With a Primer in Deep Learning for Medical Imaging
By Martin Willemink  |  Aug 05, 2021
With a Primer in Deep Learning for Medical Imaging
Image courtesy of and under license from Shutterstock.com
Artificial intelligence (AI) has garnered substantial interest in medical imaging circles. The engineering of computerized systems to perform tasks previously requiring human intelligence, popular applications of machine learning (ML) and deep learning (DL) in medical imaging include workflow improvements, automatic lesion detection, and automatic quantification.

PALO ALTO, CALIFORNIA - Artificial intelligence (AI) has garnered substantial interest in the medical imaging field. Broadly defined as the engineering of computerized systems to perform tasks that previously would require human intelligence,1 popular applications of machine learning (ML) and deep learning (DL) in medical imaging include workflow improvements, automatic lesion detection, and automatic quantification. Examples of workflow improvements are prioritizing radiologists’ worklists,2 triaging breast cancer screening mammography exams,3 and radiation dose reduction of computed tomography (CT) exams.4 

Automatic lesion detection has been applied to intracranial hemorrhage5 and pneumothorax,among other uses. Coronary calcium quantification with CT7 and prostate classification with magnetic resonance imaging (MRI)8 are two examples of automatic quantification.

Potential applications of ML and DL in medical imaging thus clearly abound, but many clinicians, radiologists, and other healthcare professionals do not know how start with DL for medical imaging.


Starting Off

The process of developing a DL algorithm requires multiple steps, as the table below shows. The first is to determine project scope. Select a clinically relevant problem that can be tackled by one ML/DL approach or a combination of multiple ones: segmentation, detection, classification, or prediction. The second need is to build a team with the right people to ensure the availability of clinical, imaging, and technical expertise. The third step is to select data required, whether they are public data or from a local hospital. The final stage is to train, validate, and test, and apply the model in its clinical setting.

The content herein is subject to copyright by The Yuan. All rights reserved. The content of the services is owned or licensed to The Yuan. The copying or storing of any content for anything other than personal use is expressly prohibited without prior written permission from The Yuan, or the copyright holder identified in the copyright notice contained in the content.
Continue reading
Sign up now to read this story for free.
Get started