We wanted to build some model using machine learning that would help the society in predicting some information. We came across the diabetic retinopathy(DR) dataset that can be used to predict if the disease has occurred. Currently, detecting DR is a time-consuming and manual process that requires a trained clinician to examine and evaluate digital color fundus photographs of the retina. By the time human readers submit their reviews, often a day or two later, the delayed results lead to lost follow up, miscommunication, and delayed treatment. With our model, this time consumption and manpower can be reduced to a great extent.
What it does
We train on images of eyes that are affected and not affected by DR and build a predictive model. The clinician can use the image of patient as input to the predictive model to predict the occurrence of DR.
How I built it
I used image processing techniques like image denoising, discrete wavelet transform using Haar wavelet,etc to retrieve the main information that is affecting the label. We then run SVM and KNN machine algorithms on this data to build a predictive model. This model is used to predict if there is occurrence of DR or not based on the input image given.
Challenges I ran into
The dataset is very large. It took a lot of time to run. We had to use the subset of a dataset to train our model. Integrating the technologies of front end and back end.
Accomplishments that I'm proud of
We are still in the process of building the model. If this works with great accuracy then I would be proud to be contributing something in the field of health.
What I learned
Learnt about how so many technologies can be used to achieve the task. Also integration of these technologies, from front end to back end.