Louis Blankemeier, CEO and Co-Founder of Cognita, describes the significant gap between the number of radiologists and the rising volume and types of medical imaging that need to be reviewed. The Cognita generative visual language model is an advancement over earlier radiology AI applications that were designed to flag single findings. This integrated approach emulates a radiologist by analyzing complex images and generating full draft radiology reports, demonstrating reduced time per case, increased detection of critical findings, and decreased cognitive burden on radiologists.

Louis explains, "Radiologists look at a number of different imaging types. So there are X-rays, and these are basically 2D images or 2D projections of the inside of the body. So you see all the organs superimposed on each other in a 2D image. And the radiologists have to take this 2D image and create almost a 3D representation in their head to understand what's going on in the body. They also read CT scans, which use X-ray radiation but take multiple images from different angles of the body. And then you basically reconstruct a 3D video that shows you what the inside of a body looks like. And then there are also MRI scans, which use magnetism to create a video of the inside of the body. And CT and MRI are 3D imaging modalities. So they're both kind of like videos. And you have an ultrasound, which uses sound waves, and you have a long tail of different imaging types. But radiologist spend most of their time looking at X-rays, CT scans, and MRI images."

"A vision language model is a model that has eyes, so it can actually look at images, and then the language part describes how we are generating text in the output. And we can actually add one more descriptor to vision language. We can add the term generative. So we're actually generating text in the output of our model. We're generating the radiology report. And this is in contrast to most models out there today that are discriminative. There are these classifier models that are saying, " Is there a disease present or not? And the output is binary. It's zero one. It's not a text report in the output."

#Cognita #AIinRadiology #GenerativeAI #VisionLanguageModels #RadiologyWorkflow #DiagnosticImaging #MedicalAI #HealthcareInnovation #RadiologistSupport #ClinicalDecisionSupport #PatientSafety #BurnoutReduction #FDA #BreakthroughDevice #DigitalHealth #HealthTech

Cognita.ai

Listen to the podcast here