Using YOLO Few_Shot Learning for Labeling Inflammatory Bowel Diseases
Bibliographic record
Abstract
Inflammatory bowel disease (IBD) is a chronic deteriorating condition that affects millions of people all over the world. The application of machine learning to IBD has the potential to arrive at effective treatment response, accurate diagnosis and improving patient survival rates. Inflammatory Bowel Disease (IBD) includes Crohn's disease and ulcerative colitis, two chronic conditions causing inflammation of the digestive tract. Crohn's disease can affect any part of the GI tract with deep, patchy inflammation, while ulcerative colitis is limited to the large intestine's inner lining, causing continuous inflammation. A third, less common type is indeterminate colitis, where features of both are present. Machine learning models can automate the review of images and biopsy slides taken from IBD procedures, identify novel biomarkers, and create predictive models for patient outcomes, helping clinicians tailor therapies and interventions for better patient care. Indeed, machine learning algorithms perform poorly when applied to medical applications due to unavailability of large annotated samples. However, learning algorithms like YOLO is inherently better at learning from small sample sizes than other machine learning models. This study adopt a variant of the YOLO model (originally pre-trained using the COCO dataset) that has been fine-tuned using the Hyper-Kvasir IBD dataset. Moreover, the fine-tuned model is used then to classify the GI conditions using a Triplet-Loss Siamese Neural Network (SNN). The integration between the YOLO model and the SNN has been simplified using the RoboFlow environment to assist expert to verify the YOLO annotations that will involve not only the ground truth but also cases that may be mistakenly classified as one of the model classes. Our fine tuning and SNN model shows much better classification of IBD region of interest even for those having a variable shape like ulcerative colitis.
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 | 0.002 | 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".