A 19th-century physician would be quite shocked by today’s changes in medicine. Back in that time, the doctor noted the symptoms, diagnosis, and treatment of the patient by relying on personal experience, professional literature, and pen and paper. With the advent of computers in 1980, the EHR (electronic health record) became a frequent practice. However, only large hospitals had computers at that time. Only 11 years later, when the Internet appeared, the US Institute of Medicine recommended every physician use computers by the year 2000 to improve patient treatment.
The availability and spread of computers enabled practical work on the creation of artificial intelligence. Thanks to advances in computer science and informatics, artificial intelligence (AI) is quickly becoming an integral part of modern healthcare. Currently, AI offers absolutely new cutting-edge solutions in the field of medicine. Remember The Hitchhiker’s Guide to the Galaxy by Douglas Adams? After seven and a half years of work, the “deep thought” computer on “the ultimate question of life, the universe, and everything” came up with the answer—42. Smart algorithms can give answers to questions without explanation. Data is the only thing you need.
The health care sector annually generates a huge amount of data resulting from patients’ Electronic Health Records (EHRS), administrative data, surveys and reviews, clinical trial data, etc. According to the OECD, currently, we face an increase in the amount of collected information including clinical, behavioral, genetic, financial, operational, etc. data. There is no way to stop data growth. It will keep on growing at an incredible rate. It is next to impossible to process all these huge amounts of data and accomplish many tasks in healthcare without the aid of artificial intelligence.
The aim of this article is to summarize recent developments of AI in medicine, provide the main use-cases where AI-powered medical technologies can already be used in data management, patient-management, clinical practice and etc.; show benefits of using artificial intelligence in medicine and so on.
Artificial Intelligence in Medicine
In medicine, artificial intelligence works with both structured data and unstructured data. For instance, ECG data is structured data, while a prescription or patient record is unstructured. In the first case, the data immediately enters the neural network for analysis. In the second case, unstructured data is first brought to a form that can be analyzed using natural language processing (NLP) methods.
The Accenture report states that by 2026 AI applications in healthcare can potentially comprise up to $150 billion in annual savings in the US.
Allied Market Research reports that the value of AI in healthcare comprises $8.23 billion in 2020 and is forecasted to reach $194.4 billion by 2030.
The classic medicine and healthcare tasks intelligent algorithms can address include:
- Improvement of patient care and treatment practices by evaluating outcomes based on medical literature and doctors’ recommendations.
- Assist insurance companies to manage preventive care (annual checkups, flu shots, specific tests, and screenings) for large population groups by analyzing patient datasets.
- Improvement of physicians’ time management and pharmaceutical research coverage.
- Simplification of administrative processes in hospitals: aid patient management and electronic health record systems.
The number of businesses offering intelligent algorithms for medical purposes is increasing year by year. Not only business people but also politicians and regulatory bodies recognized the huge potential of AI to streamline processes in medicine, pharmaceuticals, or general administration. This is evidenced by the high growth in the number of algorithms approved by the FDA—agency of the U.S. federal government authorized by Congress to inspect, test, approve, and set safety standards for foods, drugs, chemicals, cosmetics, and medical devices. In 2014 AliveCor was the only approved algorithm. This algorithm was related to cardiology. It evaluated heart rate data and informed patients about potential heart rhythm disorders. Two years later 23 algorithms were approved by the U.S. Food and Drug Administration. This is a growing trend and dozens of new AI algorithms appear in various fields of medicine every year.
AI in Medicine: App and Project Ideas
There are plenty of ways in which artificial intelligence can positively impact processes in medicine, whether its research acceleration or helping doctors make better-informed decisions. Below are some examples of how artificial intelligence might be applied.
Health Data Management is a systematic organization of medical data. The wave of digitization in healthcare made data management more challenging for hospitals and healthcare institutions. New patient information is daily collected from various sources such as electronic medical records (EMR) and other internal databases across institutions.
Apart from bringing data together, AI simplifies the search, filtering, and analysis of information. This helps to collect more complete and accurate information about patients, and thus provide better medical treatment.
AI use cases:
For instance, the Google Deep Mind Health project is used to collect data from medical records in order to provide better and faster medical services.
In the Netherlands, almost all hospitals store medical invoices in digital form. These invoices include information about the treatment, the individual physician, and the hospital. At Zorgprisma Publiek they analyze invoices and leverage IBM Watson to collect the data. Such analysis helps to determine whether a doctor or a hospital is making repeated treatment mistakes.
Benefits of AI in data managemnet:
- Higher personal health data security.
- Better and faster provision of medical services.
- Higher patient satisfaction indices.
Clinical pathway design
Clinical pathway (CP) is a way to provide medical treatment for a patient with a specific diagnosis. The clinical pathway aims at predicting the course of the disease: it predicts what, when, and to whom it should happen.
CP enables understanding of how the treatment affects the patient. The collection of data on “deviations” in each individual patient is an essential part of CP development. AI analyzes the deviations to identify peculiarities that can be further considered to improve the next clinical path.
AI use cases:
IBM Watson launched an oncologists-oriented program capable of providing treatment options. This program is able to quickly processes patient data, medical literature, recommendations from recognized scientists, and the experience of qualified doctors. Based on this data, IBM Watson provides personalized patient care recommendations to be reviewed and used by oncologists.
Hanover is another great example of machine learning research by Microsoft, aiming at the revision of all available documents, and assistance in predicting what medicines and combinations of preparations would be most effective in cancer treatment.
Benefits of AI in clinical trials:
- Medical treatment quality improvement.
- Results optimization: reduced duration of hospitalization, mortality, and readmission rates.
- Cost reduction up to 30%. Annual unreasonable healthcare expenses in the USA comprise $1 trillion. The bigger part of waste results from differences in clinical practice, inappropriate treatment leading to death, that could be avoided through better treatment recommendations.
Computer vision, being one of the most advanced areas of ANI (Automatic Number Identifier), has a significant impact on diagnostics, leading to a revolution in medical visualization. Intelligent algorithms can analyze MRIs, CT scans, X-rays, and any other medical images.
AI use cases:
Scientists from the University of Adelaide were running an artificial intelligence system-based experiments to predict if a person is going to die in the near future. The deep learning system analyzed more than 16,000 images showing the impact of pathologies in various organs. The goal was for the algorithm to learn how to measure overall health conditions rather than detecting a single disease. By analyzing CT scans of 48 patients, deep learning algorithms were able to predict whether patients would die within the next five years at 69% accuracy.
The IBM Watson platform leverages medical imaging as well. IBM enables dermatologists to use Watson’s results to diagnose melanoma and other skin cancers faster and more accurately. This approach doesn’t require a large number of biopsies. Thus the experts found that their deep learning system was able to achieve 76 percent accuracy in diagnosing melanoma cases based on dermatological images. Apart from IBM, such giants as Philips, Agfa, and Siemens started artificial intelligence integration into their medical imaging software systems.
In 2016, Google developed the retinal scan method to detect diabetic retinopathy. This disease is quite common among diabetic patients. If not detected at an early stage, it can lead to blindness. The machine-learning algorithm uses Google’s method to label millions of web images. The next step is to examine photographs of the patient’s retinas to look for tiny aneurysms being indicators of diabetic retinopathy early stages. A year later, Google announced that they started working on integrating the technology into a network of eye clinics in India. Currently, more research is being done to step beyond diabetic retinopathy. For instance, predicting risk factors of cardiovascular pathologies by examining retina images.
Google was not the only one that worked on identifying this pathology. In 2016, 16-year-old Kavya Kopparapu created Eyeagnosis. The fact that her grandfather was diagnosed with DR motivated her to create it. They lived in a small Indian town with almost no ophthalmologists leading to diagnosis at the final stage when it is too late. Therefore she developed a smartphone app that can detect the disease using artificial intelligence program and a simple 3D-printed lens mount.
Benefits of AI in transforming diagnostics:
- Reduced medical errors and more accurate diagnosis.
- Development of individual treatment strategies.
Health assistance and administration
Primary medical care does not always require doctors’ intervention. Sometimes prescriptions or some guidance should be good enough. Administrative, organizational, and paperwork take much of doctors’ time that could be spent on treatment.
Doctors spend around 16 minutes per patient just to fill out the EHR forms. The good news is that with AI-powered workflow optimization, doctors can stop worrying about administrative tasks and devote their time to patients.
The Brookings Institution estimates that 40% of tasks performed by ancillary personnel and 33% of tasks performed by medical practitioners can be automated.
AI use cases:
For instance, Augmedix from San Francisco leverages the power of Google Glass to make healthcare more patient-centric and reduce paperwork. Google Glass is equipped with a camera, speaker, and microphone. The doctor wearing these glasses can identify the patient using facial recognition software. After recognition, the patient’s electronic medical record and the history of the patient’s last doctor’s visit will be shown on the screen. And this is just one of the scenarios in which this device could be used.
Technologies converting voice to text is an alternative to handwriting. However, voice recognition solutions do not completely eliminate transcription errors, resulting in extra time for medical professionals spent on verification. Even one letter, recognized differently, can endanger a patient’s life. Perhaps a voice recognition system with an AI-based proofreading system should be considered? This can finally solve the problem of medical administration for you.
Benefits of AI in health assistance and administration:
- Time-saving up to 30%.
- Improvement of medicare quality since doctors can focus solely on their patients.
- Patients with critical conditions receive timely medical care through request prioritization and automatic listing of patients.
- Reductions in operational costs due to fewer human resources required to organize the process.
Innovative technologies in patient care enable patients to take care of their illnesses themselves.
AI use cases:
OrCam, BeMyEyes, and Aira offer solutions for the visually impaired and enable them to live more independent lives. They use various algorithms to describe the environment to the user, read text, recognize faces and objects, and notify of obstacles.
Florence is a practical chatbot for elderly patients. Practically, this is a “personal nurse” who reminds patients to take their pills.
Some institutions already recognize the potential of AI-based chatbots for their patients. The UK National Health Service (NHS) uses a chatbot application to provide medical advice in order to reduce the load on the “111” helpline. The NHS developed the app in partnership with Babylon Health. Users report their disease symptoms to the app, which checks them against a database of diseases. The chatbot suggests an appropriate action plan based on the data received.
Benefits of AI in patient management:
- 24/7 access to medical support.
- Round-the-clock monitoring of the patient’s condition.
- The ability to provide quick answers about illnesses and medicines.
- Monitoring adherence to the required medication regimen.
- Reduction of ambulance personnel workload
The field of precision medicine is also experiencing rapid growth. The National Institutes of Health (NIH) defined precision medicine as “an emerging approach for disease treatment and prevention that takes into account individual variability in genes, environment, and lifestyle for each person”.
Currently, genotype-based treatment is the most well-studied impact of precision medicine on healthcare. Researchers have already benefited from the development of programs that provide specific assistance.
AI use cases:
For example, Deep Genomics is inventing a new generation of computing technologies that can tell doctors what will happen inside a cell when DNA is altered by genetic changes. The goal is to develop genetically based pathology treatments for diseases that are currently deemed untreatable.
Cambridge Cancer Genomics develops oncology solutions to improve the treatment of patients suffering from cancer. They believe that the more clinical and genomic data oncologists have, the smarter decisions on medicines application they can make.
Benefits of AI in precision medicine:
- Development of individual prevention and treatment strategies.
- Efficient drug development
Artificial intelligence aids the pharmaceutical industry to revolutionize research and development. This is especially true in the early stages of drug development.
New drugs are approved through human trials: a procedure that begins with animal testing and gradually moves to human patients. Bringing a drug to market is a time-consuming process. Typically the research costs billions of dollars and takes many years to complete. On top of that, patients involved in trials are exposed to side effects that cannot be fully predicted. And even if the study is successful, it must be approved by regulatory authorities such as the Food and Drugs Administration (FDA). Despite time and money investment, only 1 in 10 medicines will be approved by the FDA.
The introduction of artificial intelligence helps to predict drug toxicity at the very beginning of its creation. By using AI to study compounds’ molecular behavior/compound interactions, AI can spot patterns faster. Moreover, the use of AI during animal testing can predict drug behavior in a faster and more successful way.
AI use cases:
The number of existing AI drug companies reflects the diversity of the technology application patterns: there are many already and they keep evolving.
For instance, Insilico Medicine, working with researchers at the University of Toronto, got on the news by announcing that it takes just 46 days to develop a new drug leveraging AI. The system of analysis is similar to the work of a laboratory chemist who is looking for new treatments—just way faster.
Benefits of AI in drug development:
- Saving time and money by avoiding working on compounds that are unlikely to be effective.
- Generating ideas for completely new compounds leads to the acceleration of the new effective drugs discovery.
- Save hundreds of man-hours of lab work. Eliminates the need for repetitive tasks such as analyzing thousands of histological images.
- Minimizing animal testing of new drugs.
FDA-approved AI-based algorithms in healthcare
AI can have its drawbacks, creating a greater risk of misdiagnosing a patient and leading to inappropriate treatment as a result. Therefore, every AI-based technology in healthcare must be regulated, effective, and evidence-based.
When it comes to artificial intelligence in healthcare, the FDA is one of the most reliable indicators in medicine. And also the only criteria for reliable and accurate medical software.
In this module, we will touch on algorithms that have already proven themselves efficient for clinical use. A database with all FDA-approved smart algorithms doesn’t exist so far. We’ve compiled the latest FDA-approved data in AI-assisted medicine into one infographic. A brief data survey was based on several sources: U.S. Food and Drug Administration and The Medical Futurist.
It is obvious that AI algorithms are leveraged in various medical fields. Radiology is the most common area for AI-enabled devices, accounting for 70% of solutions, followed by cardiology with 12% of solutions. Good reasons are standing behind these figures. First of all, computer vision is one of the fastest-growing areas in artificial intelligence development, and medical imaging has both data and visibility. This directly affects the development of intelligent algorithms.
The art of medicine begins with the era of Artificial Intelligence
In the coming years, artificial intelligence will undoubtedly transform medicine. It will find new medicines and treatment approaches. We have covered the benefits of artificial intelligence in medicine, however, in general, the integration of medical artificial intelligence into your project can:
- Improve the quality of patient care: artificial intelligence can provide better patient care by revealing pathologies at an early stage and suggesting more effective treatments;
- Document and provide more information on patient’s condition and help physicians make better decisions;
- Save time and money on administrative tasks: artificial intelligence in medicine can perform administrative tasks itself;
- Help provide 24/7 support using chatbots that can answer most frequently asked questions;
- Provide personalized recommendations on patient treatment;
- Predict drug toxicity at the inception, including human exposure and side effects well as facilitate new drug testing.
In the global trend of continuously evolving healthcare services, especially during and after the pandemic, data access and use to enable faster and better decision-making comes out on top. Thus, by bringing together different sets of patient data in one place and making it easier for professionals to interpret the collected information, software not only aids healthcare organizations to survive but also helps them and their patients thrive.