Co-authored by Clifford Morris, PhD and Thomas Kent

Paradigm of Artificial Intelligence

Technology moves only in one direction—forward. In general, the observation of this fact is welcome. After all, the ability to develop and apply technology is a defining aspect of our species. This fact becomes unwelcome when outcome-driven industries trade their lots for technological efficiency. In medicine, the efficiency economy has resulted in doctors spending less time with patients. This is highlighted by an average first appointment time of only twelve minutes.1 Before we continue, it is important to acknowledge that the gains we have made in efficiency are important to total patient outcomes, as more efficiency equates to more treated patients. Instead, we are concerned with individual patient outcomes.

For too long, medicine, enabled by the worst aspects of technological advancement, was trending in the direction of cookie-cutter solutions to the unique needs of individual patients. When pondering how to rehumanize medicine, AI is often the last thought on one’s mind. Yet, the current data may lead one to the opposite conclusion. AI seems to be able to give us this lost time back, allowing for doctors to give patients the time they truly need. The extra human interaction time is not just beneficial for the patients, but for the physicians as well. A study conducted at the University of Colorado showed that taking the computer out of the exam room and supporting doctors with human medical assistants significantly reduced burnout rates.2 If AI could use its data-processing ability to help with the more clerical work nurses do, it could permit nurses more time with both doctors and patients. The placement of AI in a clinical setting may facilitate more one on one time between doctors and patients, leading to a reduction in nosocomial infections and hospital readmissions as well.1,3 This could save the health care industry millions, while providing superior patient care. All that is needed is for someone to commit to the first push.

The Role of AI in the Clinical Lab Field

Although AI may not be prevalent in the clinical laboratory sector of medicine yet, saying that it is non-existent altogether is far from the truth. For example, AI is already prevalent in the laboratory setting. A great example of this is AI in the Liquid Chromatography Mass Spectrometry (LC-MS) field.4 LC-MS is a great tool used to measure various compounds in the human body, including everything from hormone levels to trace metals. One of the ways AI has already integrated with LC-MS is how it cuts down on the rate limiting steps of LC-MS, which more often than not are sample prep and LC separations. One system that Physicians Lab has made use of is parallel processing using SCIEX MPX 2.0 High Throughput System. This system can couple parallel runs with one LCMS instrument, resulting in twice the speed with no loss to accuracy. It can do this by staggering two runs either using the same method, or different methods entirely. What really makes this system great is its ability to automatically detect carryover and inject solvent blanks to clean the instrument. The system will then continue its analyzing, while automatically reinjecting samples that may be affected by the carryover. It will also flag high concentration without user input, allowing for easy detection of possibly faulty samples. This allows it to operate without users from startup to shut down. Some of the other ways that it can be used to increase efficiency are by using integrated network features to work on anything from streamlining management to increased throughput. Physicians Lab is also taking advantage of AI systems by incorporating the ASCENT software from Indigo Bioautomation. This software uses smart algorithms and machine learning to automatically integrate the LCMS peaks and turn them into results – a tedious job normally done by people. A major issue in laboratories using LC-MS is that the result interpretation of peaks is highly user-subjective and dependent on their attention, training, and behaviors.5 Using ASCENT eliminates user-to-use variability, improves workflow speed, and allows the users to focus only on problematic result interpretations. The result is a highly dependable and consistent result interpretation system, as well as a better utilized staff. Finally, just about every process from plate and sample vial changing to gradient formation can be automated, allowing for fast and accurate results. AI has allowed all of these changes, and it has shown what the power of a machine that can adjust to new circumstances can do. These features can almost double lab work speeds in some cases, allowing doctors to diagnose quicker and attack the problems faster.

AI has played a large role in advancing LC-MS instrumentation, but that is not the only aspect of clinical laboratory life that AI has enhanced. All AI-based lab software works through a term known as “computational pathology”. This means that through the use of visuals and machine learning, the AI makes the product of the data more useful and easily understood for practicing physicians. One of the ways this is achieved is through wearables that can measure blood glucose levels, heart rate, and temperature, as well as other factors. This data can be uploaded to a mass cloud. Another way AI could easily influence lab settings is through tumor detection.4 Tumor detection is generally achieved by analyzing set genes that are already associated with tumor growth, and comparing the genes that are mutated in patients with tumors. This allows for personalized therapy, depending on what tumors the patient has. AI will use machine learning along with traditional techniques for more accurate tumor identification and diagnostic process. Also, the ability to integrate pre-existing algorithms with machines gives them the ability to have a greater pattern recognition than even the best doctors.

The Greater Extent of AI in Healthcare.

AI in the lab will also help doctors design more accurate treatment paths by analyzing real-time data. The data constructed from laboratories are generally broken up into three sectors: patients about to be diagnosed, patients previously diagnosed who are responding to treatment, and patients who have been diagnosed that are not responding to treatments. If one could take this data and integrate it into AI, you could get more effective treatments to patients in a shorter time span. For example, instead of going with a primary treatment that works 80% of the time, AI may catch something that would instead recommend a less-used course of action that would be more effective. So instead of going through a trial and error of multiple treatments, AI could help find the better solution. As well, if the first line therapy fails, this could provide a faster way to reach the second line therapy. In order for AI to truly reach its maximum potential, there must be a shift in thinking. Instead of relying solely on past data, doctors would be able to rely on real-time, ever-changing data to get the best results possible.

So with AI becoming prevalent in the clinical lab, how does this translate into doctor-to-patient interaction? Theoretically, the AI would use one of two ways to treat a patient in a clinic.1 The first being the flowchart method. This is where a doctor would essentially transfer all of their data from past interviews of patients and the results into the AI. The AI would then have this knowledge, and be able to ask patients questions and form conclusions in an “If x, then y” format. There are two main issues with this method, the first being the sheer amount of data that needs to be integrated. The second is the AI’s inability to detect when patients may be hiding something. For example, someone lying to get painkillers. This could be where cohesiveness steps into play, as you could have the AI ask questions, with the doctor overriding some of the patients answers. The second method is known as database learning. This requires the machine to be shown the same image over and over again, until it can memorize that image. This could be useful if a doctor sees a mole that he’s not sure is a melanoma or not. The doctor could then go to the computer for a second, and possibly more accurate, opinion. Some of this AI is already here. There is already basic flowchart AI, robotic surgical systems, AI therapy, and AI scheduling programs. AI is bound to be even more integrated into clinical medicine as time marches on. One example of this is Stanford University’s Program in AI-Assisted Care (PAC). One of these programs allows seniors who live alone to be able to reach immediate help if needed, along with monitoring behavioral and movement patterns to spot irregularities. Other examples of future AI in the medical field involve “Molly”, a virtual nurse that can monitor and follow up with patients. This could free up an immense amount of doctors’ time. Not to mention that although certain AI has already been able to help schedule, there are more advanced AI’s which will be able to adapt to the struggles of a busy ward, freeing up more time for nurses.

Although AI may be originally thought of as a bane to physicians and the remaining human aspect of medicine, this is far from the truth. Throughout the continued advancement of AI in the clinical lab, and the newer yet fascinating use of AI in the clinic itself, medicine has a chance to regain its once human form. The abilities of AI to help with anything from scheduling to diagnostics should reduce the combination of doctor’s stress levels, burnout, and hospital readmission rate. This, in turn, should give the doctors much more one on one time with patients, drastically increasing the quality of patient care. The re-humanization of medicine starts not with man, but with the machine.

References:

1.        Topol EJ. High-performance medicine: the convergence of human and artificial intelligence. Nat Med. 2019. doi:10.1038/s41591-018-0300-7

2.        Wright AA, Katz IT. Beyond burnout – Redesigning care to restore meaning and sanity for physicians. N Engl J Med. 2018. doi:10.1056/NEJMp1716845

3.        Naugler C, Church DL. Automation and artificial intelligence in the clinical laboratory. Crit Rev Clin Lab Sci. 2019. doi:10.1080/10408363.2018.1561640

4.        Workman TE, Hirezi M, Trujillo-Rivera E, et al. A Novel Deep Learning Pipeline to Analyze Temporal Clinical Data. Proc – 2018 IEEE Int Conf Big Data, Big Data 2018. 2019:2879-2883. doi:10.1109/BigData.2018.8622099

5.        Gaona García PA, Montenegro Marin CE, Gaona García EE. Model of Learning Objects Exchange between LCMS Platforms through Intelligent Agents1. Ing y Univ. 2015. doi:10.11144/javeriana.iyu19-2.mloe