Enhancing Choroidal Nevi Segmentation in Fundus Images Using YOLO
Bibliographic record
Abstract
Choroidal nevus often appears as a darkly pigmented, benign ocular lesion, which may progress to malignant forms, such as choroidal melanoma. Prompt and precise diagnosis of choroidal melanoma cannot be overstated as untreated cases can lead to vision loss and even life-threatening metastasis, under-scoring the importance of regular screening of the eye. However, this procedure is performed manually, which can be time-consuming and prone to human errors. Recent advancements in deep learning show potential for detecting eye diseases, including choroidal nevi. However, these models require extensive labelled data, which can be difficult to acquire due to the associated labelling costs. In this paper, we utilize two approaches to tackle these challenges. Firstly, we leverage a pre-trained YOLOv8 segmentation model and train it on both patches and full-size high-resolution colour fundus images. This strategy effectively expands the dataset size and allows the model to focus on the fine details of lesions within individual patches while understanding the general shape of the lesions by analyzing the entire image. Secondly, we use data augmentation to further expand the dataset size and tackle the class imbalance problem. Additionally, through the utilization of post-processing techniques, we enhance the predicted masks by addressing any potential flaws. This approach resulted in a 0.833 Dice Coefficient Score and a 0.714 Intersection Over Union (IOU) in our initial dataset and a Dice score of 0.764 and IOU of 0.618 on our second test set collected from a different site. In both datasets, this approach surpassed the models trained exclusively on full-size 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 | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.002 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.001 | 0.000 |
| 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".