Even though the healthcare industry is worth **$8 trillion **[1] only 20% of people have **access to quality healthcare. **We have an aging population in the world. There is a pressing need for value-based healthcare [2]. The healthcare industry is going towards a **data-driven approach. **The amount of personal health and population healthcare data that is available today is growing at a rapid speed. The healthcare industry has resources and access constraints. On one hand, there is a shortage of pathologists, radiologists, and other clinicians and on the other hand, the number of procedures and diseases such as cancer is increasing. The only way to solve it is through technology.

In this article, the focus is:

What are the different types of Healthcare data

What kind of Deep Learning(DL) techniques are used in healthcare and

How can we use that data and techniques to build Top 10 applications in Healthcare

Let us start with different types of data that we use to build real-world machine learning / deep learning models. Deep learning is a subset of machine learning. Handcrafting of features is not required in deep learning at the cost of providing more data to build the models.

There are 7 types of data namely, numerical, categorical, text, image, video, speech, and signals irrespective of the domain to build deep learning models. Table 1 summarizes the types of different data from the healthcare domain. Based on the type of data the pre-processing steps may differ. Then we summarize the deep learning-based healthcare applications for each type of healthcare data.

Table 1: Types of data and healthcare applications

Next, I summarize various types of deep learning techniques and how can we use those techniques in healthcare applications along with specific examples in table-2. The techniques range from simple Feed-Forward Networks, Convolutional Neural Networks (CNN) to Recurrent Neural Networks(RNN), and latest Attention Networks.

Table 2: Type of deep learning techniques and applications in healthcare

Another dimension to look for the deep learning applications in healthcare is based on the various stages of the healthcare system, which is summarized in table-3. Prevention is better than cure. DL plays an important role both in early stages as well as advanced stages in the healthcare system [3].

Table 3: Heath sequence and deep learning applications

With this background, let us look into Top 10 Deep Learning applications in Healthcare using various types of data described earlier:

1. Medical Imaging:

• Convolutional Neural Network (CNN) 2D/3D plays an important role in medical imaging [4]
• We formulate Classification, Object detection and Segmentation kind of problems in medical imaging using advances in CNN
• This involves the processing of a huge number of images, refine its understanding and interpretation of the information
• **_Transfer learning _**from AlexNet, GoogleNet helps to build many image classification problems
• The DL models are getting ported to the Computed Tomography (CT), Magnetic Resonance Imaging (MRI) boxes to identify the **_quality of the reconstructed image _**and check for any issues such as motion detection
• Real-time image reconstruction — Can do a better reconstruction of the images in CT. This can reduce patient radiation exposure.

2. Faster Diagnosis:

• Analyzing medical images/data can often be a difficult task and time-consuming process
• GPU-accelerated **_DL to automate analysis _**and increase the accuracy of diagnostician
• DL helps doctors **_to analyze the disease better _**and provide patients with the best treatment
• Can act as a second objective opinion

#personalized-treatment #deep-learning-in-health #dl-in-healthcare #medical-imaging #ai-for-doctors #deep learning