A new study of more than 500 medical students, published in the journal Academic radiology, students find that emerging technology like AI will reduce job prospects for pathology, diagnostic imaging and anesthesia department.
Experts say this perception is not only untrue, but has the potential to jeopardize the global healthcare industry. There has been a severe shortage of pathologists, resulting in delayed outcomes and treatments. In fact, a study in JAMA is open found in the US, the number of pathologists decreased by nearly 18 percent between 2007 and 2017.
This is why we spoke with Dr. Michael Donovan, co-founder and chief medical officer at PreciseDx, a healthcare IT company looking to personalize medicine through artificial intelligence. Donovan seeks to demystify AI in healthcare.
Donovan is vice president and professor of translational studies in the department of pathology at the University of Miami. In addition to her previous academic career at Harvard Medical School and Boston Children’s Hospital, Donovan has over 20 years of experience in the biotechnology industry, serving in a variety of senior management roles.
Donovan earned a Bachelor of Science degree in Zoology, a Master of Science degree in Endocrinology, and a PhD in cell and developmental biology from Rutgers University. He received his medical degree from Rutgers New Jersey Medical School.
Q. A new study of medical students shows that they think emerging technology like AI will reduce job opportunities for pathology, imaging and anesthesiology. You say this perception is not true. Why?
ONE. Emerging technologies like AI really present opportunities for both new and established physicians, especially in service-oriented specialties such as pathology, radiology and even anesthesiology. because they will be at the forefront of advanced medicine.
Today, the reflex reaction of most doctors when they hear words like ‘effective’ or ‘decision aid’ and ‘machine learning’ is a software package that will replace what they trained to deliver. It’s not much different from robotics on an assembly line or the use of supply chain management tools to track and catalog the components needed to make a smartphone.
So the perception is very true; however, the reality is completely different. The advent of AI and machine learning in radiology is perhaps the best example of the immediate benefits, not only for radiologists but also for patients.
Often, what is forgotten is the availability of a tool that can help focus radiologists’ attention on a specific anatomical site or lesion in imaging, providing a measure of accuracy. accuracy that humans cannot reliably and consistently provide, without fatigue and error.
Immediate and long-term benefits include improved diagnostics for management decision-making, while also improving radiologists’ diagnostic skills. This scenario holds true for pathology as well, where digitalization is slowly starting to take over the entire field, and although behind radiology, has proven to be advantageous on many fronts.
Machine learning image analysis devices can highlight “hot spots” in images of tissues or cytological samples for further evaluation, while minimizing some of the risks of missing an important procedure. very important when case and specimen volumes become insurmountable. In both environments, doctors become better diagnoses while staying up-to-date and leading in a rapidly changing field.
The role of artificial intelligence and machine learning in anesthesia is comparable, but with different data streams, where radiologists and pathologists use patient-provided parameters To understand the underlying disease, their main focus is on the image in front of them.
For the anesthesiologist, a more synchronous assimilation between clinical and laboratory values is required to meet immediate clinical needs and determine appropriate pain control and a safe operation. There were no complications during or after the procedure.
The new development of innovative drugs, which incorporate comorbidities and the patient’s current physiological state, while avoiding contraindications, requires advanced real-time data analysis, which is beyond the scope of most doctors. Machine learning and artificial intelligence in this context are intended to ensure patient management and risk reduction, while enhancing the skills and knowledge of anesthesiologists.
The bottom line is that, through the advantages and efficiency of AI-powered care for physicians, the ultimate arbiter, from a medical diagnostic, ethical and legal perspective, is the physician. For example, AI alone cannot make a final diagnosis on an X-ray scan or make a definitive diagnosis on a patient’s needle biopsy specimen.
The tool also does not deliver medication to the patient. The message to medical students considering any of these careers is the following – only the physician can make these final decisions so the “role” Yours will not be replaced but improved and supported with less risk while increasing your knowledge and skills overtime.
Q. As a digital pathologist, why do you believe AI is a big opportunity in the future?
ONE. Digital pathology has demonstrated significant advantages for creating clinical digital archives, providing a mechanism to facilitate log-off of cases, and making records accessible. of any particular case, even if slides and or blocks are missing. Challenges, however, is how to improve the initial assessment, focus on what is important in any given image, and promote diagnostic accuracy, while enhancing the phenotypic characterization of the process. disease process.
Machine learning and AI are available to assist pathologists in their diagnostic evaluation through digital annotation of specific regions. It is also possible – for some disease conditions such as breast cancer and prostate cancer – that the phenotype and even grade of the cancer is based on well-established histological structures, but in a standard manner. quantification and quantification.
The goal of AI in digital pathology is first and foremost to aid the diagnostic process by enhancing the “art” of pattern recognition and bringing the concepts of standardization and quantification into the histological assessment process. cytology.
Q. How can health IT leaders at service delivery organizations help convince caregivers to accept AI rather than fear it?
ONE. Medical IT leaders at various delivery organizations will benefit from spending time deconstructing AI and machine learning processes for end users.
First, they have to define AI and machine learning. The next step will be to outline the benefits of implementing AI and machine learning in organizations and their clinical practices, including data processing, analytics, system accessibility and navigation. electronic health records, with very real examples of everyday implementations.
Additionally, health IT leaders must take the time to reaffirm that the ultimate goal is not to reduce headcount, but to promote a more productive and healthier environment for all employees. . There should also be an education and ongoing reinforcement component highlighting physician-level stress reduction, patient satisfaction, cost-effective care, and positive health economics.
Q. Please give some real-world examples of how AI helps caregivers do their jobs without “replacing” them.
ONE. In the radiological context, there are many examples where AI is playing a very active role in determining response to therapy. Recent advances – particularly in determining responses to immunotherapy – have emphasized the importance of the type of therapy and its nuanced response.
This includes not only variation in size but also in overall appearance or spatial heterogeneity of the tumor on CT imaging. Once again, radiologists are at the forefront and central in leveraging data provided by AI and machine learning to report more accurately response to a particular therapeutic agent, and enhance their own knowledge to meet the demands of the field that is flushing.
Another real-world proof-of-concept example is in the pathological assessment of breast tumor classification where features can be used to determine the grade and differentiation score for cancer.
Currently, these tasks are based on subjectively assessed criteria that tend to be inconsistent, both within and among pathologists. Machine learning image analysis tools have deconstructed layer components and made them objective, normalized, and quantifiable.
Thereby, taking the “guess” and subjectivity out of the scoring and creating a level of diagnostic accuracy that in the near future will likely be included in the management of patients with invasive breast cancer.
Follow Bill’s HIT coverage on LinkedIn: Bill Siwicki
Email the writer: [email protected]
Healthcare IT News is a publication of HIMSS Media.
