AI (artificial intelligence) technology creates smart machines capable of performing tasks that typically require human intelligence. Simply put, with AI, computers and other machines can learn, think, make decisions, and take actions. And AI systems can store vast amounts of data. Importantly, AI can improve healthcare since it is more efficient, exponentially faster, and cheaper than humans. For instance, AI can find patterns which can be used to diagnose patients and predict outcomes.
As I discussed in a previous post, using AI to analyze voices holds promise for the early detection of many diseases, including Parkinson’s disease, Alzheimer’s disease, multiple sclerosis, and coronary artery disease. However, voice analysis is only the tip of the iceberg when it comes to how AI can improve healthcare. AI can help with diagnosis, improve treatments, aid in decision making, and more. And it’s a rapidly growing field, with many new applications on the horizon.
Here are just a few of the ways AI can improve healthcare:
AI can improve healthcare by aiding with early disease detection and diagnosis.
AI can help doctors more easily, and more accurately, detect diseases in their early stages. Additionally, AI can help doctors diagnose patients, and predict who is at risk for certain conditions in the future.
Below find a few examples of how AI can help healthcare by aiding in detecting, diagnosing, and predicting medical conditions:
AI can help pathologists analyze samples.
When doctors suspect you might have an illness, they may send a biopsy or sample to a pathologist for analysis. Although we all deserve an accurate diagnosis, pathologists must rely on their judgement, which can lead to mistakes which in turn can cause us significant harm. For instance, pathologists can miss cancer cells (false-negative) or erroneously report the presence of cancer (false-positive).
Of course, an accurate diagnosis is critical to receiving the appropriate and timely treatments. However, research shows that pathology reports contain inaccurate diagnoses 3-9% of the time. Fortunately, AI-based pathology can produce more accurate results, reducing the subjectivity that can lead to errors.
For example, PathAI is creating AI driven digital pathology tools that could help pathologists make faster, more accurate diagnoses. Additionally, their technology could make it easier to accurately measure how well a treatment is working for a patient. Importantly, this technology could help doctors create personalized treatment plans for each patient, making modifications as needed.
AI can predict risk of atrial fibrillation and associated strokes.
Atrial fibrillation (AFib), the most common cardiac arrhythmia, is associated with many health risks, including stroke and death. Although over 300 million electrocardiograms (ECGs) are performed every year in the US to measure electrical signals from the heart, ECGs cannot “generally detect future potential for negative events like atrial fibrillation or stroke.”
Using data gathered from 1.6 million ECGs over 35 years, researchers from Geisinger and Tempus created AI that makes ECGs smarter and more capable of predicting future cases of AFib. Excitingly, this technology can identify patients who are likely to develop AFib, including those at an increased risk of AFib-related strokes.
Importantly, the AI exceeded current clinical models for predicting AF risk. This development will help doctors and patients take steps early in the process, potentially preventing AFib-related strokes.
AI can predict heart attacks.
A team at Johns Hopkins University created AI technology that can predict if and when a patient could die of cardiac arrest. The technology, which uses raw images of patient’s diseased hearts and patient backgrounds, significantly improves on doctor’s predictions regarding risk of cardiac arrest. In fact, this technology provides each patient with a highly accurate prediction regarding their chance for a sudden cardiac death over 10 years, and when it’s most likely to happen.
Importantly, this development could revolutionize how doctors make treatment decisions and increase survival rates associated with sudden, deadly cardiac arrhythmias.
AI can improve breast cancer detection.
Breast cancer is the most frequently diagnosed cancer. And it’s the leading cause of cancer-related deaths among women worldwide. Clearly, identifying breast cancer at an early stage before metastasis allows patients to receive more effective treatments which significantly improve survival rates.
Although mammograms and ultrasounds are widely used to screen for and diagnose breast cancer, neither of them is perfect. In some cases, cancer is missed, and in other cases, patients receive unnecessary biopsies due to false positive results.
Fortunately, researchers are using AI to improve the accuracy, consistency, and efficiency of cancer diagnoses based on mammograms and breast ultrasounds, including the following developments.
AI may improve mammogram accuracy.
Although mammography has been widely used to screen for breast cancer, it is not perfect. For example, accuracy drops dramatically for dense breasts.
Overall, the sensitivity of mammography is about 87 percent – meaning it correctly identifies about 87% of people who truly have breast cancer. In other words, screening mammograms miss about 1 in 8 breast cancers, giving people false negative results.
Conversely, the false positive rate is 7-12% after one mammogram (younger women are more likely to have a false positive results). But, after 10 yearly mammograms, the chance of having at least one false positive result is about 50-60%.
Fortunately, researchers found that AI can review and translate mammograms 30 times faster than humans, with 99% accuracy, reducing the number of unneeded biopsies. Furthermore, the AI software can accurately predict breast cancer risk, helping doctors closely monitor closely those most at risk of developing this potentially deadly cancer.
AI may improve breast ultrasound accuracy.
Sometimes patients need a breast ultrasound if a problem is detected during a mammogram or physical exam. Although ultrasounds can detect cancers obscured on mammography, including in dense breasts, they also have a high false-positive rate.
Fortunately, AI may reduce the number of false positives. For instance, in one study, when radiologists used AI to analyze breast ultrasounds, their false positive rates decreased by 37.3%, the number of biopsies recommended decreased by 27.8%. Moreover, they were able to maintain the same level of accuracy.
Interestingly, in May, 2022, the FDA approved iSono Health’s ATUSA – a whole breast ultrasound system that combines AI with their unique wearable scanner. This portable ultrasound system generates whole-breast images in just 2 minutes, compared to 10-15 minutes with traditional ultrasound techniques. Importantly, the system uses AI to generate 3D images, which can help doctors with decision making and disease management. And they are working to validate their AI to improve the process of locating and classifying breast lesions.
AI can predict lung cancer recurrences.
Lung cancer is a deadly disease. And for some types, there is a very high rate of recurrence. Logically, if doctors can predict which patients will face a recurrence, they could catch regrowth early, and potentially improve survival rates.
Researchers found AI could predict a lung cancer patient’s risk of recurrence two years after radiation therapy. Importantly, the AI made more accurate predictions than the system currently used. The team created AI to analyze the patient’s tumor size and stage, as well as the type and intensity of radiotherapy they underwent. Additionally, the AI considers the patient’s smoking history, their body mass index, and age.
AI can improve prostate cancer detection.
Prostate cancer, one of the most common cancers among men, is generally diagnosed with a blood test for Prostate-Specific Antigen (PSA) levels. Unfortunately, PSA blood test are not very accurate, which can lead to unnecessary, invasive biopsies (often accompanied with side effects) or missed cancer. Importantly, about 75% of men with a raised PSA level will not have cancer. Conversely, the PSA test can also miss about 15% of cancers.
Interestingly, reliance on a PSA test may become a thing of the past. Researchers in Korea recently developed a technique using AI to diagnose prostate cancer from urine. They can detect prostate cancer in 20 minutes with almost 100% accuracy. Certainly, when the use of this technology becomes widespread, men could receive a prostate cancer diagnosis without the risks associated with biopsies.
AI can make MRIs faster.
MRIs are not fun for most people. The scans often take up to an hour. It can feel claustrophobic, and the noises can seem unbearably loud. Never mind having to hold still the entire time – not even allowed to scratch the itch that inevitably appears when you are told not to move for any reason.
Excitingly, a new collaboration between Facebook’s AI research team and radiologists at NYU Langone Health can make MRIs more tolerable. The project, called fastMRI, aims to investigate the use AI to make MRI scans up to 10 times faster than now.
Simply put, fastMRI uses AI to create images in a new way that requires about 4x less data from the MRI machine. Importantly, radiologists found the fastMRI images were diagnostically identical to traditional MRI images.
The benefits of faster MRI scans are many, particularly less stress-inducing time in MRI machines. Additionally, faster scans will allow greater patient throughput, reducing the wait for an available machine and improving hospital revenue.
AI can detect acute kidney disease.
Acute kidney injury (AKI) is the sudden onset of kidney failure or kidney damage that happens within a few hours or a few days. AKI (also known as ARF – acute renal failure) causes a build-up of waste products in the blood and makes it hard for kidneys to keep the right balance of fluid in the body. AKI can also impact the brain, heart, and lungs.
Sadly, AKI is common among hospitalized patients, particularly among older patients and/or those in the ICU. Moreover, AKI leads to 500,000 deaths in the US each year.
Fortunately, DeepMind developed an AI technology that can predict the presence of AKI up to 48 hours before it happens by analyzing medical records and interpreting test results. Since AKI requires treatment within hours, or even minutes, of diagnosis, this technology could potentially save thousands of lives.
AI can improve healthcare by improving treatment options.
Firstly, AI’s predictive analytics can provide information that can help patients and doctors make informed decisions about treatments. Additionally, AI technology can improve the efficacy of treatments and monitor patient progress. Here are a few examples:
AI can enhance electrical brain stimulation.
Electrical brain stimulation helps people with epilepsy and movement disorders such as Parkinson’s disease. In the future, electrical stimulation may help people with psychiatric illness and direct brain injuries, such as stroke.
Interestingly, Mayo Clinic and Google Research have developed a new AI algorithm to improve the electrical brain stimulation process. AI can help doctors understand how different regions of the brain are connected and how they interact with each other. With this information, doctors can make better decisions about where to place electrodes during the stimulation.
AI can help with IVF process.
During in vitro fertilization (IVF), doctors retrieve mature eggs from women and fertilize them sperm in a lab. Then doctors choose one or more fertilized egg (called an embryo) and transfer the egg(s) into a uterus.
Unfortunately, getting pregnant through IVF is stressful and costly. For most women, the chances of conceiving are 20-35% per IVF cycle. However, the likelihood of getting pregnant decreases with each successive round, while the cost increases. For 3 full IVF cycles, most women have a 45-53% chance for a successful pregnancy.
Therefore, anything that improves the chances of a successful pregnancy is an excellent development.
Fortunately, doctors can use AI to predict how likely an embryo is to develop as far as the stage of having a fetal heart. Then they can use this information to select the best embryo for transfer, with a goal of helping women get pregnant more quickly.
Currently, an embryologist uses a standard grading system to assess the appearance of each embryo under the microscope. In contrast, AI assesses images from a 5-day growth period and identifies the embryo with the greatest chance of fetal heart development.
Researchers found the AI system, which reviews hundreds of images from each embryo, was a significant improvement of any other system used to select embryos.
Although this technology is currently not widely available, it will hopefully be in widespread use soon.
AI can improve access to mental health therapy.
Accessing help for mental health has become increasingly difficult, with an increased demand due to COVID-related anxiety and depression. In fact, experts believe only half of people with mental health issues receive treatment, primarily due to clinician shortages, fragmented care, and societal stigma.
Fortunately, AI can help. For example, AI could improve treatment effectiveness by providing more insight into patients’ needs. And it could improve therapists’ techniques.
There are many ways in which AI can help therapists and patients. For example, AI can:
- Analyze language used in therapy sessions, to improve quality control and for training purposes.
- Help doctors identify mental illness earlier and help match patients with the “right” therapists.
- Analyze troves of data, which can help doctors create effective treatment plans based on which treatment techniques work best for specific combinations of symptoms, histories, and responses to prior treatments.
- Help identify when it’s time to alter a treatment, or if it’s time for a different therapist.
Additionally, AI could fill gaps in care, helping moderately ill people access effective care. For instance, researchers are developing an AI agent for those with moderate depression or anxiety. The app will act as a virtual mental health agent. As such, it will take patients through steps and strategies using a validated treatment protocol.
Although it is in the early stages, if the app proves to be effective, it could provide easily accessed treatment for people with moderate mental health issues.
AI can improve healthcare by identifying new treatments.
AI can help researchers develop new treatments and medications. Using traditional methods, it can take more than a decade and billions of dollars to develop and test new medications. However, using AI can speed up this process, not only making it more cost-effective, but it also helping new innovations reach patients more quickly.
For instance, Insilico Medicine used AI to identify a new drug in only 46 days to target a protein linked with fibrosis (tissue scarring)! Their software analyzed vast amounts of data that would take human years to evaluate. Similarly, using AI, it took Atomwise less than a day to identify 2 drugs with significant potential to reduce the spread of Ebola.
Additionally, Alphabet subsidiary DeepMind used AI to solve a decades-old protein-folding challenge. This approach will hopefully help identify new treatments for diseases with faulty proteins.
And Relay Therapeutics developed AI technology that finds ways to reach hard-to-hit protein targets with medications. Using this technology, they have two cancer drugs in early clinical development, with a 3rd close behind.
Lastly, BERG is an AI-powered biotech company using AI to gather trillions of data points to map diseases and revolutionize treatments for patients. They aim to speed up the discovery and development of treatments for those who need novel treatment options. Additionally, they aim to provide more effective personalized precision treatments, and to reduce healthcare costs.
Although it appears certain that AI can improve healthcare, it’s always important to engage in your care for the best outcomes. Read these posts for helpful tips:
- 10 Steps to Reduce Your Risk of Diagnostic Error.
- Why Take Detailed Notes at Doctor Appointments?
- Should You Record Medical Appointments?
- Can You Trust Medical Information Online?.
- Can You Trust Advice from Other Patients?
- Understanding Medical Information Is Harder Than Most Realize.
- The Dangers of Too Many Tests and Treatments for Patients.