Artificial intelligence (AI) is reshaping healthcare, and its use is becoming a reality in many medical fields and specialties. AI, machine learning (ML), natural language processing (NLP) and deep learning (DL) enable healthcare stakeholders and medical professionals to identify healthcare needs and solutions faster with more accuracy, using data patterns to make informed medical or business decisions quickly.
AI is able to analyze large amounts of data stored by healthcare organizations in the form of images, clinical research trials and medical claims, and can identify patterns and insights often undetectable by manual human skill sets.
AI algorithms are “taught” to identify and label data patterns, while NLP allows these algorithms to isolate relevant data. With DL, the data is analyzed and interpreted with the help of extended knowledge by computers. The impact of these tools is huge, considering a Frost & Sullivan analysis indicated artificial intelligence and cognitive computing systems in healthcare will account for $6.7 billion this year from the market compared to $811 million in 2015.
The use of AI is supporting many stakeholders in healthcare:
AI is used as a tool for case triage. It supports a clinician reviewing images and scans. This enables radiologists or cardiologists to identify essential insights for prioritizing critical cases, to avoid potential errors in reading electronic health records (EHRs) and to establish more precise diagnoses.
A clinical study can result in huge amounts of data and images that need to be checked. AI algorithms can analyze these datasets at high speed and compare them to other studies in order to identify patterns and out-of-sight interconnections. The process enables medical imaging professionals to track crucial information quickly.
For example, Hardin Memorial Health (HMH) needed to find a way to extract relevant data from EHRs in a concentrated form for imaging professionals. The hospital’s Emergency Room (ER) was handling more than 70,000 patients per year and decided to partner with IBM to implement “The Patient Synopsis”. This product identifies patient information relevant to the imaging procedure conducted on that patient.
Patient Synopsis digs into past diagnostics and medical procedures, lab results, medical history and existing allergies, and delivers to radiologists and cardiologists a summary that focuses on the context for these images. The product can be integrated with any medical unit system structure, accessed from any communication workstation or device in the network, and upgraded without affecting the daily activity of the medical unit.
Detecting relevant issues and presenting them to radiologists in a friendly summary view enables the design of more customized, targeted and accurate report used in diagnostic decision process.
Supercomputers have been used to predict from databases of molecular structures which potential medicines would and would not be effective for various diseases. By using convolutional neural networks, a technology similar to the one that makes cars drive by themselves, AtomNet was able to predict the binding of small molecules to proteins by analyzing hints from millions of experimental measurements and thousands of protein structures.
This process enabled convolutional neural networks to identify a safe and effective drug candidate from the database searched, reducing the cost of developing medicine.
In 2015, during the West African Ebola virus outbreak, Atomwise partnered with IBM and the University of Toronto to screen the top compounds capable of binding to a glycoprotein that prevented Ebola virus penetration into cells in an in vivo (in the living body of an animal or plant) test. From the tested compounds, the one selected was chosen because it acted on other viruses with a similar mechanism of cell penetration. This AI analysis occurred in less than a day, a process that would have usually taken months or years, enabling the development of a treatment for the Ebola virus.
Clinicians often struggle to stay updated with the latest medical advances while providing quality patient-centered care due to huge amounts of health data and medical records. EHRs and biomedical data curated by medical units and medical professionals can be quickly scanned by ML technologies to provide prompt, reliable answers to clinicians.
In many cases, health data and medical records of patients are stored as complicated unstructured data, which makes it difficult to interpret and access. AI can seek, collect, store and standardize medical data regardless of the format, assisting repetitive tasks and supporting clinicians with fast, accurate, tailored treatment plans and medicine for their patients instead of being buried under the weight of searching, identifying, collecting and transcribing the solutions they need from piles of paper formatted EHRs.
AI algorithms are able to identify new drug applications, tracing their toxic potential as well as their mechanisms of action. This technology led to the foundation of a drug discovery platform that enables the company to repurpose existing drugs and bioactive compounds.
By combining the best elements of biology, data science and chemistry with automation and the latest AI advances, the founding company of this platform is able to generate around 80 terabytes of biological data that is processed by AI tools across 1.5 million experiments weekly.
The ML tools are created to draw insights from biological datasets that are too complex for human interpretation, decreasing the risk for human bias. Identifying new uses for known drugs is an appealing strategy for Big Pharma companies, since it is less expensive to repurpose and reposition existing drugs than to create them from scratch.
Acute kidney injury (AKI) can be difficult to detect by clinicians, but can cause patients to deteriorate very fast and become life-threatening. With an estimated 11% of deaths in hospitals following a failure to identify and treat patients, the early prediction and treatment of these cases can have a huge impact to reduce life-long treatment and the cost of kidney dialysis.
In 2019, the Department of Veterans Affairs (VA) and DeepMind Health created a ML tool that can predict AKI up to 48 hours in advance. The AI tool was able to identify more than 90% of acute AKI cases 48 hours earlier than with traditional care methods.
The partnership between VA and DeepMind Health continues. Its next target is to identify how this ML tool can be installed in medical units. A user-friendly platform is also targeted in order to support clinicians in their treatment decisions that would improve the quality of life for Veterans suffering from AKI.
During a sudden heart attack, the time between the 911 call to the ambulance arrival is crucial for recovery. For an increased chance of survival, emergency dispatchers must be able to recognize the symptoms of a cardiac arrest in order to take appropriate measures. AI can analyze both verbal and nonverbal clues in order to establish a diagnostic from a distance.
Corti is an AI tool that assists emergency medicine staff. By analyzing the voice of the caller, background noise and relevant data from medical history of the patient, Corti alerts emergency staff if it detects a heart attack. Like other ML technologies, Corti does not search for particular signals, but it trains itself by listening to many calls in order to detect crucial factors.
Based on this learning, Corti improves its model as an ongoing process. The technology Corti is equipped with can detect the difference between background noise, such as sirens, and clues from the caller, or the patient sounds in the background.
In Copenhagen, emergency dispatchers are able to identify a cardiac arrest based on the description provided by the caller around 73% of the time. But AI can do better. A small-scale study conducted in 2019 revealed that ML models were able to recognized cardiac arrest calls better than human dispatchers by using speech recognition software, ML and other background clues.
ML can play an essential role in supporting emergency medical staff. In the future medical units could use the technology to respond to emergency calls with automatic defibrillators equipped drones or with CPR-trained volunteers, which would increase the chances for survival in cases of cardiac arrest that take place in the community.
In some cases, radiation therapy can lack a digital database to collect and organize EHRs, which makes the research and treatment of cancer difficult. In order to assist clinicians to make informed decisions regarding radiation therapy for cancer patients, Oncora Medical delivered a platform that collects the relevant medical data of patients, evaluates the quality of care provided, optimizes treatments and provides thorough oncology outcomes, data and imaging.
Automatic generation of clinical notes integrated with EHRs led to a reduction of time spent by clinicians in managing patient documentation, which improves medical operations and health outcomes.
Turning EHRs into an AI-driven predictive tool allows clinicians to be more effective with their workflows, medical decisions and treatment plan. NLP and ML can read the entire medical history of a patient in real time, connect it with symptoms, chronic affections or an illness that affects other members of the family. They can turn the result into a predictive analytics tool that can catch and treat a disease before it becomes life-threatening.
In essence, chronic diseases can be predicted and their progression rate tracked. CloudMedX is a company that focuses on decoding unstructured data – data stored as notes (clinician notes, discharge summaries, diagnosis and hospitalization notes, etc.).
These notes are used alongside EHRs as a source to generate clinical insights for medical professionals, allowing for data-driven decisions to improve patient outcomes. CloudMedX solutions have already been applied in several high-risk diseases such as renal failure, pneumonia, congestive heart failure, hypertension, liver cancer, diabetes, orthopedic surgery and stroke, with the stated objective to lower costs for patients and clinicians by assisting in early and accurate diagnoses of patients.
AI is also used to help rapidly discover and develop medicine, with a high rate of success. Genetic diseases are favored by altered molecular phenotypes, such as protein binding. Predicting these alterations means predicting the likelihood of genetic diseases emerging. This is possible by collecting data on all identified compounds and on biomarkers relevant to certain clinical trials.
This data is processed, for example, by the AI system of Deep Genomics. The company designs proprietary AI and uses it to discover new methods to fix the consequences of genetic mutations, while developing customized therapies for people suffering from rare Mendelian and complex disease.
The company tests identified compounds in order to develop faster genetic medicine for conditions with high unmet need. The company’s experts are working on “Project Saturn,” a drug system based on AI molecular biology that assesses more than 69 billion oligonucleotide molecules in silico (conducted or produced by means of computer modeling or computer simulation) against 1 million target sites in order to monitor cell biology to unlock greater potential treatments and therapies.
The discovery and development of genetic medicine brings benefits to patients and clinicians by decreasing the costs associated with the treatment of rare diseases.
The AI and ML industry has the responsibility to design healthcare systems and tools that ensure fairness and equality are met, both in data science and in clinical studies, in order to deliver the best possible health outcomes. With more use of ML algorithms in various areas of medicine, the risk of health inequities can occur.
Those responsible for applying AI in healthcare must ensure AI algorithms are not only accurate, but objective and fair. Since many clinical trial guidelines and diagnostic tests take into account a patient’s race and ethnicity that a debate has arisen:
Is the selection of these factors evidence-based? Is race and ethnicity data more likely to solve or to increase universal health inequities? It is established that ML comprises a set of methods that enables computers to learn from the data they process. That means that, at least in principle, ML can provide unbiased predictions based only on the impartial analysis of the underlying data.
AI and ML algorithms can be educated to decrease or remove bias by promoting data transparency and diversity for reducing health inequities. Healthcare research in AI and ML has the potential to eliminate health-outcome differences based on race, ethnicity or gender.
AI adoption in healthcare continues to have challenges, such as lack of trust in the results delivered by an ML system and the need to meet specific requirements. However, the use of AI in health has already brought multiple benefits to healthcare stakeholders.
By improving workflows and operations, assisting medical and nonmedical staff with repetitive tasks, supporting users in finding faster answers to inquiries, and developing innovative treatments and therapies, patients, payers, researchers and clinicians can all benefit from the use of AI in healthcare.