Detection of prostate cancer in 3D pathology datasets via generative immunolabeling
Bibliographic record
Abstract
Abstract Recent advancements in nondestructive 3D pathology offer a complement to standard histology by enabling comprehensive volumetric analyses of intact clinical specimens (e.g. biopsies). Prior studies have demonstrated the added prognostic value of 3D pathology for prostate cancer risk stratification by correlating 3D microarchitectural features with long-term patient outcomes. However, these analyses relied on coarse manual annotations of cancer-enriched regions for downstream analysis without fine-grained delineation between often-intermixed cancerous and benign glands. To address these limitations, we have developed a 3D computational pipeline: Synthetic Immunolabeling for Generative Heatmaps of Tumor (SIGHT). SIGHT relies on deep learning-based 3D image translation models, trained in a fully supervised fashion, to convert H&E-analog 3D pathology datasets into multiplexed 3D immunofluorescence datasets that facilitate tumor detection. Our implementation of SIGHT synthetically labels two cytokeratin markers that are differentially expressed in cancerous and benign prostate glands, which are used to generate explainable 3D heatmaps of cancer-enriched regions in prostate tissues. Validation of SIGHT against ground-truth annotations from a panel of genitourinary pathologists yields an average F1 score of 0.88 which is comparable to the average inter-pathologist agreement F1 score of 0.90. To demonstrate the value of SIGHT, we developed machine classifiers of recurrence risk based on 3D glandular histomorphometric features from 75 patients. Volumetric glandular analysis in SIGHT-identified cancer-enriched regions vs. all tissue regions yields an average Kaplan-Meier hazard ratio of 3.57 (1.6 – 7.9 CI) vs. 0.92 (0.45 – 1.89 CI).
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.002 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.002 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".