Multi-Class ensemble classification framework with multi-layered feature extraction and ranking for Diabetic retinopathy data
International Journal of Drug Delivery Technology · 17.06.2026 · 16 (55s)
DOI: 10.25258/ijddt.16.55s.74 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Drug Delivery Technology, online erschienen am 17.06.2026. Verfasst von Deepthi Kalwa und 2 weiteren Autor:innen.
Abstract (Originalsprache)
Diabetic retinopathy (DR) remains one of the most prevalent causes of vision loss among adults, especially those with long-standing diabetes. It manifests through progressive damage to the small blood vessels in the retina, resulting in visible lesions like microaneurysms, hemorrhages, and both soft and hard exudates. Fundus imaging, a non-invasive diagnostic method, has become essential in identifying these retinal abnormalities. Recent improvements in image contrast techniques have enhanced the visualization of subtle lesions, enabling more precise and earlier detection of DR. By highlighting the morphological differences in the retinal structure, contrast-enhanced fundus images serve as a robust foundation for automated diagnostic systems and clinical evaluations, ultimately aiding in the timely management of DR to prevent irreversible blindness.Diabetic Retinopathy (DR) is a microvascular complication arising from prolonged diabetes, primarily affecting the retinal blood vessels. The probability of developing DR increases significantly with the patient's age and the number of years since being diagnosed with diabetes. Elevated and poorly managed blood glucose levels, along with erratic fluctuations in blood pressure, contribute to vascular damage, making the eyes more susceptible to retinopathy. The interplay between these physiological factors leads to the progressive degeneration of retinal capillaries, emphasizing the importance of long-term metabolic control in preventing or delaying the onset of DR.Early detection and classification of Diabetic Retinopathy (DR) are crucial in preventing irreversible vision loss. DR progresses through distinct stages, ranging from mild non-proliferative signs to advanced proliferative forms, each requiring specific clinical interventions. Manual diagnosis using retinal fundus images is not only time-consuming and labor-intensive but also depends on the availability of trained ophthalmologists. The inherent subjectivity in human assessment and the growing number of diabetic patients make manual screening an unsustainable solution. Therefore, automated systems based on image classification offer a promising alternative for timely and accurate identification of DR stages, enabling scalable and efficient screening protocols in clinical practice.Diagnosing Diabetic Retinopathy (DR) is often hindered by irregularities in contrast and brightness in fundus images, which obscure the detection of lesions. This study introduces a novel technique involving multi-level threshold-based filtering combined with an ensemble feature extraction framework. Four distinct sets of image features are utilized to accurately segment the retinal regions. A region-growing method, refined through threshold optimization, eliminates over-segmented and noisy areas, thereby enhancing the reliability of feature selection. These refined features are then fed into an ensemble classification model to improve diagnostic accuracy. Experimental validation on diverse DR datasets confirms that this method surpasses traditional techniques in precision and robustness, offering a more dependable solution for early DR detection.
Zitieren
APA
Kalwa, D., Josephine, M., & JeyabalaRaja, V. (2026). Multi-Class ensemble classification framework with multi-layered feature extraction and ranking for Diabetic retinopathy data. International Journal of Drug Delivery Technology, 16(55s). https://doi.org/10.25258/ijddt.16.55s.74
Vancouver
Kalwa D, Josephine M, JeyabalaRaja V. Multi-Class ensemble classification framework with multi-layered feature extraction and ranking for Diabetic retinopathy data. International Journal of Drug Delivery Technology. 2026;16(55s). doi:10.25258/ijddt.16.55s.74
BibTeX
@article{Kalwa2026,
author = {Kalwa, Deepthi and Josephine, M.S. and JeyabalaRaja, V.},
title = {Multi-Class ensemble classification framework with multi-layered feature extraction and ranking for Diabetic retinopathy data},
journal = {International Journal of Drug Delivery Technology},
year = {2026},
volume = {16},
number = {55s},
doi = {10.25258/ijddt.16.55s.74}
}
EUVIMED