Out-of-distribution generalization for segmentation of lymph node metastasis in breast cancer
Bibliographic record
Abstract
Pathology provides the definitive diagnosis, and Artificial Intelligence (AI) tools are poised to improve accuracy, inter-rater agreement, and turn-around time (TAT) of pathologists, leading to improved quality of care. A high value clinical application is the grading of Lymph Node Metastasis (LNM) which is used for breast cancer staging and guides treatment decisions. A challenge of implementing AI tools widely for LNM classification is domain shift, where Out-of-Distribution (OOD) data has a different distribution than the In-Distribution (ID) data used to train the model, resulting in a drop in performance in OOD data. This work proposes a novel clustering and sampling method to automatically curate training datasets in an unsupervised manner with the aim of improving model generalization abilities. To evaluate the generalization performance of the proposed models, we applied a novel use of the Two One-sided Tests (TOST) method. This method examines whether the performance on ID and OOD data is equivalent, serving as a proxy for generalization. We provide the first evidence for computing equivalence margins that are data-dependent, which reduces subjectivity. The proposed framework shows the ensembled models constructed from models that generalized across both tumor and normal patches enhanced performance, achieving an F1 score of 0.81 for LNM classification on unseen ID and OOD samples. Interactive viewing of slide-level segmentations can be accessed on PathcoreFlow™ through https://web.pathcore.com/folder/18555?s=QTJVHJuhrfe5 . Segmentation models are available at https://github.com/IAMLAB-Ryerson/OOD-Generalization-LNM .
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 | 0.006 | 0.010 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.001 | 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".