Processing images and building a Neural Network to identify diseases from images of Eye Sonography

The aim is to create a system that uses NLP on patient reports to classify patient data according to diseases, and to process the sonographic images of the eye, to confirm the presence of disease; to enable the doctor to compare the image analysis with the report and to localize the presence of disease within the sonographic image. The sonographic images of the eye are generated by doctors using ultrasound machines. Based on the image generated, they make inferences if diseases are present or not; if present, they are specified in the inference and stored within the patient's diagnostic report. Manually traversing through a considerable number of existing patient files for the purpose of arranging according to disease is infeasible. So a large number of patient reports are processed using NLP, and each patient data is classified according to the disease inferred by the doctor within the report [1]. Further, the model trained on the dataset of diseases is used to detect the presence of a particular disease without a doctor's intervention. The report generated by the doctor after identifying the disease and the disease detected by the neural network model is used to verify the doctor's report and vice-versa. An accurate model may be used to identify diseases in absence of the doctor or it may assist the doctor in some manner. The Webapp User Interface is associated with viewing the classified data quickly. The collective dataset of sonographic images [2],[3] of existing patient records is further used to train a model; a classifier model that takes the sonographic image of the eye then gives the result for the presence of disease from the trained classes.