Deep Learning: Types and Applications in Healthcare | Technology and Health

Deep Learning: Types and Applications in Healthcare | Technology and Health
Deep Learning: Types and Applications in Healthcare

Deep learning is a growing trend in healthcare artificial intelligence, but what are the use cases for the different types of deep learning?

Deep learning (DL), also known as deep structured learning or hierarchical learning, is a subset of machine learning.  It is loosely based on the way neurons connect to each other to process information in animal brains.  To mimic these connections, DL uses a layered algorithmic architecture known as artificial neural networks (ANNs) to analyze the data.  By analyzing how data is filtered through the layers of the ANN and how the layers interact with each other, a DL algorithm can 'learn' to make correlations and connections in the data.

These capabilities make DL algorithms an innovative tool with the potential to transform healthcare.  The most common types found in industry have different use cases.

READ MORE: The Future of Healthy Eating: The artificial intelligence nutritionist is here | Health and Technology

Deep Neural Network

Deep Neural Network (DNN) is a type of ANN, but it is classified as 'Deep' because it has greater depth than some other neural networks.  These layers perform mathematical translation functions that allow the raw data to be translated into meaningful outputs.  Additional layers allow slightly different translations in each layer with the intention of further refinement of the output.

A recent example from a DNN use case shows how these algorithms can be used to predict the no-shows of a pediatric appointment.

In a study published in NPJ Digital Medicine, researchers hypothesized that patient EHR data and local weather information could be used as predictions for pediatric no-shows and that a prediction model using this information could be used by providers.  can help to implement no-show prevention measures.

The researchers developed their model by retrospectively collecting EHR data for 19,450 patients between January 10, 2015 and September 9, 2016, at Boston Children's Hospital's primary care pediatric clinic.  Of the 161,822 appointments involving these patients, 20.3 percent did not show up.  Local weather information on the day of appointment was also included.

DNN outperformed the traditional no-show prediction approach, and showed that atmospheric pressure and patients' history of no-show records were the most important predictors of whether patients would show up for their next appointment.

Using a DNN, researchers could more easily evaluate the relationship between various factors related to a patient's no-show and determine which factors were most correlated with outcome.

READ MORE: Deep Learning: An in-depth look at AI-powered Technology | Technology

Conventional Neural Networks

Convolutional Neural Network (CNN) is a type of DNN used to understand visual data.  CNNs analyze images and extract features they can use to classify images into categories.

Classification is important in areas such as medical imaging, where a doctor will look at an image, such as a CT scan or X-ray, to diagnose various conditions.  Applying the CNN algorithm to aid medical imaging tasks has the potential to improve clinical decision support and address issues related to population health.

In a study that examined the relationship between retinal disease and lack of access to care, researchers sought to develop a CNN model that could detect multiple retinal diseases simultaneously.

They developed their model using 120,002 ocular fundus images that were evaluated by a group of certified ophthalmologists and labeled according to retinal disease diagnosis.  To evaluate the model's ability to accurately detect retinal disease, its performance was compared to that of a group of retinal specialists.

CNN achieved accuracy greater than or equal to that of the expert group in seven of the 10 retinal diseases being evaluated.  The model also outperformed physicians when comparing image evaluation speed, analyzing an image in less than a second while the fastest specialist took 7.68 seconds with the same image.

The researchers said the success of their model could help improve access to care in underdeveloped areas where low screening rates and late diagnosis of retinal disease are associated with an increased risk of irreversible vision loss.

READ MORE: Artificial Intelligence in the Future of Sports | Technology and Sports

Recurrent Neural Network

Recurrent Neural Networks (RNNs) are another type of ANN that use sequential or temporal data.  They are often used for problems related to language translation, natural language processing, speech recognition, and image captioning.  Unlike other neural networks, where inputs and outputs are independent of each other, RNNs take information from inputs in previous layers to influence the current inputs and outputs.

RNNs are useful for healthcare providers to assist with tasks such as clinical trial group selection.

In clinical trials, a group or group of patients sharing similar relevant characteristics are selected to participate in the research.  Because the success of a clinical trial depends on the precise selection of a patient group, the process can be time-consuming and costly.

The researchers, trying to minimize time and cost, set out to test whether different DL models could successfully identify key features for group selection.  They trained and tested a simple CNN, a deep CNN, an RNN, and a hybrid model that combines both a CNN and an RNN.  Patients' records to the model were manually labeled by experts to indicate whether patients met one or more of the 13 possible criteria for a clinical trial.

Overall, the RNN and hybrid models outperformed the CNN model.  However, the researchers noted that their study was limited by the small dataset they used and indicated that further studies would be needed before a group selection model could be implemented.

READ MORE: Mental Health leads the list of the most significant business concerns for 2022 | Health

Generative Adversarial Network

Generative Adversarial Networks (GANs) use two neural networks to generate synthetic data that can be used in place of real data.  GAN is commonly used in image, video and voice generation.

GANs have great potential for use in healthcare due to their ability to generate synthetic MRI images.  Using medical images to train AI models for diagnosis and predictive analysis poses several challenges to researchers because their quality can vary, they can be subject to patient privacy regulations, and image datasets are often imbalanced.  There are.

To address these concerns, researchers have explored using MRI images generated by GANs to train deep-learning models for clinical decision support.  In one study, researchers trained a GAN to produce abnormal MRI images using publicly available scans from the Alzheimer's Disease Neuroimaging Initiative (ADNI) and the Multimodal Brain Tumor Image Segmentation Benchmark (BRATS).

GAN-constructed MRIs had some features that allowed researchers to differentiate them from real MRIs, but the team pointed out that a larger training dataset could eliminate that problem in the future.

Synthetic medical images will potentially allow researchers to create larger datasets with higher-quality images and a more balanced distribution of pathological findings.

In addition to improving the imaging data used to train other AIs, GaN-generated medical images can help protect patient privacy by reducing the need to use real medical images for research.

Source: Health IT Analytics, Direct News 99