Redundant co-training: Semi-supervised segmentation of medical images using informative redundancy
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Pseudo-labeling, consistency regularization, and co-training are common paradigms for semi-supervised learning. In this paper, we propose a novel method based on co-training and pseudo-labeling for the semi-supervised segmentation of the left ventricle. Our co-training strategy is novel and unlike most previous works does not rely on using multiple-view datasets, performing weak/strong augmentations on the input images or perturbations on the networks. We proposed creating redundant labels by utilizing the provided ground-truths and training networks segmenting different overlapping regions corresponding to the created labels. Although the new labels seem to be redundant, we demonstrated that they provide valuable information to the networks. The predictions of the redundant networks (which are trained on the redundant labels) can be used in the pixels where the primary network’s predictions are not reliable. This enables extracting a secondary source of information without requiring any additional ground-truths. The common practice in pseudo-labeling is using the reliable predictions of the unlabeled data and discarding the unreliable ones. However, we proposed utilizing predictions from the redundant networks to generate pseudo-labels for the unreliable pixels in the primary network’s predictions, rather than simply discarding them. We validated our method on two left ventricle segmentation datasets, and it surpassed the state-of-the-art semi-supervised learning approaches. Furthermore, we conducted extensive studies to analyze the proposed method from different aspects. Implementation of our work is available at https://github.com/behnam-rahmati/redundant-cotraining .
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.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
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 it