Medical Image Search Powered by Deep Learning

Facing increasing workloads, radiologists must choose between longer working hours or decreased time spent evaluating images. About 20% of cases require additional research from many sources, requiring up to 20 minutes each with a questionable rate of success.

contextflow’s image search engine uses deep learning to put the knowledge encoded in millions of medical images and reports at radiologists' fingertips, saving them time and money. Simply mark a region of interest in an image, and our search engine instantly returns reference cases and associated knowledge. No more paging through books, guessing the right keyword or consulting various text-based reference search engines. Case relevant information is now only a few clicks away! In addition, our software can be configured to allow for prioritization of cases where patterns of interest are detected.

contextflow’s patent-pending 3D image-based information search solution is currently being validated on Lung CT scans and can be extended to additional pathologies and organs. We enable clinics to share information across institutional borders and benefit from a collaborative growing knowledge base, generating clinical value from existing data.

Schlagworte

Website

http://www.contextflow.com

Temporal Trajectories for Radiology Image Search (TeTRIS)

The TeTRIS project will develop a plug-in for contextflow's RadiologyExplorer software that supports pulmonologists in the analysis of temporal imaging information, that is, Computed Tomography (CT) Images of a single person taken at multiple points in time to follow disease or treatment progression. The main development is a structured patient trajectory that automatically aligns multiple CT Images taken at different times with key information extracted from the text of clinical reports associated with these images.

The plug-in will also provide a search functionality that will look through hospital archives for the anonymised patient trajectories most similar to the query trajectory to provide pulmonologists with key information to assist them in selecting the best treatments. The development will be done in an iterative way with extensive input from pulmonologists at each step, from the specification of the system to its evaluation. The technology developed in TeTRIS will be essential to using medical imaging optimally in the rapidly emerging field of personalised medicine.

Alleinstellungsmerkmal

contextflow is unique in providing large-scale, real-time visual similarity search technology for radiology. The TeTRIS project takes this technology a step further, allowing information related to changes over time in images of a particular patient to also become searchable as a support for diagnosis.

Schlagworte

deep learning (-), image processing (Bildverarbeitung), medical imaging (-), radiology (Radiologie, Bildgebende Verfahren), diagnostics (Diagnostik), machine learning (ml), Artificial intelligence (AI)