Feature | Information Technology | July 01, 2019 | By Greg Freiherr

Are Shallow Networks Better Than Deep Ones?

“Shallow” learning promises better artificial intelligence, say SIIM presenters

David Anderson at SIIM

David Anderson presents information about the development of artificial Intelligence (AI) at SIIM 2019 as his father, Charles, looks on from a front row seat. The father-son team presented information at the SIIM annual meeting about their use of relatively shallow Deep Learning networks to analyze chest radiographs with AI.

Greg Freiherr

Greg Freiherr

Two simple neural networks are better than one complex one, according to a father-son team of entrepreneurs. On June 27 the two described “A Two-Stage Deep Learning Approach to Chest X-Ray Analysis” during the 2019 meeting of the Society of Imaging Informatics in Medicine in Denver.

 

Diagnostic decisions coming from models built using two “shallow” neural networks can lead to faster, more accurate and more interpretable artificial intelligence for radiography and other imaging modalities, according to David Anderson.

“When you have a literal deep (learning) network you have lots and lots of computational nodes and lots of layers of those nodes. A shallower network means that there are fewer computations so it is less math going on, which means you can look inside of it and see what it is doing — you can pull it apart,” Anderson told Imaging Technology News after the SIIM presentation.

He and his father, Charles, described their work at SIIM 2019 as a tag team — David presenting technical information first; Charles Anderson following with broader perspectives. Both men work at privately held Pattern Exploration, where Charles Anderson, Ph.D., is CEO; David is the machine learning developer.

David cited research1 & 2 into the automated analysis of chest X-ray images as the basis for the work done at Pattern Exploration. The use of “shallow” networks allows training with unlabelled data, which makes training less difficult and more efficient, he said. The use of these networks can also make the logic underlying AI less opaque.

 

Making the “Box” Transparent

Many have described deep learning as leading to algorithms that are “black boxes.” If shallower networks are used, however, the result “is not black at all — if you have the right tools,” David Anderson told ITN. “You can say, ‘Here is what this network is doing.’”

The two, five-layer networks that the Andersons developed to model X-ray analysis represent deep learning despite their relatively shallow depth. (Deep learning is a subset of machine learning, which is a type of artificial intelligence.) These two networks have separate tasks. One aligns the images into a standard orientation so the patient spine, for example, is presented vertically. The other does the classification.

Charles, who also serves as a professor of computer science at Colorado State University, focused on the potential impact of the research. He also raised the potential utility of a verbal interface. The computer science professor described during his presentation how such a verbal interface might be used. “It would allow you to ask a quick question — like you do with Siri — and come back with an answer in a spoken language,” he summarized later for ITN.

Supporting verbal Q & A would allow the radiologist to focus on the screen, rather than intermittently shifting attention to input devices. Radiologists have told Charles that their priorities are speed and confidence that their decisions are correct.

 

Greg Freiherr is a contributing editor to Imaging Technology News (ITN). Over the past three decades, he has served as business and technology editor for publications in medical imaging, as well as consulted for vendors, professional organizations, academia, and financial institutions.

 

Related content and references:

1. P. Rajpurkar, et al. (2017) CheXNet: Radiologist-Level Pneumonia Detection on Chest X-Rays with Deep Learning. https://arxiv.org/abs/1711.05225v3

2 I. M. Baltruschat, et al. (2018) Comparison of Deep Learning Approaches for Multi-Label Chest X-Ray Classification. https://arxiv.org/abs/1803.02315v1

How AI Might Provide a Safety Net for Patients and Providers

Editor’s note: This article is the ninth piece in a content series by Greg Freiherr covering the Society for Imaging Informatics in Medicine (SIIM) conference in June.

 

Related content:

How AI Might Provide a Safety Net for Patients and Providers 

Smart Algorithm Extracts Data from Radiology Reports 

PODCAST: Why Blockchain Matters In Medical Imaging

PODCAST: How to Fix Your Enterprise Imaging Network

PODCAST: 5 Low-Cost Ways To Slow Hackers

Cinebot: Efficient Creation of Movies and Animated Gifs for Presentation and Education Directly from PACS

DeepAAA Uses AI to Look Automatically For Aneurysms

Making AI Safe, Effective and Humane for Imaging

Related Content

Novel Coronavirus 2019-nCoV Pneumonia

Image by _freakwave_ from Pixabay 

News | Computed Tomography (CT) | February 16, 2020
February 16, 2020 — The following statement was issued by the U.S.
negative RT-PCR results and chest CT findings compatible with 2019-nCoV pneumonia

Figure 1: Patient flowchart. Of 167 patients screened, 5 (3%) had negative RT-PCR results and chest CT findings compatible with 2019-nCoV pneumonia. Chart courtesy of Radiology

Feature | Computed Tomography (CT) | February 14, 2020
As the 2019-nCoV Pneumonia is taking the world by storm, researchers have found a possible way to predict this virus
SIR-Spheres Y-90 resin

SIR-Spheres Y-90 resin microspheres are released into the hepatic artery.

News | Nuclear Imaging | February 14, 2020
February 14, 2020 —  ...
MolecuLight’s i:X procedure in wound care
News | Radiology Imaging | February 13, 2020
February 13, 2020 — MolecuLight Inc., a leader in handheld fluorescence imaging for real-time detection of bacteria i
Hyperfine Research, Inc. announced that it has received U.S. Food and Drug Administration (FDA) 510(k) clearance for the world’s first bedside Magnetic Resonance Imaging (MRI) system

Hyperfine's point-of-care MRI wheels directly to the patient’s bedside, plugs into a standard electrical wall outlet, and is controlled via a wireless tablet. Photo courtesy of Business Wire

News | Magnetic Resonance Imaging (MRI) | February 12, 2020
February 12, 2020 — Hyperfine Research, Inc. announced that i
Mobile devices proved both reliable and accurate for the clinical decision to administer IV thrombolysis in patients with acute stroke

Appearance of same unenhanced CT scan on three reading systems: E-2620 monitor (Barco) (A), Galaxy S8 Plus (Samsung) smartphone (B) and ThinkPad T460s laptop computer (Lenovo) (C).

News | Computed Tomography (CT) | February 12, 2020
February 12, 2020 — Mobile devices proved both reliable and accurate for the clinical decision to administer IV throm
The Candelis ImageGrid Plus PACS Server is an ultra-high-performance platform that can support high volume healthcare environments of 1,000 plus modalities
News | PACS | February 12, 2020
February 12, 2020 — The Candelis ImageGrid Plus...
The magnetic resonance imaging (MRI) contrast agents market is expected to grow rapidly

Image courtesy of GE Healthcare

News | Magnetic Resonance Imaging (MRI) | February 11, 2020
February 11, 2020 — The magnetic resonance imaging (MRI) contrast agents market is expected to grow rapidly in the fo
CT image of Novel Coronavirus 2019-nCoV from the Radiology article showing a baseline CT image of a 75 year old male with multiple patchy areas of pure ground glass opacity (GGO) and GGO with reticular and/or interlobular septal thickening. Follow-up CT images on day 3 after admission show an overlap of organizing pneumonia with diffuse alveolar damage in that it is more diffuse and associated with underlying reticulation. Read more and see 15 more images from novel coronavirus patients in the article.

An image from the Radiology article showing a baseline CT image of a 75 year old male with multiple patchy areas of pure ground glass opacity (GGO) and GGO with reticular and/or interlobular septal thickening. Follow-up CT images on day 3 after admission show an overlap of organizing pneumonia with diffuse alveolar damage in that it is more diffuse and associated with underlying reticulation. Read more and see 15 more images from novel coronavirus patients in the article.

Feature | Computed Tomography (CT) | February 11, 2020
February 11, 2020 — The Radiological Society of North America (RSNA) jo