Add image overrides and local image support to projects list

tkapur · claude · tkapur · commit 89d807055ad9 · 2026-06-25T15:47:06.000-04:00
- Add IMAGE_OVERRIDES dict for manual image selection (projects 5, 9, 17, 32)
- Support local repo images via raw GitHub URLs
- Regenerate with latest project page updates

Co-Authored-By: Claude Sonnet 4.6 &lt;noreply@anthropic.com&gt;
diff --git a/PW45_2026_Boston/gen_projects_list.py b/PW45_2026_Boston/gen_projects_list.py
@@ -4,6 +4,19 @@
 PROJECTS_DIR = "/Users/tkapur/repos/ProjectWeek/PW45_2026_Boston/Projects"
 OUT_PATH = "/Users/tkapur/repos/ProjectWeek/PW45_2026_Boston/projects_list.md"
 BASE_URL = "https://projectweek.na-mic.org/PW45_2026_Boston/Projects"
+RAW_BASE = "https://raw.githubusercontent.com/NA-MIC/ProjectWeek/master/PW45_2026_Boston/Projects"
+
+# Manual image overrides: folder name → image URL to use instead of auto-selected
+IMAGE_OVERRIDES = {
+    "CastInterfaceModuleFor3DSlicerServiceProvidersImageDisplayClientAndHub":
+        "https://github.com/user-attachments/assets/7c8aefc0-d697-4813-9193-53fe08944c99",
+    "FanMaskAndOcrBoundBoxEditingForOhifBasedDicomDeIdentificationVerificationMode":
+        "https://github.com/user-attachments/assets/f4a099f4-5f71-40a3-a22f-f84ace253d0e",
+    "Mr2CbctRestoringAndExtendingAutomatedCbctMriRegistrationForTmjAnalysis":
+        "https://github.com/user-attachments/assets/f8aa1b2d-785f-422d-9fc3-d4350c1f5b37",
+    "Vox2SeglamProtocolGuidedSubcorticalSegmentationIn3DSlicer":
+        "https://github.com/user-attachments/assets/51fcdf47-e1a7-4318-ac84-38637db7b6bc",
+}
 
 
 def extract_section(text, start_pattern, stop_patterns):
@@ -50,28 +63,33 @@ def get_images_from_section(text, start_pattern, stop_patterns):
     return results
 
 
-def pick_best_image(text):
+def resolve_img_url(url, folder):
+    """Convert a relative image path to a raw GitHub URL."""
+    if url.startswith('http'):
+        return url
+    return f"{RAW_BASE}/{folder}/{url}"
+
+
+def pick_best_image(text, folder):
     """Pick the best image: prefer Illustrations section, then largest width."""
     stop = [r'^# ', r'^## ']
 
     # Try Illustrations section first
     illus = get_images_from_section(text, r'^# Illustrations', stop)
     if illus:
-        # Pick the one with the largest declared width; fall back to first
         best = max(illus, key=lambda x: x[0])
-        return best[1]
+        return resolve_img_url(best[1], folder)
 
     # Fall back to any image in the file
     all_imgs = get_images_from_section(text, r'^---\s*$', [r'^\Z'])
     if not all_imgs:
-        # Try whole file
         results = []
         for tag in re.finditer(r'<img\b[^>]+>', text):
             url_m = re.search(r'src="(https://[^"]+)"', tag.group())
             if url_m and not re.search(r'youtube|youtu\.be', url_m.group(1), re.I):
                 w_m = re.search(r'width="([0-9]+)"', tag.group())
                 results.append((int(w_m.group(1)) if w_m else 0, url_m.group(1)))
-        for m2 in re.finditer(r'!\[[^\]]*\]\((https://[^)]+)\)', text):
+        for m2 in re.finditer(r'!\[[^\]]*\]\(([^)]+)\)', text):
             url = m2.group(1)
             if not re.search(r'youtube|youtu\.be', url, re.I):
                 results.append((0, url))
@@ -80,7 +98,7 @@ def pick_best_image(text):
     if not all_imgs:
         return None
     best = max(all_imgs, key=lambda x: x[0])
-    return best[1]
+    return resolve_img_url(best[1], folder)
 
 
 projects = []
@@ -126,7 +144,7 @@ def pick_best_image(text):
             break
 
     proj_url = f"{BASE_URL}/{name}/"
-    img_url = pick_best_image(text)
+    img_url = IMAGE_OVERRIDES.get(name) or pick_best_image(text, name)
 
     desc = extract_section(text, r'^# Project Description', [r'^## ', r'^# '])
     # Convert any numbered list items in description to bullets so they
diff --git a/PW45_2026_Boston/projects_list.md b/PW45_2026_Boston/projects_list.md
@@ -36,7 +36,7 @@ The goal of this project is to further increase availability of anatomic segment
 
 5. **[Cast interface extension for  Slim and 3D Slicer: Hub,  Resource Servers and Image Display client.](https://projectweek.na-mic.org/PW45_2026_Boston/Projects/CastInterfaceModuleFor3DSlicerServiceProvidersImageDisplayClientAndHub/)** (Martin Bellehumeur @ Radical Imaging, +) ([mbellehumeur/SlicerCastInterface](https://github.com/mbellehumeur/SlicerCastInterface))
 
-<a href="https://projectweek.na-mic.org/PW45_2026_Boston/Projects/CastInterfaceModuleFor3DSlicerServiceProvidersImageDisplayClientAndHub/"><img src="https://github.com/user-attachments/assets/76dbe891-9ef5-49c9-839c-827d4a13cc88" style="max-width:480px;width:100%"></a>
+<a href="https://projectweek.na-mic.org/PW45_2026_Boston/Projects/CastInterfaceModuleFor3DSlicerServiceProvidersImageDisplayClientAndHub/"><img src="https://github.com/user-attachments/assets/7c8aefc0-d697-4813-9193-53fe08944c99" style="max-width:480px;width:100%"></a>
 
 Try the new “Cast Interface” 3D slicer extension to integrate different AI engines with VolView and OHIF without a DICOM server.  
 
@@ -74,7 +74,7 @@ This project focuses on using existing foundational deep learning models in 3D S
 
 9. **[Fan mask and OCR bound box editing for OHIF-based DICOM De-Identification verification](https://projectweek.na-mic.org/PW45_2026_Boston/Projects/FanMaskAndOcrBoundBoxEditingForOhifBasedDicomDeIdentificationVerificationMode/)** (Dave Dinh @ BWH, +) (no github repo yet)
 
-<a href="https://projectweek.na-mic.org/PW45_2026_Boston/Projects/FanMaskAndOcrBoundBoxEditingForOhifBasedDicomDeIdentificationVerificationMode/"><img src="https://github.com/user-attachments/assets/b95e956f-f422-494c-8c52-91b3d1fbd174" style="max-width:480px;width:100%"></a>
+<a href="https://projectweek.na-mic.org/PW45_2026_Boston/Projects/FanMaskAndOcrBoundBoxEditingForOhifBasedDicomDeIdentificationVerificationMode/"><img src="https://github.com/user-attachments/assets/f4a099f4-5f71-40a3-a22f-f84ace253d0e" style="max-width:480px;width:100%"></a>
 
 Optimal DICOM de-identification for downstream research requires balancing removal and retention. We must scrub embedded PHI in both tags and pixels while preserving valuable non-PHI metadata (e.g., scan settings like depth). Even with both addressed, human review is still needed to catch misses.
 
@@ -144,7 +144,7 @@ During Project Week, we aim to develop and evaluate a more reliable and integrat
 
 17. **[MR2CBCT: Restoring and Extending Automated CBCT-MRI Registration for TMJ Analysis](https://projectweek.na-mic.org/PW45_2026_Boston/Projects/Mr2CbctRestoringAndExtendingAutomatedCbctMriRegistrationForTmjAnalysis/)** (Eduardo Duarte Caleme @ University of North Carolina, +) (no github repo yet)
 
-<a href="https://projectweek.na-mic.org/PW45_2026_Boston/Projects/Mr2CbctRestoringAndExtendingAutomatedCbctMriRegistrationForTmjAnalysis/"><img src="https://github.com/user-attachments/assets/6ff6b1e8-a796-4fd8-8bec-5be64b6bd389" style="max-width:480px;width:100%"></a>
+<a href="https://projectweek.na-mic.org/PW45_2026_Boston/Projects/Mr2CbctRestoringAndExtendingAutomatedCbctMriRegistrationForTmjAnalysis/"><img src="https://github.com/user-attachments/assets/f8aa1b2d-785f-422d-9fc3-d4350c1f5b37" style="max-width:480px;width:100%"></a>
 
 MRI-to-CBCT registration remains a pressing challenge in medical imaging. Simultaneous visualization of hard and soft tissue structures benefits both clinicians and patients in the diagnosis of temporomandibular degenerative joint disease. Challenges remain in accurately registering these two modalities due to differences in intensity distributions that complicate mutual information optimization, as well as the necessity for initial manual alignment, which can prove unintuitive and challenging for clinicians using current 3D Slicer tools.
 
@@ -214,6 +214,8 @@ IDC already offers a customized agent skill in [https://github.com/ImagingDataCo
 
 25. **[Robust Segmentation Experience in OHIF Viewer](https://projectweek.na-mic.org/PW45_2026_Boston/Projects/RobustSegmentationExperienceInOHIFViewer/)** (Jaeyoung Cho @ University Hospital Bonn, +) ([CCI-Bonn/OHIF-AI](https://github.com/CCI-Bonn/OHIF-AI))
 
+<a href="https://projectweek.na-mic.org/PW45_2026_Boston/Projects/RobustSegmentationExperienceInOHIFViewer/"><img src="https://raw.githubusercontent.com/NA-MIC/ProjectWeek/master/PW45_2026_Boston/Projects/RobustSegmentationExperienceInOHIFViewer/ohif-segmentation-stats-mpr.png" style="max-width:480px;width:100%"></a>
+
 OHIF Viewer is a widely used web-based medical image viewer built on Cornerstone3D. Segmentation is central to many clinical and research workflows, yet reliability and robustness in OHIF/Cornerstone still lag behind user expectations — with recurring errors and edge cases in everyday use ([OHIF/Viewers#5453](https://github.com/OHIF/Viewers/issues/5453)).
 
 During Project Week 45, I would like to focus on improving the stability and user experience of segmentation in OHIF/Cornerstone. This builds on ongoing work in [OHIF-AI](https://github.com/CCI-Bonn/OHIF-AI) — an OHIF-based viewer with server-side AI integration (SAM2, nnInteractive, MedSAM2, and related tools) at University Hospital Bonn — where segmentation workflows are central to day-to-day use. I am looking forward to collaborating with others interested in segmentation, digital pathology, and web-based imaging viewers.
@@ -307,7 +309,7 @@ The work in progress in [SlicerWGPU](https://github.com/pieper/SlicerWGPU), whic
 
 32. **[Vox2SegLAM: Protocol-Guided Subcortical Segmentation in 3D Slicer](https://projectweek.na-mic.org/PW45_2026_Boston/Projects/Vox2SeglamProtocolGuidedSubcorticalSegmentationIn3DSlicer/)** (Ahmed Rekik @ École de technologie supérieure, +) (no github repo yet)
 
-<a href="https://projectweek.na-mic.org/PW45_2026_Boston/Projects/Vox2SeglamProtocolGuidedSubcorticalSegmentationIn3DSlicer/"><img src="https://github.com/user-attachments/assets/d4ad19a7-6048-4d48-aa48-f893331e8d80" style="max-width:480px;width:100%"></a>
+<a href="https://projectweek.na-mic.org/PW45_2026_Boston/Projects/Vox2SeglamProtocolGuidedSubcorticalSegmentationIn3DSlicer/"><img src="https://github.com/user-attachments/assets/51fcdf47-e1a7-4318-ac84-38637db7b6bc" style="max-width:480px;width:100%"></a>
 
 3D SlicerVox2SegLAM is an extension that brings an AI-assisted, protocol-guided workflow to subcortical brain segmentation and landmarking. A joint CNN+GNN model predicts 26 subcortical structures and 20 anatomical landmarks directly from a T1 MRI, following the Harvard-Oxford Atlas (HOA 2.0) neuroanatomical protocol.
 
