Diabetic Retinopathy Features Segmentation without Coding Experience with Computer Vision Models YOLOv8 and YOLOv9 DOI Creative Commons
Nicola Rizzieri, Luca Dall’Asta,

Māris Ozoliņš

et al.

Vision, Journal Year: 2024, Volume and Issue: 8(3), P. 48 - 48

Published: Aug. 23, 2024

Computer vision is a powerful tool in medical image analysis, supporting the early detection and classification of eye diseases. Diabetic retinopathy (DR), severe disease secondary to diabetes, accompanies several signs eye-threatening conditions, such as microaneurysms (MAs), hemorrhages (HEMOs), exudates (EXs), which have been widely studied targeted objects be detected by computer models. In this work, we tested performances state-of-the-art YOLOv8 YOLOv9 architectures on DR fundus features segmentation without coding experience or programming background. We took one hundred images from public MESSIDOR database, manually labelled prepared them for pixel segmentation, abilities different model variants. increased diversity training sample data augmentation, including tiling, flipping, rotating images. The proposed approaches reached an acceptable mean average precision (mAP) detecting lesions MA, HEMO, EX, well hallmark posterior pole eye, optic disc. compared our results with related works literature involving neural networks. Our are promising, but far being ready implementation into clinical practice. Accurate lesion mandatory ensure correct diagnoses. Future will investigate further, especially MA improved extraction techniques, pre-processing, standardized datasets.

Language: Английский

Diabetic Retinopathy Features Segmentation without Coding Experience with Computer Vision Models YOLOv8 and YOLOv9 DOI Creative Commons
Nicola Rizzieri, Luca Dall’Asta,

Māris Ozoliņš

et al.

Vision, Journal Year: 2024, Volume and Issue: 8(3), P. 48 - 48

Published: Aug. 23, 2024

Computer vision is a powerful tool in medical image analysis, supporting the early detection and classification of eye diseases. Diabetic retinopathy (DR), severe disease secondary to diabetes, accompanies several signs eye-threatening conditions, such as microaneurysms (MAs), hemorrhages (HEMOs), exudates (EXs), which have been widely studied targeted objects be detected by computer models. In this work, we tested performances state-of-the-art YOLOv8 YOLOv9 architectures on DR fundus features segmentation without coding experience or programming background. We took one hundred images from public MESSIDOR database, manually labelled prepared them for pixel segmentation, abilities different model variants. increased diversity training sample data augmentation, including tiling, flipping, rotating images. The proposed approaches reached an acceptable mean average precision (mAP) detecting lesions MA, HEMO, EX, well hallmark posterior pole eye, optic disc. compared our results with related works literature involving neural networks. Our are promising, but far being ready implementation into clinical practice. Accurate lesion mandatory ensure correct diagnoses. Future will investigate further, especially MA improved extraction techniques, pre-processing, standardized datasets.

Language: Английский

Citations

4