Artificial Intelligence in Healthcare: A Primer

There is a lot of buzz surrounding Artificial Intelligence (AI) nowadays. This article will provide a brief description about AI in healthcare.

Background Information:

The term ‘Artificial intelligence’ was coined by John McCarthy in the year 1956 at Dartmouth College at the first-ever AI conference.

Artificial Intelligence (AI): refers to systems or machines that mimic human intelligence to perform tasks and can iteratively improve themselves based on the information they collect.

Put simply, it is the ability of a computer program to learn and think.

Machine Learning: A subfield of artificial intelligence, which is broadly defined as the capability of a machine to imitate intelligent human behavior.

Put simply, machine learning is the field of study that gives computers the ability to learn without explicitly being programmed.

Types of AI:

  • Reactive Machines: Deal with simple classification and pattern recognition tasks, and are useful when all parameters are known. These cannot handle imperfect information.
  • Limited Memory: Can handle complex classification tasks by using historical data to make predictions. This is the current state of AI.
  • Theory of Mind: Understands human reasoning and motives, so needs fewer examples to learn because it understands motives.
  • Self-Awareness: Refers to AI with Human-level intelligence that can even by-pass human intelligence. Here, AI has a sense of self-consciousness. At present, this does not exist.

Artificial Intelligence (AI) in Healthcare:

Over time, several AI systems have been developed in the medical field.

These include AI systems that automate processes or procedures, those that assist healthcare professionals, or even operate in the absence of healthcare professionals.

I will mention some examples of AI systems here to illustrate the above.

Disclaimer: Mentioning any Trademark name does not imply that I am endorsing the same. All trademarks belong to their respective owners, and I am not making any claim on trademark. Readers are advised to make independent enquiries and consult a qualified professional before making any financial or health related decisions.

Dermatology: An AI system ‘DeepDerm’ uses deep convolutional neural networks to analyse photographs of skin lesions and determine the condition.

Diabetic retinopathy: EYEART is used for screening of retina for diabetic retinopathy.

Radiology: AI systems have been developed for Xray, CT, and MRI (Siemens Healthineers). MIMIC-CXR (large scale automated reading of frontal and lateral CXR using Dual convoluted neural networks) and QureAI (automated reading of CXR for tuberculosis and other conditions) are well-known. QureAI has been recognized by the World Health Organization (WHO) for screening of Tuberculosis.

Chexnext (Deep learning for chest radiograph diagnosis)

Mammography: AI system for breast cancer screening has been developed that exceeds the performance of radiologists.

Gastro endoscopy: Virtual endoscopy (3D images from CT/ MRI are used to simulate endoscopy virtually). Capsule endoscopy (wireless) uses a capsule to perform endoscopy in patients. The capsule contains a light source, camera, and transmitter. Live feed from the camera may be obtained. Alternately, the video thus obtained may be analysed by an AI system to detect polyps. GastricNet is an AI system for detection of Gastric cancer. Analysis of endoscopy videos using Deep learning is on the rise.

Pathology: Machine learning for digital histopathology involves creation of 3D images from a tissue section. These 3D images are then analyzed using an AI system. This is more accurate since a greater tissue sample is utilized for the process than conventional histopathology.

Neurosurgery: AI is being used for brain tumour diagnosis. Optical imaging and AI are making brain tumor diagnosis quicker and more accurate.

Cardiology: AI systems for automatic detection and classification of coronary artery plaques and stenosis using Deep Learning have been developed.

Even in the field of medical education, AI systems have been developed for a variety of activities.

Resource management for students often requires complex decision-making. AI systems for this purpose have been developed, and automate the process.

Skill training: Simulation technology, chatbots in medical education, AI systems to teach communication skills, AI systems for Direct Observation of Procedural Skills (DOPS), virtual reality (VR), laparoscopy simulators (da Vinci), real time actual dynamic operators, AI guided surgical orientation system, simulator for GI endoscopy, VirtaMed Arthroscopy simulator, UroS (Urology simulator), OB/GYN simulator, etc. all use AI to help develop and assess skills.

AI in DOPS: Video analytics for DOPS- hand motion entropy and timing metrics discriminate levels of surgical skill. Synchronized video and motion analysis are now possible due to AI systems.

There is research showing that training on VR gives better training. AI and VR used in neurosurgery has been shown to have 97.6% accuracy in one study.

Machine learning to assess surgical expertise (MLASE) is designed to bridge the gap between specialties.

Surgical skill assessment using robotics and AI is already in place. Haptic feedback can be provided through robotics, enhancing the learning experience for learners, and generate key data for assessors.

AI systems in clinical diagnostic skills: i-Doctor (algorithm based diagnosis) has been combined with point of care tests to develop standalone kiosks to diagnose common conditions and dispense drugs in primary care settings. The system can be used to teach and evaluate diagnostic skills also.

AI system in communication skills: MPathic-VR virtual human simulation is being used to teach communication skills.

Key Considerations:

Although there has been considerable progress regarding the development and deployment of AI systems in healthcare, much of the progress has occurred in developed countries.

These technologies have considerable initial cost on account of technology and infrastructure requirements, so are unaffordable for most institutions and individuals. However, in some instances, government funded projects employing AI systems are helping bring such technology to the masses. This is particularly true of AI systems in primary health care including telemedicine.

Many of the AI systems mentioned in this article have received FDA approval. However, not all are commercially available at his point.

AI systems may assume greater significance in medical education in the coming years.

Useful Link:

Link to a talk by Prof. Dr. Arun Jamkar (Vice-Chancellor, Maharashtra University of Health Sciences) on Artificial Intelligence in Medical Education and Skill Training:

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.