Pipeline Progress & Blockers
Status as of 2026-03-23
§1 Route A Pipeline Status
Table 1. Route A pipeline steps
| # | Step | Status | Output |
|---|---|---|---|
| 1 | DICOM → NIfTI conversion | Done | 114 patient directories in ReMIND_nifti/ |
| 2 | First-surgery patient filtering | Done | first_surgery_patients.json (62 patients) |
| 3a | Registration V1 (baseline) | Done | registration_results_v1.csv (204 rows, 6.9% conv.) |
| 3b | Registration V2 (+mask, 3k iter) | Done | registration_results_v2.csv (114 rows, 56.1% conv.) |
| 3c | A/B test: optimizer configs | Done | ab_test_20260317_075447.csv (70 rows, 7 configs) |
| 3d | Registration V3 (2 mm dil.) | Done | registration_results_v3.csv (114 rows, 73.7% conv.) |
| 3e | Rescue passes | Done | 84/114 valid after QC exclusion (73.7%) |
| 4 | Pseudo-label generation | Done | 84 labeled volumes → 44,502 2D slices (43 patients) |
| 5 | nnU-Net data formatting | Done | Dataset001_iUS (43 patients) |
| 6 | nnU-Net training | Blocked | — |
| 7 | Evaluation on RESECT-SEG | Pending | — |
§2 Route B Pipeline Status
Table 2. Route B pipeline steps
| # | Step | Status | Output |
|---|---|---|---|
| 1 | Virtual sweep generation (v3) | Done | 1,090 sweeps across 110 cases |
| 2 | Best-effort deep tumor fix | Done | 4 cases verified (v3_deep_fix/) |
| 3 | 2D→3D NIfTI conversion | Done | 1,878 MRI + 1,090 seg volumes (33.7 GB) |
| 4 | MMHVAE data normalisation | Pending | — |
| 5 | MMHVAE inference | Blocked | — |
| 6 | Synthetic iUS → nnU-Net training | Pending | — |
| 7 | Evaluation on RESECT-SEG | Pending | — |
§3 Current Blockers
Critical GPU Resource Required — Blocks Route A Step 6
Route A nnU-Net 5-fold cross-validation training requires GPU compute node access — current personal machine lacks sufficient GPU resources. Remote environment (x99-debian) is configured with CUDA 12.2 / Driver 535.247.01 / PyTorch 2.5.1+cu124, but lacks dedicated GPU allocation. Need to secure university compute node or equivalent GPU resource to proceed with training.
Blocked MMHVAE Pre-trained Weights Unavailable — Blocks Route B Step 5
The MMHVAE repository (github.com/ReubenDo/MMHVAE) documents pre-trained weights at pretrained/mmhvae_f0/models/CP_main_1000.pth but provides no download link and no GitHub Release. Email sent to first author R. Dorent, who replied that he is waiting for group lead approval to share weights. Author provided training code as fallback — if weights remain unavailable, can train from scratch once GPU resource is secured, but this also requires compute node access.