DEEP-ENIGMA

We are happy to announce that we received a University Utrecht Special Interest Group seeding grant of €5,000 for our proposal entitled DEEP-ENIGMA: A DEEP nEural Network for Image seGMentation to classify and quantify Atherosclerotic disease based on high-resolution scanned histological slides in collaboration with Tim G.M. van de Kerkhof and Dr. Ayoub Bagheri, PhD.

Executive summary

Cardiovascular diseases (CVD) will top global charts of morbidity and mortality in all socio-economic classes by 2030. Atherosclerosis is the major underlying cause of cardiovascular diseases and results in atherosclerotic plaque formation. The extent and type of atherosclerosis is manually assessed through histological analysis, and histological characteristics are linked to major acute cardiovascular events (MACE).

Example of how image segmentation identifies and quantifies different plaque characteristics.

Currently, we are developing CONVOCALS (as part of the SIG-grant program), a convolutional neural network (CNN) aimed at processing and analyzing high-resolution images from scanned histological slides of plaques in order to predict MACE. A spin-off from CONVOCALS was the development of an image segmentation-based masking method to obscure non-tissue image-artefacts from analysis. A by-product of the image segmentation was the identification and quantification of cryptic substructures plaques, not readily visible to the naked eye. Pilot studies suggest these are representations of atherosclerotic characteristics that constitute a plaque, e.g. inflammatory cell influx and intraplaque hemorrhage.

Here we propose “DEEP-ENIGMA: a DEEP nEural network for Image seGMentation to classify and quantify Atherosclerotic disease based on high-resolution scanned histological slides”. DEEP-ENIGMA is a double edge-sword: it aims 1) to identify, classify, and quantify atherosclerotic subtypes using image-data from plaques, 2) use the newly defined subtypes to predict MACE.