr(_train_loader, contrast_value, height=256, width=256, rotation=0):
-    _volume, _target = sampleVolume(_train_loader, contrast_value=contrast_value)
-    subject = read(
-        tensor=_volume[0], label_tensor=_target[0], bone_attenuation_multiplier=5.0
-    )
-    return (
-        create_drr(
-            subject,
-            device="cpu",
-            height=height,
-            width=width,
-            mask_to_channels=False,
-            rotations=torch.tensor([[rotation, 0.0, 0.0]]),
-        ),
-        _volume,
-        _target,
+@st.cache_data(show_spinner="Collecting the data...")
+def cached_get_demo_data():
+    return get_demo_data()
+
+
+@st.cache_data(show_spinner="Generating the results...")
+def cached_demonstration(
+    _volumes, _targets, _model, rotation, ef, deformation, initial_contrast, device
+):
+    # Your existing code to generate the figure
+    fig = demonstration(
+        _volumes, _targets, _model, rotation, ef, deformation, initial_contrast, device
     )
+    return fig
 
 
-@st.cache_data
-def cached_loadData():
-    return loadData()
+st.session_state.deformation_checkbox = False
 
+# Load the model
+model = cached_load_model(
+    deformation=st.session_state.deformation_checkbox, device=device
+)
 
-@st.cache_data
-def cached_sampleVolume(_train_loader, contrast_value):
-    return sampleVolume(_train_loader, contrast_value=contrast_value)
+# Load the demo data
+volumes, targets = cached_get_demo_data()
 
 
 # Show dialog
@@ -83,8 +68,6 @@ def terms_conditions():
 if "terms_conditions" not in st.session_state:
     terms_conditions()
 
-st.logo("https://avatars.githubusercontent.com/u/8323854?s=200&v=4")
-
 # Display the DRR and enhanced DRR in the main layout
 st.title("DRR Enhancement Model Demo")
 
@@ -108,63 +91,38 @@ def terms_conditions():
     )
 
     st.slider(
-        label="Select initial contrast",
-        min_value=0,
-        max_value=4000,
-        value=0,
-        step=500,
+        label="Select contrast agent reduction ratio",
+        min_value=0.0,
+        max_value=1.0,
+        value=1.0,
+        step=0.1,
         key="contrast_slider",
-        help="Select the intital contrast",
+        help="Select the contrast reduction ratio for the vessels",
         on_change=None,
         label_visibility="visible",
     )
 
-    "---"
-
-# Load your model
-model = load_model()
-model.to(device)
-
-# Load CT data
-train_loader, val_loader = cached_loadData()
-# volume, target = cached_sampleVolume(
-#     train_loader, contrast_value=st.session_state.contrast_slider
-# )
-
-# Initialize the DRR module for generating synthetic X-rays
-drr, volume, target = cached_create_drr(
-    train_loader,
-    contrast_value=st.session_state.contrast_slider,
-    rotation=st.session_state.rotation_slider,
-)
+    st.checkbox(
+        label="Enable deformation",
+        value=st.session_state.deformation_checkbox,
+        key="deformation_checkbox",
+        help="Enable deformation of the subject",
+    )
 
+    "---"
 
-col1, col2, col3 = st.columns(spec=[0.4, 0.2, 0.4])
+if st.sidebar.button("Generate"):
+    # Assuming demonstration now returns a figure directly
+    fig = cached_demonstration(
+        _volumes=volumes,
+        _targets=targets,
+        _model=model,
+        rotation=st.session_state.rotation_slider,
+        ef=st.session_state.contrast_slider,
+        deformation=st.session_state.deformation_checkbox,
+        initial_contrast=4000,
+        device=device,
+    )
 
-with col1:
-    st.header("Original DRR")
-    axs = plot_drr(drr, ticks=False)
-    fig = axs[0].figure
-    fig.set_size_inches(2, 2)  # Adjust the figure size
+    # Display the figure in Streamlit's main view
     st.pyplot(fig)
-
-with col3:
-    st.header("Enhanced DRR")
-    enhanced_placeholder = st.empty()  # Placeholder for enhanced image
-
-with col2:
-    # Add an "Enhance" button
-    if st.button("Enhance"):
-        with st.spinner("Enhancing DRR..."):
-            prediction, latent = model(
-                volume[0].unsqueeze(0).to(device), drr.to(device)
-            )
-            axs = plot_drr(prediction, ticks=False)
-            fig = axs[0].figure
-            fig.set_size_inches(2, 2)  # Adjust the figure size
-            enhanced_placeholder.pyplot(fig)
-
-col4, col5, col6 = st.columns(3)
-col4.metric("Temperature", "70 °F", "1.2 °F")
-col5.metric("Wind", "9 mph", "-8%")
-col6.metric("Humidity", "86%", "4%")

src/demo.py

@@ -24,6 +24,7 @@ def load_model(deformation, device):
     )
     model = TACEnet().to(device)
     model.load_state_dict(torch.load(model_path))
+    model.eval()
     return model
 
 
@@ -52,26 +53,37 @@ def generate_drr(subject, rotation, ef, device):
     return drr_combined, drr_vessels.to(device)
 
 
-def demonstration(rotation, ef, deformation=False, initial_contrast=4000, device="cpu"):
-    transform = get_transforms(
-        resize_shape=[512, 512, 96], contrast_value=initial_contrast
-    )
+def get_demo_data():
+    transform = get_transforms(resize_shape=[512, 512, 96])
     train_ds, _ = get_datasets(
         root_dir="../data081",
         collection="HCC-TACE-Seg",
         seg_type="SEG",
         transform=transform,
         download=False,
         download_len=1,
-        val_frac=0.2,
+        val_frac=0.0,
         seed=42,
     )
     train_loader, _ = get_dataloaders(train_ds, _, batch_size=1)
     batch = next(iter(train_loader))
     volumes, targets = batch["image"], batch["seg"]
+    return volumes, targets
+
+
+def demonstration(
+    volumes,
+    targets,
+    model,
+    rotation,
+    ef,
+    deformation=False,
+    initial_contrast=4000,
+    device="cpu",
+):
+
     volumes = add_vessel_contrast(volumes, targets, contrast_value=initial_contrast)
 
-    model = load_model(deformation, device)
     subject = read(
         tensor=volumes[0],
         label_tensor=targets[0],
@@ -83,5 +95,8 @@ def demonstration(rotation, ef, deformation=False, initial_contrast=4000, device
         subject = apply_deformation(subject)
     drr_combined, drr_target = generate_drr(subject, rotation, ef, device)
 
-    prediction, latent = model(targets.to(device), drr_combined)
-    plot_results(drr_combined, drr_target, prediction, latent, vmax=25)
+    with torch.no_grad():  # Disable gradient computation
+        prediction, latent = model(targets.to(device), drr_combined)
+    return plot_results(
+        drr_combined.cpu(), drr_target.cpu(), prediction.cpu(), latent.cpu(), vmax=25
+    )

src/pages/Surgeon Demo.py

@@ -0,0 +1,91 @@
+import random
+import time
+
+import streamlit as st
+
+st.set_page_config(page_title="Surgeon Demo", page_icon="👋")
+
+# Mock Patient Details
+patient_info = {
+    "Name": "John Doe",
+    "Age": 34,
+    "Gender": "Male",
+    "MRN": "123456789",
+    "Admission Date": "2023-04-01",
+    "Diagnosis": "Appendicitis",
+    "Surgeon Assigned": "Dr. Jane Smith",
+    "Procedure": "TACE",
+}
+
+# Vital Signs (Mock, could be dynamically updated)
+vital_signs = {
+    "SpO2": "98%",
+    "Blood Pressure": "120/80 mmHg",
+    "Respiratory Rate": "16 breaths/min",
+}
+
+# Display Patient Information in Sidebar
+st.sidebar.title("Patient Information")
+for key, value in patient_info.items():
+    st.sidebar.text(f"{key}: {value}")
+
+st.sidebar.title("Vital Signs")
+for key, value in vital_signs.items():
+    st.sidebar.text(f"{key}: {value}")
+
+# Initialize session state variables if they don't exist
+if "rotation" not in st.session_state:
+    st.session_state.rotation = -45  # Start at 0 degrees
+if "rotation_direction" not in st.session_state:
+    st.session_state.rotation_direction = 1  # Start moving towards -45
+if "confidence" not in st.session_state:
+    st.session_state.confidence = 98  # Start with high confidence
+if "spo2" not in st.session_state:
+    st.session_state.spo2 = random.randint(
+        80, 88
+    )  # Random SpO2 value between 94% and 98%
+
+st.image(r"D:\Programming\AI\5ARIP10-ITP-T3G3\src\pages\xray_demo.gif", width=400)
+
+
+# Placeholder for dynamic metrics
+col1, col2, col3 = st.columns(3)
+rotation_placeholder = col1.empty()
+confidence_placeholder = col2.empty()
+spo2_placeholder = col3.empty()
+
+while True:
+    # Update rotation
+    if st.session_state.rotation_direction == 1:
+        if st.session_state.rotation + 10 <= 45:
+            st.session_state.rotation += 10
+        else:
+            st.session_state.rotation = 45  # Correct to max 45 if overshooting
+            st.session_state.rotation_direction = -1
+    else:
+        if st.session_state.rotation - 10 >= -45:
+            st.session_state.rotation -= 10
+        else:
+            st.session_state.rotation = -45  # Correct to min -45 if overshooting
+            st.session_state.rotation_direction = 1
+
+    # Update confidence based on rotation
+    angle_from_zero = abs(st.session_state.rotation)
+    st.session_state.confidence = round(
+        98 - (15 * (angle_from_zero / 45)), 2
+    )  # Decreases from 98% to 65%
+
+    # Randomly vary SpO2 value
+    st.session_state.spo2 = random.randint(80, 88)
+
+    # Display updated metrics
+    rotation_placeholder.metric(
+        "Rotation", f"{st.session_state.rotation} °C", delta=None
+    )
+    confidence_placeholder.metric(
+        "Confidence", f"{st.session_state.confidence}%", delta=None
+    )
+    spo2_placeholder.metric("Heart Rate", f"{st.session_state.spo2} BPM", delta=None)
+
+    # Sleep for the duration of the GIF's rotation cycle or any desired update interval
+    time.sleep(0.157)