Headquartered in North Chicago, Illinois, AbbVie is a Fortune 500 biopharmaceutical company dedicated to advancing healthcare through innovative research and the development of lifesaving treatments.

AbbVie researchers often work with special liquids that separate into two layers, called biphasic solutions. Measuring properties of biphasic solutions, such as phase boundaries (where two layers meet) and how layers are blended, is a manual process that limits high-throughput experimentation. The time this process takes inhibits AbbVie’s mission of providing quality healthcare.

Our Image Analysis Tool automates this process by using a deep learning model to acquire metrics from images of biphasic solutions that AbbVie’s chemists need to make critical decisions. First, a robotic arm positions each vial for the camera, capturing high-resolution images which are then processed by the model. Solution properties are located and displayed to the user for quick and accurate insights into the composition of the biphasic solution. Chemists save the image with its associated information, such as sample number or the chemicals used.

Our software also includes a model training feature, enabling researchers to adapt the model to specific chemicals and environments. This ensures that the model is as robust as possible, leading to the most precise and realistic results.

To support the model training feature, the application also supports the saving of images to a database. This enables users to create new datasets that can be used to train models. It also enables users to revert the software to use previous versions of the model depending on the current use scenario.

Flask, a Python webserver, constructs the back-end interface which applies the model to the acquired data. The front-end framework, Angular, enables users to view model results with ease.