Retinal Fundus image processing and ensemble learning: optic disc and optic cup detection
Bibliographic record
Abstract
Ophthalmologists have widely used retinal fundus imaging systems to examine the health of the optic nerve, vitreous, macula, retina and their blood vessels. Many critical diseases, such as glaucoma and diabetic retinopathy, can be diagnosed by analyzing retinal fundus images. Retinal image-based glaucoma detection is a comprehensive diagnostic approach that examines the head cup-to-disc ratio (CDR) as an important indicator for detecting the presence and the extent of glaucoma in a patient. The accurate segmentations of the optic disc (OD) and optic cup (OC) are critical for the calculation of CDR. Machine learning based algorithms can be very helpful to efficiently exploit the vast amounts of retinal fundus data. In this thesis project, the main goal is to develop image processing and machine learning algorithms to automatically detect OD and OC from fundus images. This goal has been achieved by developing and applying several image enhancement techniques. First, an algorithm is proposed and tested on several fundus images to detect OD. The proposed algorithm is based on a combination of Contrast Limited Adaptive Histogram Equalization (CLAHE), Alternating Sequential Filters (ASF), thresholding, and Circular Hough Transform (CHT) methods. The results section highlights that the proposed algorithm is highly efficient in segmentation of OD from other parts of the fundus image. Several classification and modeling methods are studied in order to classify detected OD into OC and non-OC regions. In this thesis project three main ensemble modeling algorithms are studied to segment OC. The studied ensemble models are Random Forest, Gradient Boosting Machines (GBM), and Extreme Gradient Boosting Machines (XGBoost). The comparison between these models shows that they have more accurate results than conventional classification methods such as Logistic Regression (LR) or Support Vector Machines (SVM). This study shows that XGBoost is the fastest and most accurate approach to segment optic cup within the optic disc region.
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 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) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| 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 itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, 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".