“We are a society that waits to be sick before we take action. We don’t practice healthcare, we practice sick care, and that should change.”
Daniel Kraft

Healthcare – What is it?

From the definitional standpoint, healthcare is the organized provision of medical care to individuals or communities through “prevention, diagnosis, treatment, recovery, or cure of disease, illness, injury, and other physical and mental impairments” (Source: Wikipedia)

A solid healthcare system implementing continuous improvement in the health of citizens leads to a strong economy. The famous Iron Triangle of Healthcare consists of Quality, Access, and Cost, three competing priorities of healthcare. Best access to the healthcare system of the highest quality at the least cost is the idealistic goal of any healthcare system.

What’s not going well?

“Medicine itself is at a moment of crisis.” – Abraham Verghese, MD, Department of Medicine, Stanford University, wrote in the foreword of the book “Deep Medicine” by Eric Topol, mentioning that the current medical community fails to live the truth in what Hippocrates (460 BC – 370 BC) once said, “It’s far more important to know what sort of person has the disease than what sort of disease the person has.”

The medical community is over-burdened all across the globe. Even a doctor’s eye contact with the patient is nowadays a rare occurrence. Healthcare cost is soaring. We sadly witness the deaths of medical practitioners during the ongoing fight with the pandemic.

Can AI help? Is a robot assistant to a doctor without the chance of getting infected with the deadly virus the need of the moment?

We are also aware of the deaths of patients because of misdiagnosis by physicians are not the things of the past. 

Considering the divided societies of the world and pervasive inequalities, the expectation from AI – Fair and Equal Treatment of all people.

How Technologists can collaborate with Medical Practitioners

There will never be completely AI-driven automated healthcare. Always there will be Human in the Loop delivery of AI-enabled explainable healthcare to the patients. The humans here are no other than human doctors, nurses, and the like.

The goals of the collaboration among Medical Practitioners and Technologists with the use of AI are listed below:

• Improve Patient Experience considering Iron Triangle of Healthcare  of Quality, Access, and Cost
• Improve Health 
• Reduce healthcare Cost
• Achieve Work-life balance of healthcare providers
• Predict Insight and Risk in all areas of healthcare
• Provide Clinical Decision Support System (CDSS) to healthcare providers
• Provide assistance with automation in Medical Imaging and Diagnostics
• Provide assistance with automation in Drug Discovery. Please refer to the blog How Computer Technology Helps Vaccine Development
• Provide tools for better Patient and Consumer Engagement
• Provide tools for better Clinician Engagement

Artificial Intelligence technologies have the capabilities of analyzing a patient, starting from the stage of genomic analysis to Electronic Healthcare Records analysis through Natural Language Processing and delivering low-cost accessible high-quality healthcare by clinician process automation and healthcare fraud management.

Deep Learning-based Medical Imaging using Imaging Studies – MRI, CT  Scans, Echocardiograms, angiograms, etc. has made huge headway. Virtual Reality/Augmented Reality is being used in Medical Education/training. AI-driven Decision Support/Patient Decision/Risk Prediction using Electronic Health Records/Electronic Medical records like Lab results, prescriptions, vital signs/details, patient encounters/visits, etc. are already in place and are being extensively used. Diagnostics with Data Analytics and Explainable AI are making inroads into the day-to-day life of doctors. Precision Medicine, Population Health, Digital Health, and Telehealth are becoming all-pervasive. Wearable Technology with IoMT (Internet of Medical Things) consists of Wearables like Smartwatches with the capability like ECG at Wrist. IoMT will become very efficient with the advent of faster 5G technologies. Robotic Surgery is being used on a case-to-case basis. There is no dearth of Virtual Assistants/Conversational Chatbots in the life of doctors.

How AI can impact Major Specialities in Medicine

Impacts of AI on some of the major medical specialties are briefly described below:

Radiology

It is the most impacted specialty with the progress of AI. Radiology is changing its working methodologies as it is more and more leveraging Computer Vision/Deep Learning-based Medical Imaging technology.

Examples of Radiology Imaging use cases impacted by AI:

• Chest Pathology
• Cardiovascular Diseases
• Breast Cancer
• Neurological Diseases

Oncology

This is an area where AI already has a big presence. IBM Watson is a big player with the use of AI here. There are a number of failures for IBM Watson in meeting its promises in Oncology. This paper Concordance Study Between IBM Watson for Oncology and Real Clinical Practice for Cervical Cancer Patients in China: A Retrospective Analysisanalyzes the performances of IBM Watson in the context of China.

Cardiology

Cardiology is an area of medicine that extensively uses imaging products like EKG  (tracing of the rhythms of heart), CT, MRI, echocardiogram (ultrasound study of the heart). Lots of clinical data is already present in repositories related to treatments and surgeries of cardiovascular diseases.

The possible use cases for AI in Cardiology are depicted in the following diagram (taken from the Journal of the American College of Cardiology with Open Access license for the paper Artificial Intelligence in Cardiology  by Johnson and others).

Gastroenterology

There have been recent interests in the use of AI in Gastroenterology as the Deep Learning-based Computer Vision and Medical Imaging Technologies can assist the decision-making processes of Gastroenterologists. For example, this paper Artificial intelligence in gastrointestinal endoscopy: The future is almost here analyses how AI can be utilized for Automatic colonic polyp detection, Optical biopsy, etc.

Pediatrics Critical care medicine

Research is being conducted extensively to protect human life during this critical phase with data-driven intelligent decisions. Intelligent NICU (Neonatal Intensive Care Unit) solution is able to improve care time, enable remote monitoring, identify the early onset of disease, and reduction in neonatal mortality. The details are available in this paper
iNICU – Integrated Neonatal Care Unit: Capturing Neonatal Journey in an Intelligent Data Way

Dermatology

The activities in which Dermatologist uses observational skills followed by analysis for diagnosis of the types of diseases have the potential to be automated with AI Image Analysis and Computer Vision technologies.  For example, this paper Man against machine: diagnostic performance of a deep learning convolutional neural network for dermoscopic melanoma recognition in comparison to 58 dermatologists compares the diagnostic performances of human doctors and machine and concludes that “Most dermatologists were outperformed by the CNN. Irrespective of any physicians’ experience, they may benefit from assistance by a CNN’s image classification.”

(CNN is Convolutional Neural Network that is a type of artificial neural network mainly used in image recognition and processing.)

Endocrinology

This specialty primarily deals with diabetes and there are four ways AI is expected to impact endocrinology, for example, automation of retina screening, Clinical Decision Support System, Predictive risk stratification of population, and Guidance to the patient providing self-management tools. This paper Artificial Intelligence and Machine Learning in Endocrinology and Metabolism: The Dawn of a New Era | Endocrinology (frontiersin.org)describes the research outcome of impacts of AI on Endocrinology.

Anesthesiology

Though the job of Anesthesiologists is called “99% boredom and 1% terror”, the activities of anesthesiologists are difficult to predict. Anesthesiologists get very limited time to know each individual patient. But there is hope around the researches like this: Prediction of Bispectral Index during Target-controlled Infusion of Propofol and Remifentanil: A Deep Learning Approach

IoMT – Internet of Medical Things

IoMT (Internet of Medical Things) is a sub-field under the broad scope of Artificial intelligence that deals with all wearables and sensors-driven medical devices helping monitoring, treatment and patient-doctor interactions, and so on.

Objectives of IoMT

• Decision-making using biometric sensors/wearables
• Fitness wearables-driven wellness
• Investigations based on brain sensors/neuro sensors
• Smart self-monitoring like ECG using wristwatch, body temperature monitoring using the smart thermometer app
• Infant monitoring using IoMT-aided child wearables
• Clinical Efficiency through automation

AI during Covid-19

There are research papers being published regarding the effectiveness of AI to help manage the ongoing fight with the pandemic.

For example, a recently published paper, Artificial Intelligence Applications for COVID-19 in Intensive Care and Emergency Settings: A Systematic Review is discussing prognostic or diagnostic AI predictive models, the prognostic utility of deep learning-based lung segmentation software, an AI-based system for resource optimization in the ICU.

Barriers to introducing AI in Healthcare

• Human Factors- Skepticism, Attitude of doctors, Resistance to Change, Biased Behavior, especially of surgeons with high switching costs of surgical processes/procedures, Emotion like greed, etc.
• Environmental/Business/Regulatory barriers- Countries/Hospitals/Legal Infrastructure are not ready to use AI-based treatment of patients.
• Technical Barrier with Explainability of AI-driven decisions- AI solutions/tools at times fail to justify the AI-based decisions with causal analysis and traceability of factors. 
• Lack of available data for automation- For Machine Learning to work, it needs huge historical data. Lack of existing data or the presence of corrupted data can completely spoil the promise of AI.

Ethics of AI in Medicine

There should be human judgment to validate AI-driven decisions before any execution or implementation. The medical community should consider AI as their support mechanism. All AI systems should be developed to support fairness and equality. Human doctors or nurses can evaluate the presence of bias in AI-led decisions with rigorous observations on a continuing basis.

Broadly, these potential ethical issues with AI can fall under the following areas:

• Bias in training data for AI Systems
• Unintended uses of the AI tools/techniques not serving the mankind
• Impact of the AI Systems on the physician-patient relationship, primarily from the money and payment standpoint

There should be established governance (for example, boards, committees, etc.) around the use of AI in Medicine, before any serious use of AI in any particular specialty. Some of the thoughts may be around the following terms:

• AI Lawyers- How is the Legal Infrastructure around AI in the country?
• AI Taxation- As AI can replace a number of Medical practitioners, what is the plan around supporting them with new skill development, etc?
• US Federal Drug Administration (FDA) review process for Self Learning AI Tools/Solutions – For AI Tools/Solutions with Self-Learning Algorithms, what will be the approval process of the tool/process of the new algorithms by the US FDA or Drug Administration Authority of the country?

Going Forward

The presence of Artificial Intelligence in Medicine is inescapable. Countries are getting ready. US Federal Drug Administration (FDA) has established an AI action plan early this year. The objective of this action plan is “a total product lifecycle-based regulatory framework for these technologies that would allow for modifications to be made from real-world learning and adaptation while ensuring that the safety and effectiveness of the software as a medical device are maintained”. There should also be a use case-driven approach to evaluate the measurable value of outcomes of AI usage before going for extensive implementation of AI in any particular Medical scenario.