Healthcare Trends in Neural Networks

Last year I had the opportunity to speak at a large healthcare technology conference. The audience was primarily comprised of healthcare professors, clinical researchers, and medical students. One of the biggest challenges for these healthcare professionals and those in healthcare research is understanding the impact Artificial Intelligence (AI) and deep learning (DL) will have in their day to day activities. Clearly AI is booming in every industry, transforming Enterprise IT, and healthcare is no different — whether it’s a medical research lab searching for faster insights or a hospital embracing AI and DL to augment practices and resources.

Healthcare offers some of the biggest opportunities for AI and DL to make positive impacts in human lives. Whether the impacts come from aiding in quicker diagnosis or assisting in high risk surgical procedures, future healthcare professionals will rely progressively more on these burgeoning technologies for positive patient outcomes.

Why Deep Learning for Healthcare?

Deep Learning is a sub branch of Machine Learning where neural networks are used to build models from large data sets. Hospitals are extremely data rich environments and DL loves to process large amounts of data. In previous decades, processing such large amounts of data using DL would have taken months or years and consumed multiple years of IT budgets. Now with the help of accelerated compute and dense storage platforms, those same processes can be done in weeks, days, or even hours for a fraction of the cost. So many more organizations can now take advantage of the advances in IT technology to deploy DL algorithms and neural networks. Let’s take a quick look at different types of neural networks and where they apply to the healthcare industry.

Neural Networks Impacting Healthcare

The first type of neural network impacting the healthcare industry is a Convolutional Neural Network (CNN). In the world of neural networks, CNNs are widely used for image classification. Recently the FDA approved AI for use in chest x-ray detection for Pneumothorax, a condition that occurs when gas accumulates in the space between the chest walls and lungs. If undetected, it can lead to lung collapse or become fatal. Pneumothorax can be often overlooked, as it is hard to detect at first glance. Now, with the use of AI, the image can be flagged for a deeper look by doctors, which leads to easier detection and better outcomes for the patients. Notice here that the image is simply flagged and then still must be reviewed by medical staff. This is an AI augmentation use case and not a replacement for hands-on medical care.

Another workload seeing the benefits of AI on image analysis is Digital Pathology. This practice allows pathologists to digitize whole slide images allowing for AI algorithms to be run against these images. This can accelerate time to diagnosis leading to better and faster patient care.

The second type of neural network is a Recurrent Neural Network (RNN) where the sequence of the data matters, such as in verbal communication. Natural Language Processing (NLP) is a common technique used in RNNs to build voice recognizing applications. If you’ve ever talked into a virtual assistant like Siri or Alexa, you have used an RNN. The Healthcare industry is being completely transformed using NLP and voice recognition applications.

For instance, a couple weeks ago I was in the doctor’s office and he was using a voice recorder to record our session for his notes. He explained that he tried using tablets to jot down consultation notes, but found himself staring at the tablet instead of patients. In the end it was easier to record the meetings then have the notes transcribed.  Short-term automation through AI will help with dictation and transcription via the use of virtual assistants. Doctor’s notes will be captured and transcribed in near real-time. The impact will be better care and more face time for doctors to be in front of their patients instead of behind a keyboard or desk.

The last neural network being implemented in the healthcare industry is the Generative Neural Network (GAN). A GAN is actually two neural networks: one is a generator that creates fake data and the second is a discriminator which attempts to tell if the data is real or fake. The process pitting the generator and discriminator against each other help build better outcomes for the models. Deep fakes are a common example of GANs. While deep fakes may pose threats, there are some good use cases for GANs in Healthcare.

Drug discovery in healthcare is a long and costly process. Most drugs never make it out of the research phase let alone get FDA approval. GANs are being used now to speed along the discovery phase of approval process. Researchers can generate a list of known elements for use in a GAN to build out millions of different possibilities for element combination that will be the next to treat breast cancer, prostate cancer, or other diseases. The use of GANs in drug discovery offers a ton of upside and is something that the Dell Technologies Healthcare IT teams will monitor closely. These three neural networks showcase the immense potential of AI and Deep Learning in Healthcare; and this is just the beginning.

Start Your AI Healthcare Journey 

The science behind these Healthcare advances can be difficult to understand however architecting the right IT Infrastructure for your AI initiatives doesn’t need to be as challenging. At Dell Technologies we have been helping customers to unlock the value in their data capital with the right technology to suit their needs and use cases. To learn more about how we can assist on your AI Journey in Healthcare, Life Sciences or any other enterprise click the link below:

About the Author: Thomas Henson

Thomas Henson an Unstructured Data Solutions Systems Engineer with a passion for Streaming Analytics, Internet of Things, and Machine Learning at Dell Technologies. He brings experience in Machine Learning Anomaly Detection, Open Source Data Analytics Frameworks, and Simulation Analysis. Thomas is also heavily involved in the Data Analytics community.