Synthesized X-Ray Images Help Train AI Programs
|
By HospiMedica International staff writers Posted on 17 Jul 2018 |
|
Image: Real X-ray image (L) next to a synthesized X-ray created by DCGAN. Underneath the X-ray images are the corresponding heatmaps (Photo courtesy of Hojjat Salehinejad/MIMLab).
A new study describes how computer generated X-rays can be used to augment artificial intelligence (AI) training sets.
In order to generate and continually improve artificial X-rays, researchers at the University of Toronto (Canada) used deep convolutional generative adversarial network (DCGAN) algorithms, which are made up of two networks: one that generates the images, and the other that tries to discriminate synthetic images from real images. The two networks are continuously trained until they reach a point in which the discriminator cannot differentiate real images from synthesized ones. Once a sufficient number of artificial X-rays are created, they are used to train another DCGAN that can classify the images accordingly.
The researchers then compared the accuracy of the artificially augmented dataset to the original one when fed through their AI system, and found that classification accuracy improved by 20% for common conditions. For some rare conditions, accuracy improved up to 40%. An advantage of the method is that as the synthetic X-rays are not real, the dataset can be readily available to researchers outside hospital premises without violating privacy concerns. The study was presented at the IEEE International Conference on Acoustics, Speech and Signal Processing, held during April 2018 in Calgary (Canada).
“In a sense, we are using machine learning to do machine learning,” said senior author and study presenter Professor Shahrokh Valaee, PhD, of the Machine Intelligence in Medicine Lab (MIMLab). “We are creating simulated X-rays that reflect certain rare conditions so that we can combine them with real X-rays to have a sufficiently large database to train the neural networks to identify these conditions in other X-rays.”
“Deep learning only works if the volume of training data is large enough, and this is one way to ensure we have neural networks that can classify images with high precision,” concluded Professor Valaee. “We've been able to show that artificial data generated by deep convolutional GANs can be used to augment real datasets. This provides a greater quantity of data for training and improves the performance of these systems in identifying rare conditions.”
Related Links:
University of Toronto
In order to generate and continually improve artificial X-rays, researchers at the University of Toronto (Canada) used deep convolutional generative adversarial network (DCGAN) algorithms, which are made up of two networks: one that generates the images, and the other that tries to discriminate synthetic images from real images. The two networks are continuously trained until they reach a point in which the discriminator cannot differentiate real images from synthesized ones. Once a sufficient number of artificial X-rays are created, they are used to train another DCGAN that can classify the images accordingly.
The researchers then compared the accuracy of the artificially augmented dataset to the original one when fed through their AI system, and found that classification accuracy improved by 20% for common conditions. For some rare conditions, accuracy improved up to 40%. An advantage of the method is that as the synthetic X-rays are not real, the dataset can be readily available to researchers outside hospital premises without violating privacy concerns. The study was presented at the IEEE International Conference on Acoustics, Speech and Signal Processing, held during April 2018 in Calgary (Canada).
“In a sense, we are using machine learning to do machine learning,” said senior author and study presenter Professor Shahrokh Valaee, PhD, of the Machine Intelligence in Medicine Lab (MIMLab). “We are creating simulated X-rays that reflect certain rare conditions so that we can combine them with real X-rays to have a sufficiently large database to train the neural networks to identify these conditions in other X-rays.”
“Deep learning only works if the volume of training data is large enough, and this is one way to ensure we have neural networks that can classify images with high precision,” concluded Professor Valaee. “We've been able to show that artificial data generated by deep convolutional GANs can be used to augment real datasets. This provides a greater quantity of data for training and improves the performance of these systems in identifying rare conditions.”
Related Links:
University of Toronto
New
Gold Member
Handheld Blood Glucose Analyzer
STAT-Site
New
Vessel Sealing Instrument
ERGOseal
New
Patient Preoperative Skin Preparation
BD ChloraPrep
Channels
Critical Care
view channel
Noninvasive Monitoring Device Enables Earlier Intervention in Heart Failure
Hospitalizations for heart failure with preserved ejection fraction (HFpEF) remain common because lung congestion often worsens before symptoms prompt treatment changes. Missed early decompensation... Read more
Automated IV Labeling Solution Improves Infusion Safety and Efficiency
Medication administration in high-acuity settings is often complicated by multiple concurrent infusions, making accurate line identification essential. In a 10-hospital intensive care unit study, 60% of... Read more
Surgical Techniques
view channel
Ultrasound Technology Aims to Replace Invasive BPH Procedures
Benign prostatic hyperplasia (BPH) is a frequent cause of lower urinary tract symptoms in aging men and often requires invasive procedures or prolonged recovery. With prevalence expected to rise as populations... Read more
Continuous Monitoring with Wearables Enhances Postoperative Patient Safety
Postoperative hypoxemia on general surgical wards is common and often missed by intermittent vital sign checks. Undetected low oxygen levels can delay recovery and raise the risk of complications that... Read more
Patient Care
view channel
Wearable Sleep Data Predict Adherence to Pulmonary Rehabilitation
Chronic obstructive pulmonary disease (COPD) is a long-term lung disorder that makes breathing difficult and often disturbs sleep, reducing energy for daily activities. Limited engagement in pulmonary... Read more
Revolutionary Automatic IV-Line Flushing Device to Enhance Infusion Care
More than 80% of in-hospital patients receive intravenous (IV) therapy. Every dose of IV medicine delivered in a small volume (<250 mL) infusion bag should be followed by subsequent flushing to ensure... Read more
Health IT
view channel
EMR-Based Tool Predicts Graft Failure After Kidney Transplant
Kidney transplantation offers patients with end-stage kidney disease longer survival and better quality of life than dialysis, yet graft failure remains a major challenge. Although a successful transplant... Read more
Printable Molecule-Selective Nanoparticles Enable Mass Production of Wearable Biosensors
The future of medicine is likely to focus on the personalization of healthcare—understanding exactly what an individual requires and delivering the appropriate combination of nutrients, metabolites, and... Read more
Business
view channel
GE HealthCare Leads Major European Initiative to Advance Cardio-Oncology Care
Cardiovascular complications are increasingly recognized among people undergoing or recovering from cancer treatment, driven by pre‑existing disease and the cardiotoxic effects of chemotherapy, radiotherapy,... Read more
