AI

Business Objective: The objective of this POC is to showcase that AI - machine and deep learning can be utilized in a reasonable way both in terms of time and cost to get insights into business. If you have an AI need but are unsure of methodology, please contact us.

Summary: This initiative's objective was to identify pneumonia in dataset of pediatric x-rays in a quick and effective manner. There are two main phases of the development.

1. Tag or label the data: The old principle of garbage in garbage out (GIGO) applies to AI as well. Before we could submit the unstructured data (x-ray images) to the AI api, it had to be enriched by tagging or labeling. It could be completed via Python, Panda libraries, or Figure 8. Click here for more information.

2. Vision API: Once the xrays were labelled we then used Google Cloud Platform(GCP) to identify the pneumonia. Actually four different sets of x-rays were used to train the model to catch discrepancies. The labelled x-ray image dataset was uploaded to the GCP platform, attached to a cloud bucket, then the CNN modeled was trained and the results (confusion matrix, precision, and recall) were presented for evaluation. The results are presented here.