Securities.io is not an investment adviser, and this does not constitute investment advice, financial advice, or trading advice. Securities.io does not recommend that any security should be bought, sold, or held by you. Conduct your own due diligence and consult a financial adviser before making any investment decisions.
AI's goal is to enable machines to mimic human thoughts and behavior, encompassing learning, reasoning, and predicting. At the heart of this endeavor is AI-based modeling, which serves as the foundation for creating automated and intelligent systems.
To build such models, a variety of techniques are utilized, which are crucial for creating smart systems applicable in various real-world sectors, including finance, healthcare, agriculture, cybersecurity, and more.
The development of these models often employs different types of analytics—descriptive, diagnostic, predictive, and prescriptive—based on the problem's nature and its solution requirements.
Today, our focus is on predictive analytics, which explores data to predict what will happen in the future. By utilizing huge amounts of data, the model interprets recognizable patterns, which facilitates accurate predictions and helps make informed decisions.
AI's Unparalleled Ability to Recognize Trends
AI has this unparalleled ability to recognize trends within a set of variables. This is because, unlike humans, the technology can process vast amounts of data quickly, then identify patterns and trends and identify what it all means.
It is through machine learning algorithms that an AI model does its job, which is predicting the output based on the provided data.
When analyzing a trend, usually human brain function-inspired neural networks are used due to their ability to learn and adapt. Other machine learning algorithms key to trend analysis are Support Vector Machines (SVM), which categorize data into distinct classes, and Random forests, which aggregate multiple decision trees for powerful predictive models. Bayesian network is another one that uses probabilistic graphical models to represent the statistical dependencies among a set of variables.
In healthcare, leveraging these machine learning algorithms to build AI-powered prediction models allows for better and more tailored individualized care and treatment plans.
For this, AI sifts through diverse healthcare data to present emerging trends in disease progression. This enables medical professionals to preemptively address health issues and adapt treatment strategies, reshaping patient care.
The Potential of AI's Ability to Detect Patterns in Healthcare
Now, let's look at the real usage of technology's ability to recognize trends within a set of variables.
AI Surveillance Tool to Detect Sepsis Before Symptom Onset
In a recent study, researchers found that they can successfully predict sepsis infection in high-risk patients and save lives by utilizing a novel AI algorithm.
Sepsis is a major cause of mortality and morbidity worldwide. Under this condition, the body's immune system responds extremely to an infection, causing damage to its own tissues and organs.
In the US, each year, at least 1.7 million adults develop sepsis, and about 350,000 die from this life-threatening blood infection. Meanwhile, an estimated 48.9 million people are affected every year worldwide, out of which approximately 11 million die.
However, researchers from the University of California San Diego School of Medicine are all set to change these figures by leveraging AI.
According to the study published in the online edition of Digital Medicine, researchers developed an AI algorithm called COMPOSER. It utilizes data in real-time to tell in advance if the patient has the condition before its clinical signs are manifested by working “silently and safely behind the scenes.”
The model continuously surveys every patient for any signs of sepsis, noted study co-author Gabriel Wardi, MD, chief of the Division of Critical Care in the UC San Diego (UCSD) Department of Emergency Medicine.
Early recognition of sepsis is critical as interventions have greater benefits when implemented earlier in the disease course.
The model was used at the emergency departments of UC San Diego Health, the region's only academic medical system, to identify patients at risk for sepsis infection and has reduced mortality by 17%.
The way it works is that after a patient checks in the emergency department, COMPOSER starts to monitor over 150 different patient variables that could be linked to sepsis. These variables include demographics, medical history, vital signs, and lab results.
If a patient is found to have multiple variables, resulting in a high risk for sepsis infection, the algorithm notifies the nursing staff via the hospital's electronic health record. The staff then reviews with the physician for confirmation and to determine appropriate treatment plans for the patient.
“These advanced AI algorithms can detect patterns that are not initially obvious to the human eye.”
– Shamim Nemati, Ph.D., The study co-author & the director of predictive analytics and associate professor of biomedical informatics at UC San Diego School of Medicine
He also added:
“The system can look at these risk factors and come up with a highly accurate prediction of sepsis.”
The study, which was funded by the National Institutes of Health, the National Library of Medicine, and the Joan and Irwin Jacobs Center for Health Innovation at UC San Diego Health, examined over 6,200 patient admissions prior to and after the AI algorithm was deployed in the Hospital's emergency departments.
It is the first study that utilized an AI deep-learning model and reported improvement in patient outcomes. The model uses artificial neural networks to identify complex and multiple risk factors in patients accurately and safely. Wardi said:
“It is because of this AI model that our teams can provide life-saving therapy for patients quicker.”
COMPOSER was first activated over a year ago, in Dec. 2022. It is currently being utilized in many hospital in-patient units throughout UC San Diego Health, and in the future, UC San Diego Health East Campus will also start using it.
Recently, the Health Care Center announced its inaugural chief health AI officer. It has also launched a pilot where cloud-based electronic health record systems Epic and ChatGPT automatically draft more compassionate message responses so that doctors and caregivers can focus on patient care.
“Integration of AI technology in the electronic health record is helping to deliver on the promise of digital health.”
– Christopher Longhurst, the study co-author & executive director of the Jacobs Center for Health Innovation and chief medical officer and chief digital officer at UC San Diego Health
Software to Analyze Genomic Data to Predict Disease With Accuracy
Yet another example of AI-powered prediction models helping the healthcare sector can be seen in a study published last week. Created at Rutgers Health, the first-of-its-kind software called IntelliGenes combines AI and machine learning to help predict diseases in individuals. The model examines genomic data to find health-related biomarkers in order to produce personalized patient predictions as well as a visual representation of them.
Multi-genomic data informs about the inherent genetic makeup of a patient and has the potential to uncover new biomarkers based on their risk of disease. The study noted:
“We believe that synergistic use of multiple AI algorithms provides more accurate results, draws insightful conclusions, and precise predictions about real-world problems compared to a single AI algorithm on its own.”
It further found that the Support Vector Machine (SVM) and Random Forest (RF) are the most successful machine learning algorithms used to make predictions with high precision and resolve the problems of regression and classification.
Currently, there are no AI tools available to investigate and interpret the complete human genome, particularly for individuals who are not experts in the field. As such, the researchers developed IntelliGenes, which can be used by anyone, including those who do not possess much knowledge of bioinformatics techniques or have access to high-performing computers.
Using a combination of machine learning algorithms, Rutgers Health researchers aim to uncover information that usually goes undetected by classical statistics and traditional bioinformatics techniques.
The researchers then applied IntelliGenes in another study to discover novel biomarkers and predict cardiovascular disease (CVDs) with high accuracy.
The proper use of AI and ML methodologies, as per the study, can provide us with new insights and understandings of cardiovascular diseases. This, in turn, will enable better personalized treatments.
The study started with a vigorous pre-processing of gene expression data and then made use of three statistical tests to evaluate the differences in transcriptomic (study of the complete set of RNA) expression and the difference between clinical characteristics of CVD patients and that of healthy individuals.
In its research, the team uncovered 18 transcriptomic biomarkers that are highly significant in the CVD population. They were then used to predict disease with up to 96% accuracy. These results were also cross-validated with clinical records collected from patients in the cohort.
“There is huge potential in the convergence of datasets and the staggering developments in artificial intelligence and machine learning.”
– Zeeshan Ahmed, lead author of the study. Ahmed is a faculty member at Rutgers Institute for Health, Health Care Policy and Aging Research (IFH) and an assistant professor of medicine at Robert Wood Johnson Medical School
According to him, the model not only supports personalized early detection of both common and rare diseases in individuals but also opens the gates for broader research that'll ultimately lead to novel interventions and treatments.
AI is Not Yet a Perfect Solution!
As we saw, AI-powered models have significant advantages in the field of healthcare. It has the ability to detect diseases well in advance with high accuracy and protect many lives. However, it is not the ultimate solution, not yet, at least. This can be seen in a study published earlier this month that showed that when AI is applied in the real world, mistakes occur.
Published earlier this month in the journal Modern Pathology, this study is the first one to examine just how the contamination in tissue influences AI's learning models.
In this study, researchers trained three AI models to analyze microscope slides of placenta tissue, with tasks including identifying blood vessel damage, estimating gestational age, and classifying macroscopic lesions. Additionally, a fourth model was trained to detect prostate cancer in tissue samples from needle biopsies.
To test their robustness, the researchers then exposed each model to small portions of contaminant tissue—such as blood and bladder tissue—sampled randomly from other slides. This step aimed to mimic real-world conditions where contaminants are often present.
Upon evaluating the AI models' performance, it was discovered that all four models diverted their focus to the tissue contaminants rather than concentrating on their primary tasks. This distraction resulted in inaccuracies in determining gestational age, diagnosing prostate cancer, identifying lesions, and assessing blood vessel damage. This is because AIs encounter a variety of materials in the real world that they aren't trained on, as opposed to the clean simulated environments in which they are trained.
“Our findings serve as a reminder that AI that works incredibly well in the lab may fall on its face in the real world. Patients should continue to expect that a human expert is the final decider on diagnoses made on biopsies and other tissue samples. Pathologists fear — and AI companies hope — that the computers are coming for our jobs. Not yet.”
– Co-author Dr. Jeffery Goldstein
Not just for AI but for even pathologists who are extensively trained to detect the same, tissue contamination is a pretty common issue. However, the study points out that doctors and non-pathologist researchers are often even amazed to know this. It noted that on a given day, out of the daily 80 to 100 slides scanned by a pathologist, 2-3 are expected to be with contaminants, which they're trained to disregard.
As per the study, AI models perform in a similar manner, much like how humans examine tissues on slides. When examining tissues on slides, humans can only look at a limited field within the microscope and then move to a new field and combine all the information gathered to make a diagnosis.
However, AI was easily found to be misled by contaminants as it had to decide what to pay attention to and what to not. “That's zero-sum,” said Goldstein, who's a director of perinatal pathology and an assistant professor of perinatal pathology and autopsy at the research-intensive Feinberg School of Medicine, Northwestern University, as he explained:
“If it's paying attention to tissue contaminants, then it's paying less attention to the tissue from the patient that is being examined.”
According to the study authors, practitioners should work on quantifying and improving upon the problem of AI models' inability to encode biological impurities.
However, the results have made Goldstein “confident that AI evaluations of the placenta are doable,” calling AI's destruction a “speed bump.” It simply means, “We're on the road to better integrating the use of machine learning in pathology,” he added.
Companies That can Benefit From AI-Powered Prediction Models
With AI models showing great promise, let's now see which companies stand to benefit from them the most:
It is a subsidiary of the tech giant Google that aims to solve intelligence in order to “advance science and benefit humanity.”
Most recently, Google DeepMind researchers revealed that they have come up with an AI assistant for doctors called Articulate Medical Intelligence Explorer (AMIE) that can help diagnose patients. Another paper from the organization proposed WARM to tackle reward hacking in LLMs.
With its shares trading at $152.59, up 8.94% year-to-date, at the time of writing, Google (GOOGL) has a market cap of $1.9 trillion. The company posted a revenue (TTM) of $297 bln while having an EPS (TTM) of 5.21 and P/E (TTM) of 29.19.
The leading healthcare company has been betting on AI to transform the healthcare industry. Siemens Healthineers has developed 84 AI-supported products as of Nov. 2023 and holds over 800 patents related to the technology.
Spun off from its parent company Siemens, the healthcare solution provider has a market cap of $63 bln as its shares trade at $51.68. Siemens Healthineers has an EPS (TTM) of 1.35 and a P/E (TTM) of 38.28.
As we saw, AI holds immense promise in trend analysis, and as technology advances, we will see even more accurate and actionable insights as well as innovation across sectors. In healthcare, AI-driven prediction models synthesize data, insights, and interventions to personalize healthcare and optimize outcomes.