This model is a proof of concept for a client and is a great example of how computer vision can automate medical scan reads. Here I am leveraging a pretrained Fast R-CNN model and fine tuning it on a Dataset consisting of Brain Scans.
The model learns to detect cancer lesions and for the larger lesions has an Average Precision and Recall of 90%. The performance significantly improved for the model in production with higher computing power and better data.
Existing Object Detection models can be leveraged to get highly accurate results for most types of data.