Automatic characterization of breast lesions using multi-scale attention-guided deep learning of digital histology images
Bibliographic record
Abstract
A multi-scale attention-guided deep learning model is proposed to characterise breast tissue in digital histology images (H&E stained) according to four different histological types including normal, benign, in situ carcinoma and invasive carcinoma. The framework includes two parallel convolutional neural networks with modified VGG16 architecture. The first network analyzes the whole-sample images at low magnification. The second network focuses on the patches extracted from the whole-sample images at high magnification. In the low-magnification network, a global average pooling layer was added at the end of the network to extract the class activation maps for the attention model. A long short-term memory network was adapted as a recurrent attention mechanism to increase the contribution of the relevant parts of each image for classification. In the high-magnification network, the probability vectors were averaged over all patches extracted from an image to obtain the probability vectors associated with the four histological types for each sample. The probability vectors for each sample from the high-magnification network and the attention model were fused using a multilayer perceptron network to generate a classification label. Obtained results on an independent test set demonstrated an average accuracy of 97.5% ± 1.0% for the proposed model. An average accuracy of 94.5%, 93.5%, and 96.3% was obtained, respectively, for the separate high- and low-magnification networks, and the multi-scale model without an attention mechanism. The results suggested that a multi-scale strategy coupled with an attention mechanism can improve the accuracy of deep learning models in classifying digital histology images.
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 | 0.001 |
| 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) | 0.001 | 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 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".