DWCLF-Net: A weighted contrastive learning feature fusion network for temporal scar image sequence classification
Bibliographic record
Abstract
Accurate staging of scars is crucial for the targeted treatment of patients. However, the task is complicated by the varied development of scars across different individuals and body parts. To address this issue, we have constructed a temporal scar sequence dataset that distinguishes both individuals and body parts, aiming to capture the complex dynamics of scars from a temporal perspective. In previous studies, deep learning sequence models have been crucial for analyzing medical temporal data, identifying disease progression patterns by processing temporal dependencies. Despite their success, previous methods showed limitations in handling data with high inter-class similarity. In response, a novel strategy, which we named DWCLF-Net, a dual-path time series classification network based on supervised contrastive learning, has been proposed to address this issue. This network focuses on the spatial and temporal features of sequences through two distinct encoders. Additionally, it incorporates a weighted supervised contrastive learning strategy designed to tackle inter-class similarity and data imbalance with dynamic weight adjustment. Moreover, an interactive feature fusion module is introduced to integrate these features effectively for robust classification. We demonstrate the performance of DWCLF-Net in scar staging using our constructed scar sequence dataset. Compared to state-of-the-art temporal models, DWCLF-Net has proven its superiority by achieving 82.21% Accuracy, 83.16% Precision, 82.21% Recall, and 82.45% F1-score, significantly outperforming existing baseline models. These results validate the effectiveness of our model. Finally, we analyzed the correlation between clinical scar indicators and the Vancouver Scar Scale, finding a significant correlation between scar phase and prognosis.
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) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| 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".