add aeon/paladin

app.py

@@ -1,3 +1,4 @@
+import os
 import gradio as gr
 import numpy as np
 import pandas as pd
@@ -5,8 +6,12 @@ from PIL import Image
 
 from mussel.utils import get_features, segment_tissue
 from mussel.models import ModelType
+from mosaic_ui.inference import run_aeon, run_paladin
 
-def analyze_slide(slide_path, model_name="RESNET50"):
+NUM_WORKERS = int(os.getenv("NUM_WORKERS", 2))
+USE_GPU = torch.cuda.is_available()
+
+def analyze_slide(slide_path, site_type):
     _, _, coords, attrs = segment_tissue(
         slide_path=slide_path,
         patch_size=224,
@@ -18,25 +23,74 @@ def analyze_slide(slide_path, model_name="RESNET50"):
         max_num_holes=2
     )
 
-    df_coords = pd.DataFrame(coords, columns=['x', 'y'])
-    features, labels = get_features(coords, slide_path, attrs, model_type=ModelType.RESNET50, use_gpu=False)
+#    features, _ = get_features(coords, slide_path, attrs, model_type=ModelType.OPTIMUS, use_gpu=USE_GPU, batch_size=64, num_workers=NUM_WORKERS)
+#
+#    # Step 3: Run Aeon to predict histology
+#    aeon_results, _ = run_aeon(
+#    │   features=features,
+#    │   model_path="data/aeon_model.pkl",
+#    │   metastatic=(site_type == "Metastatic"),
+#    │   batch_size=8,
+#    │   num_workers=NM_WORKERS,
+#    │   use_cpu=not USE_GPU,
+#    )
+#
+#    # Step 4: Run Paladin to predict biomarkers
+#    paladin_results = run_paladin(
+#    │   features=features,
+#    │   model_map_path="data/paladin_model_map.csv",
+#    │   aeon_results=aeon_results,
+#    │   metastatic=(site_type == "Metastatic"),
+#    │   batch_size=8,
+#    │   num_workers=NUM_WORKERS,
+#    │   use_cpu=not USE_GPU,
+#    )
+    # Dummy results for demonstration purposes
+    # Replace these with actual model inference results
+    aeon_results = pd.DataFrame({
+        "Cancer Subtype": ["Subtype A", "Subtype B"],
+        "Confidence": [0.95, 0.85]
+    })
+    paladin_results = pd.DataFrame({
+        "Cancer Subtype": ["Subtype A", "Subtype A"],
+        "Biomarker": ["Biomarker 1", "Biomarker 2"],
+        "Score": [0.9, 0.8]
+    })
 
-    return df_coords
+    return aeon_results, paladin_results
 
 with gr.Blocks() as demo:
-    gr.Markdown("# Whole Slide Image Featurization & Inference")
+    gr.Markdown("# Mosaic: H&E Whole Slide Image Cancer Subtype and Biomarker Inference")
     with gr.Row():
         with gr.Column():
-            input_slide = gr.File(label="Upload Whole Slide Image (.svs, .tiff)")
-            model_name = gr.Dropdown(choices=["RESNET50", "CTRANSPATH", "VIRCHOW", "VIRCHOW2", "OPTIMUS", "CLIP", "CONCH1_5"], label="Select Model", value="RESNET50")
-            run_button = gr.Button("Analyze Slide")
+            input_image = gr.File(
+                label="Upload H&E Whole Slide Image",
+                file_types=[".svs", ".tiff", ".tif"],
+            )
+            input_dropdown = gr.Dropdown(
+                choices=["Primary", "Metastatic"],
+                label="Site Type",
+                value="Primary",
+            )
+            run_button = gr.Button("Run Inference")
         with gr.Column():
-            coords_output = gr.Dataframe(headers=["X", "Y"], label="Extracted Patch Coordinates")
+            aeon_output_table = gr.Dataframe(
+                headers=["Cancer Subtype", "Confidence"],
+                label="Cancer Subtype Inference",
+                datatype=["str", "number"],
+            )
+            paladin_output_table = gr.Dataframe(
+                headers=["Cancer Subtype", "Biomarker", "Score"],
+                label="Biomarker Inference",
+                datatype=["str", "str", "number"],
+            )
+
 
     run_button.click(
         analyze_slide,
-        inputs=[input_slide, model_name],
-        outputs=coords_output
+        inputs=[input_slide, input_dropdown],
+        outputs=[aeon_output_table, paladin_output_table],
+
     )
 
 if __name__ == "__main__":

requirements.txt

@@ -16,3 +16,4 @@ scikit-learn
 transpath@git+https://github.com/msk-mind/TransPath/
 timm_ctranspath@git+https://github.com/msk-mind/timm-ctranspath
 open-clip-torch
+git+https://{GH_TOKEN}@github.com/pathology-data-mining/paladin_webapp.git@dev