abstract |
System and method for classifying pathological images. The image is received and normalized, and the normalized image is divided into regions. A disease vector is automatically determined for the plurality of regions by at least one classifier having a neural network. Each of the plurality of regions is automatically annotated based on the determined disease vector. The received image is automatically evaluated based at least on the annotation. The neural network is trained based at least on expert annotations for each region of the image according to at least one objective classification criterion. The image can be an eye image, a blood vessel image, or a fundus image. For example, the disease can be vascular disease, vascular disease, or diabetic retinopathy. [Selection diagram] Figure 1 |