Create app.py

app.py

@@ -0,0 +1,150 @@
+# app.py
+import gradio as gr
+import os
+import tempfile
+from pathlib import Path
+import numpy as np
+import matplotlib.pyplot as plt
+from lcmv_class import LCMVSourceEstimator
+
+# Predefine global paths (fsaverage already in repo root)
+FS_SRC_FILE = "fsaverage-vol-5mm-src.fif"
+GPS_TEMPLATE = "ghw280_from_egig.gpsc"
+
+def run_lcmv_app(ica_fif_file, gpsc_file, roi_indices_str="305,437,223"):
+    try:
+        # Parse ROI indices
+        roi_indices = [int(x.strip()) for x in roi_indices_str.split(",")]
+
+        # Create temp project dir
+        with tempfile.TemporaryDirectory() as tmp_dir:
+            tmp_path = Path(tmp_dir)
+            
+            # Copy/move uploaded files
+            ica_path = tmp_path / "sub-01_task-ec_ica_cleaned.fif"
+            gps_path = tmp_path / "electrodes.gpsc"
+            
+            with open(ica_path, "wb") as f:
+                f.write(ica_fif_file.read())
+            with open(gps_path, "wb") as f:
+                f.write(gpsc_file.read())
+
+            # Copy global fsaverage source space into temp dir (required by lcmv_class)
+            fs_src_dest = tmp_path / "fsaverage-vol-5mm-src.fif"
+            os.system(f"cp {FS_SRC_FILE} {fs_src_dest}")
+
+            # Config
+            config = {
+                'project_base': str(tmp_path),
+                'subject_id': 'sub-01',
+                'task': 'ec',
+                'ica_file_path': 'sub-01_task-ec_ica_cleaned.fif',
+                'gpsc_file_path': 'electrodes.gpsc',
+                'reg': 0.01,
+                'n_jobs': 1,  # Reduce parallelism for stability
+            }
+
+            # Run LCMV
+            estimator = LCMVSourceEstimator(config)
+            estimator.run_enhanced_computation()
+
+            # Run DiFuMo
+            difumo_config = {'dimension': 512, 'resolution_mm': 2}
+            time_courses, _ = estimator.run_difumo_extraction(difumo_config=difumo_config)
+
+            # Save time courses
+            tc_path = tmp_path / "difumo_time_courses.npy"
+            np.save(tc_path, time_courses)
+
+            # Generate PSD plot
+            fig = plot_psd_rois(str(tc_path), roi_indices, sfreq=500.0, cmap_name='viridis_r')
+            plot_path = tmp_path / "psd_plot.png"
+            fig.savefig(plot_path, dpi=150, bbox_inches='tight')
+            plt.close(fig)
+
+            return str(tc_path), str(plot_path)
+
+    except Exception as e:
+        return None, f"❌ Error: {str(e)}"
+
+# Reuse your plot function (slightly modified)
+def plot_psd_rois(file_path, roi_indices, sfreq=500.0, figsize=(9, 4), cmap_name='viridis_r'):
+    import numpy as np
+    import matplotlib.pyplot as plt
+    from scipy import signal
+    from nilearn import datasets
+
+    data = np.load(file_path)
+    if data.shape[0] != 512:
+        data = data.T
+
+    labels = datasets.fetch_atlas_difumo(512, 2).labels['difumo_names'].tolist()
+    cmap = plt.cm.get_cmap(cmap_name)
+    colors = cmap(np.linspace(0, 1, len(roi_indices)))
+
+    fig, ax = plt.subplots(figsize=figsize)
+    freqs = None
+
+    for i, idx in enumerate(roi_indices):
+        ts = data[idx]
+        freqs, psd = signal.welch(ts, fs=sfreq, nperseg=int(1.5 * sfreq))
+        psd_db = 10 * np.log10(psd)
+        name = labels[idx].strip()
+        ax.plot(freqs, psd_db, label=name, color=colors[i], linewidth=2)
+
+    FREQ_BANDS = [
+        ('Delta', (0, 4)),
+        ('Theta', (4, 8)),
+        ('Alpha', (8, 12)),
+        ('Low_Beta', (12, 20)),
+        ('High_Beta', (20, 30)),
+        ('Low_Gamma', (30, 50)),
+        ('High_Gamma', (50, 120))
+    ]
+    blue_shades = plt.cm.Blues(np.linspace(0.9, 0.2, len(FREQ_BANDS)))
+    band_colors = {band: blue_shades[i] for i, (band, _) in enumerate(FREQ_BANDS)}
+
+    if freqs is not None:
+        for band, (fmin, fmax) in FREQ_BANDS:
+            if fmax < freqs[0] or fmin > freqs[-1]:
+                continue
+            ax.axvspan(fmin, fmax, color=band_colors[band], alpha=0.3, zorder=0)
+
+    ax.set_xlim(0, 120)
+    ax.set_ylim(-70, -10)
+    ax.set_xlabel("Frequency (Hz)")
+    ax.set_ylabel("Power Spectral Density (dB/Hz)")
+    ax.set_title("Source-Level PSD (LCMV + DiFuMo)", fontweight='bold')
+    ax.grid(True, alpha=0.1)
+    ax.legend(loc='upper right', fontsize=9, framealpha=0.8)
+    plt.tight_layout()
+    return fig
+
+# Gradio Interface
+with gr.Blocks(title="LCMV_Xtra: EEG Source Imaging") as demo:
+    gr.Markdown("# 🧠 LCMV_Xtra: EEG Source Imaging with DiFuMo")
+    gr.Markdown("Upload ICA-cleaned EEG (.fif) and electrode (.gpsc) files to compute source-level time courses and PSD.")
+    
+    with gr.Row():
+        with gr.Column():
+            fif_input = gr.File(label="ICA-cleaned EEG (.fif)", file_types=[".fif"])
+            gpsc_input = gr.File(label="Electrode positions (.gpsc)", file_types=[".gpsc"])
+            roi_input = gr.Textbox(label="DiFuMo ROI Indices (comma-separated)", value="305,437,223")
+            run_btn = gr.Button("Run LCMV + DiFuMo")
+        
+        with gr.Column():
+            time_course_out = gr.File(label="Download DiFuMo Time Courses (.npy)")
+            plot_out = gr.Image(label="Source-Level PSD")
+
+    run_btn.click(
+        fn=run_lcmv_app,
+        inputs=[fif_input, gpsc_input, roi_input],
+        outputs=[time_course_out, plot_out]
+    )
+
+    gr.Markdown("""
+    > ⚠️ **Note**: Processing may take 1–3 minutes. Uses pre-downloaded `fsaverage` and 5mm source space.
+    """)
+
+demo.launch()
+