Over a million developers have joined DZone.
{{announcement.body}}
{{announcement.title}}

Using Machine Learning to Spot Pneumonia

DZone's Guide to

Using Machine Learning to Spot Pneumonia

Researchers hope that this tool will significantly reduce the number of pneumonia cases that go undiagnosed and speed up attempts to treat those with the condition.

· AI Zone ·
Free Resource

Start coding something amazing with the IBM library of open source AI code patterns.  Content provided by IBM.

Early forays of AI into healthcare have typically followed a similar path. Algorithms would be trained on vast quantities of medical data (images, usually) and would then use computational might to spot early signs of an illness faster than human doctors are capable of doing.

The latest project of this ilk comes from a Stanford team that has developed an algorithm called ChexNet that diagnoses pneumonia from chest X-rays.

"Interpreting X-ray images to diagnose pathologies like pneumonia is very challenging, and we know that there's a lot of variability in the diagnoses radiologists arrive at," the team says.

The algorithm was trained using the publicly available data held at the National Institutes of Health Clinical Center. The database contains over 110,000 X-ray images that are each labeled with one of 14 possible pathologies. The algorithm is capable of analyzing each image and returning a diagnosis for any of these pathologies.

Signs of Progress

The work is interesting because it builds upon initial efforts by the Center to do this themselves. The data and their own algorithm were released in September 2017, and the Stanford team has managed to significantly improve its productivity in a very short period of time.

The team believes that their work is especially useful, as spotting it early enough is very difficult from just X-ray images. They were able to rapidly improve upon the work done by the Center to the extent that within a week, they had developed an algorithm that could accurately diagnose ten of the pathologies, with the full 14 delivered in just over a month. Each of the diagnoses was more accurate than current state-of-the-art tests.

The interpretation of X-rays is a crucial part of the diagnosis of many diseases. Currently, however, that analysis largely relies on the skills of the radiologist — as talented as they are, they are nonetheless fallible.

"The motivation behind this work is to have a deep learning model to aid in the interpretation task that could overcome the intrinsic limitations of human perception and bias, and reduce errors," the authors say.

After a month of tweaking and improvements, the system was consistently able to outperform a number of experienced radiologists in detecting pneumonia.

X-Ray Heat Maps

The team has also developed a digital tool that is capable of producing a heat map-style image from the X-ray. The "temperature" of the image correlates with the areas most likely to represent pneumonia.

They hope that the tool will significantly reduce the number of pneumonia cases that go undiagnosed and speed up attempts to treat those with the condition.

"We plan to continue building and improving upon medical algorithms that can automatically detect abnormalities and we hope to make high-quality, anonymized medical datasets publicly available for others to work on similar problems," they explain. "There is massive potential for machine learning to improve the current health care system, and we want to continue to be at the forefront of innovation in the field."

Start coding something amazing with the IBM library of open source AI code patterns.  Content provided by IBM.

Topics:
machine learning ,ai ,healthcare ,algorithms ,deep learning

Published at DZone with permission of

Opinions expressed by DZone contributors are their own.

{{ parent.title || parent.header.title}}

{{ parent.tldr }}

{{ parent.urlSource.name }}