Brain diseases and diagnostics are a significant part of the healthcare market. In the US alone, over 6.5 million brain scans are performed annually (over $1.5B spent on imaging alone). This magnitude of data is hard for clinicians to thoroughly analyze, or compare with historical databases of healthy/pathological cases.
Our team is working with a US-based startup to address this issue, and develop tools that enable to autonomously analyze brain scans, screening them for a number of pathologies (such as tumors or aneurysms) and deriving insights from automated similarity comparison to datasets of pathological cases (including also comparing changes in patient images over time across the continuum of care) – potentially supporting doctors decision making on daily basis.
We’re focusing on processing medical 3D scans (with popular modalities like MRI or MRA), using image processing techniques and neural networks combined with large databases of historical medical data to segment both normal and pathological structures, as well as to detect anomalies in patient brains. Due to the regulatory reasons (FDA) and requirements we’ve taken additional steps in order to prepare a data versioning system – allowing for models and data traceability.
