Eric He is Suki’s president and tech-savvy CEO with experience in co-founding and scaling companies including Hotwire and Expedia.
I recently wrote about the promise of artificial intelligence and its potential to play an important role in transforming how clinicians interact with technology. Even today, artificial intelligence is making meaningful advances in disciplines ranging from radiology to cardiology. The ability of AI to help doctors work faster and with greater accuracy has industry analysts predicting explosive growth of more than 10-fold in this decade alone, with estimates reaching $96 billion in 2028.
However, most clinicians are starting to learn about AI and understand its use cases. This article will take a look at what AI can currently do in medical settings and where more time and innovation is needed.
Current strengths of artificial intelligenceLiterally speaking, AI can support today’s clinicians with enhanced data analysis and workflow improvements like the following.
1. Historical health data organization
Healthcare providers own huge data sets across multiple technology platforms. Advances in deep learning, a type of artificial intelligence that reflects how humans acquire knowledge, are now enabling rapid analysis of highly complex data sets to provide useful insights for clinicians.
for example, There are solutions today That process Electronic health records Data to model how the disease progresses while also providing recommendations for intervention. A major issue is the quality of the underlying information, as EHR data can be inconsistent, and the richest information is often in a free text format that is not easily parsed.
2. Process visual images and check common diagnoses
AI can process images such as radiographs and retinal images for diagnostic purposes at faster rates and in greater detail than human doctors. for example, Hospitals in India She is currently trialling an AI-based program to check images of a patient’s retina for signs of diabetic retinopathy – a condition often diagnosed too late to prevent vision loss.
3. Leverage data to suggest treatment options
AI learns as it is and becomes more accurate and insightful with larger amounts of high-quality data. During a study at UNC School of Medicine, IBM’s Watson digested more than 160,000 cancer papers published each year and leveraged their massive data-processing capabilities to offer treatment options that human doctors missed. 30% of the time. While some believe the UNC study has been overstated, it is fair to conclude that the AI’s data processing and analysis capabilities will only improve over time.
4. Improving the workflow of the doctor
In addition to detecting clinical trends, AI can also capture workflow habits and provide smart and timely suggestions to make clinicians more efficient and productive. For example, AI is now used to run scheduling and data collection applications and Perform other administrative tasks.
5. Reduce documentation time
Physicians spend two hours documenting and writing work for each hour they spend with the patient, a ratio that often results in this doctor burnout. Natural language processing (a form of artificial intelligence) can recognize a doctor’s voice and use it to document care through a health system’s electronic health records. by Simplify boring documentation workflowAI voice recognition apps allow clinicians to spend more time interacting with patients.
Current AI Improvement OpportunitiesAs far as recent advances in AI in healthcare, there are still limitations, largely in four areas.
1. Comprehensive data
Experienced doctors are incredibly adept at getting insights from their patients and their environment. Every detail noted can make a difference, and training all of an individual’s senses to aid in patient diagnosis is not learned through textbooks but instead through a long and meticulous apprenticeship model.
While AI can ingest huge amounts of EHR and textbook data, it cannot generate valuable real-time patient data that practitioners in the field acquire with all their senses.
2. Measure and respond to human emotions
AI cannot interpret human emotions to open up more information and insight in a clinical setting, which is important for gaining a broader understanding of a patient’s issue or need. Perhaps not surprisingly, AI currently lacks empathy when dealing with a patient and does not yet pick up on any social or physical cues.
However, is empathy and emotional differentials something that AI can learn? Can machines and humans communicate in ways that go beyond data and logic? The potential development of artificial intelligence in this field is one of the many reasons why the next 10 years or more will be so impressive.
3. Understand the social context
Artificial intelligence does not understand what information is useful and what is just small talk or noise. Advances in computing now make it possible to transcribe entire conversations, identify different speakers, filter out background noise, and even understand different dialects. But it is quite another task to take the texts and synthesize them in a way that tells the patient’s story as a doctor does.
For this reason, current ambient solutions that document entire patient visits remain highly dependent on humans behind the scenes to extract social chatter from actual clinical data.
4. Accounting for biases in the data
The use of AI in data sets across healthcare could lead to a problem now known asArithmetic bias“which is when an algorithm inflates existing inequalities across race, gender, socioeconomic status, or sexual orientation. Conventional medicine faces similar challenges, but it is then possible to trace what causes bias to occur. Using artificial intelligence, a “black box” component makes it difficult Understand why AI draws some conclusions.
concluding thoughtsAt first glance, the limitations of AI reflect the shortcomings of medicine in general. Most of us have, at some point, experienced a doctor who has poor bedside treatment or lacks the emotional intelligence needed to capture important clinical insights. Some of us may have experienced discrimination during a patient visit or suffered from a misdiagnosis due to our doctor not correctly calculating our race or gender.
The limitations of AI are not limited to AI, but as investment, innovation, and regulatory oversight continues, AI in healthcare can continue to take important steps forward and make the industry work better for everyone. Looking to the future, it is likely that the most important advances in healthcare for artificial intelligence will intertwine the best of humanity and computing to create a unit that works best.