Medical image processing, and in particular segmentation, is a crucial process for researchers across the globe. The time-intensiveness of segmentation creates a bottleneck for research pipelines that rely on this procedure, as manual segmentation can necessitate multiple days of human labor for a single scan.

Machine learning is a powerful class of algorithms that can learn to make predictions from
annotated data. These data-driven algorithms can augment traditional medical image processing by replacing cognitive labor with predictions from trained models. Unfortunately, model development comes with its own set of complications including specialized hardware requirements, complex data pipelines, expert knowledge, and iterative design procedures associated with hyperparameter tuning. As a result, there is significant overhead involved when implementing custom machine learning infrastructure in research environments.

The objective of Seg AI is to empower researchers to develop custom machine learning models for medical image analysis tasks. This is achieved by consolidating the entire machine learning life cycle into a single software solution such that model development is seamlessly integrated with the target processing task. Researchers can import or create annotated data, use this data to develop custom models with the Seg AI machine learning platform in the cloud, and download the trained models into the software for use in medical image processing pipelines.