MACHINE LEARNING – BRAINSCAN PROJECT

WE USE MACHINE LEARNING TO REDUCE INTERPRETATION TIME AND COST OF BRAIN COMPUTED TOMOGRAPHY

PLAY VIDEO

HOW IT WORKS?

BrainScan provides automatic search for similar computed tomography (CT) or magnetic resonance imaging (MRI) scans in large data sets. It searches for medical cases that are similar to a given scan or region of interest (ROI).

Doctors can easily find historical cases and examinations that showed similar lesions or abnormal changes to those from the examined patient, which can be extremely useful in the diagnosis process.

A doctor loads scans into the system and optionally selects a region of interest (ROI) of a lesion. BrainScan immediately searches the entire archive of computed tomography (CT) or magnetic resonance imaging (MRI) scans for similar occurrences.

In a matter of seconds, the retrieved scans are presented to the doctor along with the location of patches similar to the requested region of interest (ROI) as well as important details regarding the exams including diagnosis, treatment and outcome.

TECHNOLOGY

In order to provide instant retrieving of search results, the entire database is firstly extensively indexed with a set of dedicated signatures describing contents of a scan, which are computed over full scans as well as numerous subareas of different sizes. Similar indexing procedure is then applied for the searched region of interest (ROI) which then enables fast comparing with data in the index and returning results.

We use most efficient deep learning techniques including 3D Convolutional Neural Networks. These networks are being trained on vast amounts of 3D scans, Deep Networks provide great accuracy in summarizing 3D content and fast comparing of data samples. They are also computationally efficient methods for fast processing of large collections of data.

PARTNERSHIP

We cooperate with worldwide companies specializing in medical imaging diagnostics. Companies that usually operate for several decades on the market. They share their knowledge and experience with us as well as povide collected data sets for training.

CONTACT US