SAN DIEGO, CA / ACCESSWIRE / May 22, 2023 / From this article you will learn:
How AI diagnostics is revolutionizing healthcare by enabling medical professionals to make faster and more accurate diagnoses, leading to improved patient outcomes and customized treatment plans.
The role Diagu’s AI tools, such as Diagu Fire and GetLabTest, play in analyzing lab results and providing valuable diagnostic suggestions to improve physician decision-making.
How the implementation of AI is impacting efficiency, streamlining hospital workflows, and improving coordination between testing labs, physicians, and patients.
The potential of AI technology to make healthcare more accessible and portable, reach underserved populations, and reduce the workload of hospital staff.
How AI diagnostic tools can better predict patients’ future health problems, enabling physicians to develop innovative and proactive treatment strategies and ushering in a new era of diagnostic medicine.
The benefits of emerging AI technologies for healthcare systems, hospitals, physicians, and patients, including cost savings, more efficient administration, and improved communication.
How the role of physicians and medical professionals is evolving in the context of AI diagnostic tools, focusing on the collaboration between human expertise and machine learning to deliver better patient care.
Artificial intelligence-the application of digital computing to tasks otherwise associated with human thought, analysis, and judgment-has begun to find a place in every sector of our economy. Medical science is no exception. Around the country, doctors are worried that artificial intelligence has begun to displace them. AI’s champions are thrilled: after all, diagnosis is about analyzing data, and that’s what AI does best.
They’re all wrong.
Between dread and hype lies the truth of AI’s value to the medical community, particularly as a diagnostic tool. It’s true that AI has increasingly asserted its role in the healthcare field-more than 90% of US hospitals have a strategic plan to integrate AI with their standard modes of practice. AI helps the Mayo Clinic search for biomarkers indicating cancer, facilitates the Cleveland Clinic’s effort to mine its pathology specimens for pathology insights, and even predicts which patients at Boston Children’s Hospital will miss their appointments.
But recent studies suggest that AI’s most productive role is as a complement to human diagnosis, not as its replacement. Researchers at MIT’s Computer Science and Artificial Intelligence Lab, for example, found that a carefully calibrated hybrid model involving both AI and human participation resulted in an 8% improvement in the diagnosis of cardiomegaly, exceeding the accuracy of AI or human diagnosis alone.
A new generation of collaborative AI applications positions AI as a tool and a guide for medical personnel without presuming to encroach on the roles of doctors and nurses. Diagu leads this emerging wave with a set of AI tools designed to facilitate quicker, more accurate diagnoses of a wide range of medical conditions. Founded in 2012, Diagu has invested more than ten years of research and development into AI tools that analyze laboratory test results with exceptional sensitivity to a nearly comprehensive range of patient conditions. Its platform promises to benefit labs, patients, and doctors alike:
Testing labs receive unprecedented support from Diagu’s GetLabTest product, which helps patients or their doctors to arrange sample collection and delivers test results to their doctor or a recommended specialists.
Patients and their doctors receive quicker, more accurate suggested diagnoses of a wider range of conditions than previously possible-even conditions that elude the traditional threshold for diagnosis.
Doctors benefit from AI’s power to discover patterns in large volumes of data, while retaining oversight of the entire diagnostic process. When patients are referred through GetLabTest, their doctors or specialists begin their consultations with a rich set of data, allowing them to serve patients more quickly and accurately.
Implications for diagnostic practice
The means by which AI models are trained have significant implications their use as medical diagnostic tools. A 2022 review of AI in disease diagnosis published in the Journal of Ambient Intelligence and Humanized Computing found that “The most common challenge faced by most of the studies was insufficient data to train the model.” Various applications of AI for medical diagnosis reflect this challenge to varying degrees. In some applications-the interpretation of radiological imaging, for example-AI is extremely adept. As the authors of a study in Nature Reviews Cancer put it, “AI excels at recognizing complex patterns in imaging data and can provide a quantitative assessment in an automated fashion.” Trained on a sufficiently representative range of images, AI models have outperformed human control groups while delivering results more quickly. We must remember that this blend of speed and accuracy has as much to do with human discretion as with the inherent strengths of AI.
Speed is easy enough to measure, but assessing diagnostic accuracy requires prior knowledge of the patient’s actual condition-and this in turn demands a final, definitive assessment by a human diagnostician. As the Nature Reviews Cancer article continues, “More accurate and reproducible radiology assessments can then be made when AI is integrated into the clinical workflow as a tool to assist physicians [emphasis added].”
A recent study of AI’s ability to analyze lung scans underscores this point. Researchers at the University of South Carolina found that a well-trained AI model was able to identify certain types of lung nodules more accurately than humans, and roughly four times more quickly. The study measured accuracy against the model’s ability to identify nodules appearing on lung tissue, which it did with a sensitivity comparable to that characteristic of seasoned radiologists. Before it identified those nodules, the AI model needed to be trained to recognize them. Had the study used an imperfect set of data to train the AI model, its results may well have been significantly different.
Perhaps most strikingly, the study found that when experienced radiologists used AI as a tool for evaluating lung scans, they achieved accurate results with a greater degree of confidence in roughly a quarter of the time they spent on each scan without the assistance of machine learning. These results help describe the vanguard of AI as a diagnostic tool: far from replacing the expertise of human diagnosticians, AI works best as a facilitator. In the University of South Carolina study, AI played a discrete role: to improve the speed and confidence with which veteran radiologists accurately assessed lung scans. It accomplished this by presenting its own conclusions for the radiologist to review and confirm, greatly narrowing the scope of reasonable possibilities the radiologist was burdened with assessing. The radiologist’s role remained essentially the same. AI simply began the process of human analysis at a point further down the road to accurate diagnosis than had previously been possible. This vital interplay of machine learning and human judgment is the future of AI-assisted clinical diagnosis.
Diagu’s flagship Diagu Fire system and its GetLabTest solution embrace the fundamental reliance of AI on human discretion. Diagu’s platform uses AI to evaluate a patient’s full medical history against the knowledge base it has already established by analyzing millions of other, anonymized patient histories. When doctors order lab tests, Diagu facilitates the reporting of test results, supplementing them with its own findings and proposed diagnoses. In the process, it avoids the limitations of first-generation diagnostic AI solutions by focusing on what AI does most reliably: interpreting test results and presenting its conclusions to doctors in the ways that best support quick, accurate diagnoses.
A new generation of diagnostic AI
Diagu represents a significantly different application of AI from those that have become established in medical practice. Most of the examples cited in this paper necessarily reflect AI’s first generation as a diagnostic tool. First-generation diagnostic AI solutions typically perform-at greater speed and often with greater accuracy-interpretive functions previously performed by humans. Radiological scans, for instance, benefit greatly from this application of AI, which excels at the sort of pattern-matching necessary to identify and characterize anomalies found in narrowly defined ranges of images. Crucially, this application of AI replaces human analysis on a 1:1 basis: it seeks to render existing workflows more efficient and effective but does not substantially change them.
The emerging generation of diagnostic AI tools promises to optimize the diagnostic process much further, allowing hospitals to establish new, more effective workflows that make full use of both human and machine analysis. Diagu’s ability to convey accurate, highly informed diagnostic suggestions along with test results-and to deliver all of this information more quickly than previous workflows allowed-is a good example of diagnostic AI’s next wave. While Diagu’s approach makes it uniquely appropriate to a broad range of diagnoses and clinical settings, it shares an approach to AI with some other medical technology companies. Freenome, for example, uses computational biology and machine learning to speed the early diagnosis of cancer and to support more efficient and effective treatment planning. PinPoint Data Science offers a similar cancer-screening solution based on multiple blood analytes.
Diagu’s GetLabTest service is a unique application of AI in the healthcare industry. While some companies offer similar services, they are more narrowly focused. eConsult Health and Mobile Phlebotomist are healthcare companies operating in several European Union countries and the UK, providing digital triage and remote consultation capabilities, and home-based phlebotomy services, respectively. Both companies are now expanding their reach and seeking partnerships in the United States to bring their services to a new market. It will be interesting to see how they adapt to the US market and what opportunities they discover.
AI as an Interpretive Tool
AI’s pattern-matching capabilities make it an especially promising tool for interpreting test data. A study conducted at Massachusetts General Hospital offers a good example of how AI can extend a hospital’s existing diagnostic methods. Using low dose computed tomography (LDCT) scans gathered from three different sources, researchers developed an AI model capable of analyzing single LDCT scans in real time and determining the risk of lung cancer among the nonsmoking population. Because the study drew on historic data, researchers were able to confirm that its results were remarkably accurate in determining cancer risk up to six years into the future. Traditional techniques for assessing lung cancer risk require multiple LDCT scans for each patient and arrive at predictions only after analyzing series of scans.
As with the earlier example from the University of South Carolina, the Massachusetts General study points to a revolutionary breakthrough in diagnostic practice that brings the roles of doctors and AI tools into focus without abrogating the central role of human diagnosticians. This breakthrough begins before a radiologist or doctor receives scans or other test results: AI can make existing technologies more sensitive and more accurate. As we have seen, AI conveys its greatest benefits in diagnostic settings when it enhances the results of traditional tests such as radiological scans and gives diagnosticians a guide to their interpretation. This is especially true of LDCT scans: questions remain as to LDCT’s sensitivity, and its false-positive rate has been measured at rates of up to 26.6%. The Massachusetts General study used a highly optimized classification tool, Lung-RADS 1.0, to measure the accuracy of non-AI-guided LDCT scan evaluation; this tool yielded a false-positive rate of 0.10 on a cohort of scans known to be either positive or negative over the study’s six-year timeframe, and a FPR of 0.14 on the entire baseline cohort. AI-guided analysis yielded commensurate FPRs of 0.08 and 0.08.
More accurate results encourage timelier and more effective diagnoses. LDCT has already proven itself as the centerpiece of preventive measures for patients who may be at risk for lung cancer, but its reputation for generating false positives gives doctors reason to think carefully about how they act upon its findings. Invasive tests, for example, can greatly improve a patient’s prospects when they confirm true positives, but they disprove false positives at a high cost, both in terms of the time and money required and as negative impacts on the patient’s health. Those costs are especially cruel when invasive testing was ordered on the basis of an avoidable false positive. AI’s proven ability to improve the accuracy even of radiological techniques not known for high degrees of sensitivity could truly revolutionize the way in which initial diagnoses are confirmed, getting patients more quickly into more effective courses of treatment.
New advancements in AI facilitate more common forms of medical testing, from bloodwork and urine tests to images of all kinds, and it is in this capacity that the fullest, most reliable application of AI to diagnostic practice is likely to be realized. This view comports with the stated vision of Sumeet Chugh, director of the Division of Artificial Intelligence in Medicine at Cedars Sinai Hospital. In Chugh’s words, his team devotes itself to “solving existing gaps [emphasis added] in mechanisms, diagnostics, risk assessment and therapeutics of major human disease conditions.” In the view of the researchers most closely in tune with emerging applications of AI, then, the methods and frameworks created and maintained by human doctors still form the foundation of modern medical practice.
Seen as a complement to existing diagnostic practice, AI-enhanced methods can greatly increase the speed and accuracy with which test results are reviewed, interpreted, and used to inform diagnoses. Diagu has made this a defining characteristic of its Diagu Fire lab test analyzer. Working directly with partner hospitals in the UK and Europe, Diagu Fire is initially trained on millions of medical records, including test results and confirmed outcomes. It then pre-processes each patient’s full medical history while specimens are away at the testing laboratory. This gives Diagu’s AI model a head start toward recommending diagnoses based on the patient’s latest round of test results. Drawing on what it has learned from its baseline training, the model identifies patterns in each patient’s health history, which allows it in turn to identify ranges of test results as consistent or anomalous with the patient’s documented health to date. While the laboratory releases its testing results, Diagu Fire incorporates them into its previous analysis, delivering in near-real time a focused set of diagnostic possibilities and treatment options, complete with the assessed likeliness of each potential diagnosis.
Crucially, Diagu does not seek to replace human discretion altogether. AI can greatly reduce the rate of diagnostic errors, and it is especially adept at identifying and reconciling complex interactions of symptoms, test results, and patients’ medical histories. It does so, though, only within the limits of the presumptions that informed the development of its algorithms and the data on which it was trained. Most doctors have found themselves making informed diagnoses on the evidence available to them, only to find that information the patient deemed irrelevant is in fact hugely consequential. The same is true of AI-enhanced diagnostic tools, with one important difference: until they are re-trained on the newly discovered data, they will continue to make the same errors of omission. This further underscores the need to posit AI as a helper, guide, and tool for experienced diagnosticians, and not as a substitute for human oversight of the diagnostic process. AI models never know when they are acting on erroneous or incomplete data, and are capable of presenting their conclusions with an air of definitiveness based solely on the precision with which they analyzed the data available to them. Doctors are, like the rest of us, all too aware of our capacity for human weakness, and their resulting skepticism is an important hedge against AI’s limitations.
For that matter, a healthy skepticism toward the grander claims sometimes made for AI can help frame the relationship between doctor and machine in its most effective light. Like any tool, AI is as effective as the person using it. When medical personnel understand the strengths and limitations of the AI tools they consult, they receive the full benefit of those tools. Diagu Fire reflects both the promise of AI diagnostic technology and the need to employ it responsibly by giving doctors suggestions and ranges of possibility, not edicts or mandates. In this context, doctors are able to focus their attentions on the likeliest range of possible diagnoses while still informing their conclusions with the full benefit of their clinical experience. AI also helps diagnosticians see further into the future than older technologies currently allow, and to identify patterns in patients’ health histories that argue for preventive measures. For patients, the benefits are at once numerous and obvious: more efficient and responsive care, earlier starts on more targeted treatment plans, and better health outcomes. Hospitals, too, benefit from properly balanced, AI-assisted diagnostic methods. More accurate diagnoses allow for more effective treatment, reducing the costs associated with misdiagnosis.
The Benefits of Emerging AI Technologies
When used to its greatest potential advantage, as with Diagu’s solution, AI can convey a wide range of significant benefits to healthcare systems, hospitals, doctors, and patients. Some of these benefits are somewhat mundane: As cited in an article in Forbes on February 22, 2023, “From a financial perspective, wider adoption of AI could lead to savings of 5% to 10% in US healthcare spending-roughly $200 billion to $360 billion annually in 2019 dollars – within the next five years, according to a paper recently published by the National Bureau of Economic Research. . AI solutions are capable of streamlining intra-hospital communications, producing more efficient staff rotation schedules, facilitating patient intake, and automating much of the billing and collections process. As the backbone of natural language processing solutions, AI also holds great promise as a potential translator and facilitator of information within and across medical facilities. In the process, such applications of AI also promise to make medical care more efficient, responsive, and accurate.
The cost of healthcare services goes a considerable way toward determining patient satisfaction, especially when more efficient, lower-priced care is also more effective. AI diagnostic tools like Diagu Fire reduce the risk of diagnostic errors while supporting preventive health measures based on evidence unavailable until now to doctors and physicians. The result is a clearer and shorter path to better patient outcomes.
AI also holds the promise of more portable healthcare, further extending its benefits-especially to people in underserved areas of the country and to overworked hospital staff. Diagu’s GetLabTest system offers a glimpse of how AI might support an extended healthcare system. GLT allows patients to purchase single diagnostic tests or themed packages geared toward addressing specific health concerns or goals. When blood samples are needed, patients can visit a local clinic or arrange for a licensed phlebotomist to draw a sample right at home. Diagu’s AI models analyze the results as soon as they are ready, and immediately direct patients to video conferences with medical specialists appropriate to their diagnoses.
This approach removes many of the practical obstacles facing people across the country in light of hospital closures and consolidations. Telemedicine extends healthcare to people in regions underserved by medical facilities, and GetLabTest builds on that promise by shortening the distance among testing, diagnosis, and treatment. Like much of what AI seeks to improve, the Get Lab Test model could be replicated without the support of machine learning, but far more slowly and inefficiently. The speed with which medical conditions are diagnosed and treated is only one factor affecting health outcomes, but it can be a deciding one.
Improving healthcare access is an important public health goal. So is the kind of long-range forecasting made possible by AI. Systems like Diagu’s are capable of spotting emerging public health crises more quickly and accurately than traditional methods have allowed. As we have seen, AI can derive accurate conclusions from smaller groups of test data than some traditional diagnostic methods. It can also identify emerging health issues farther in advance than older techniques. Each of these advantages facilitates earlier detection and treatment of both chronic and emerging public health concerns, giving researchers, doctors, administrators, and public officials as much time as possible to respond to looming threats.
AI and the Role of Doctors
Like any technology, AI’s role is largely what its overseers-both technical personnel and hospital administrators-choose it to be.
On one hand, AI is able to deliver consistent, objective, evidence-based diagnoses based on volumes of data too vast for human practitioners to consider. Properly trained and implemented, AI can support real-time diagnoses of conditions that would otherwise take hours or even days to identify.
But these advantages only translate into improved patient outcomes when they support the judgment of qualified medical personnel and free doctors and nurses to provide the types of care that AI cannot. Not all patients are able to effectively communicate each of their symptoms when they first visit a clinic or hospital; an automated diagnostic tool will remain oblivious to the subtle hints such patients may send as to their real condition. A doctor, especially one who has been able to trade some time previously spent on administrative tasks for more time with patients, is far better positioned to gather all the information such patients intend to convey.
Again, Diagu’s approach offers a compelling vision of AI’s ability to facilitate established medical practice. Using HL7’s FHIR API, Diagu allows doctors to order tests, receive results from testing laboratories, and relay test results directly to doctors along with Diagu’s AI-generated predictive analysis. Patients who wish to begin the process themselves can use GetLabTest. After the lab processes their samples, patients receive SMS notifications and a PDF summarizing their test results. They also receive suggestions as to which specialists can best help them respond to any issues raised by their tests. Appropriate levels of information are made available to patients and doctors. For their part, doctors who meet with GetLabTest patients receive full documentation of each test’s results, along with Diagu’s AI-generated predictive analysis of the results, and a provisional diagnosis of the patient’s condition.
Diagu currently partners with hospitals, clinics and laboratories in the UK, Poland and other European countries. As part of its expansion, Diagu is actively seeking partnerships with laboratories and hospitals in the United States to train its AI models on anonymized datasets and pilot the use of its services such as Diagu Fire and GetLabTest. In addition, Diagu is working to integrate third-party products and platforms such as Workbeep and Gigbeep, which facilitate the work of phlebotomists in the UK. It will be interesting to see how Diagu adapts to the US market and what opportunities they discover in this new territory.
Early attempts to apply AI in hospitals and clinics were greeted with a combination of grand expectations and nervous apprehension. As it happens, each of those responses appears to have represented something of an overreaction. With time, the field has matured somewhat, and a truly revolutionary role for AI technology has emerged. Far from supplanting either doctors or existing testing methods, AI diagnostic tools have proven capable of significantly improving the performance of both. Most importantly, the advances made possible by AI improve patient outcomes by facilitating quicker, more accurate diagnoses and better-informed, more responsive treatment plans.
The new generation of collaborative AI is poised to unlock AI’s true promise and make good on some of its loftiest expectations. By combining the pattern-matching prowess of machine learning with the responsiveness and nuance only a doctor can provide, collaborative AI diagnosis represents a chance to see farther into patients’ futures than ever before. As we saw in the example from Massachusetts General, AI can diagnose some conditions just as accurately as human experts, more quickly and, crucially, with less testing data from a given patient. As companies like Diagu work to make lab testing easier, more accessible, and more efficient, AI diagnostic tools will increasingly be able to predict a patient’s future health concerns beyond the power of human doctors to do so. Doctors, for their part, will be able to develop courses of treatment that we cannot currently imagine.
This is a golden age for diagnostic medicine. Diagu invites you to meet that emerging opportunity with a suite of remarkable yet eminently practical tools for doctors and patients. We build solutions that allow the best of machine learning and human expertise-including yours. Schedule a demonstration of Diagu Fire or GetLabTest and see for yourself what the future holds in store for all of us.
The Future of AI Diagnostics is Here. Discover What it Means for You.
As AI health solutions continue to shape the future of healthcare, it is more important than ever to stay informed and engaged in the latest developments in the field. Diagu is ready to keep you up to date on the technologies and methods that shape our industry. Sign up today for regular updates on the latest news and analysis surrounding digital health solutions, and to learn how Diagu is poised to revolutionize diagnostic practice. When you do, a dedicated member of our team will personally reach out to you to discuss emerging solutions that are changing the way we diagnose disease and prepare treatment plans. Don’t miss out on this opportunity to be a part of the digital health revolution-register now on https://getlabtest.com/.
Contact: [email protected]
SOURCE: Diagu sp. z o.o.
View source version on accesswire.com: