AI as the Physician’s Assistant – Upping-their Game!

Each year, millions of Americans walk out of a doctor’s office with a misdiagnosis or missed-diagnosis. Physicians try to be systematic when identifying illness and disease, but they lack the appropriate medical devices.  

Now a group of researchers in Canada and Serbia have introduced artificial intelligence (AI) to remedy many of the all to familiar human shortcomings that lead to mis-diagnosis and missed-diagnosis in heartcare in Primary Care.

In Primary Care, heartcare is one of the biggest challenges faced by GP’s as they don’t have the specialized knowledge of the heart, making them mere gatekeepers to Cardiology. Yet, they are the first-responders for diagnosing heart disease onset – that patient both expect and rely on.  

The challenge is twofold, GP’s need to detect disease onset, and then to find a solution that would assist GP’s identify which patients need to see a cardiologist and which can simply undergo preventative or pharmacological treatments.

The role of AI to assist the GP in these challenges.

The current State-of-the-Art non-invasive diagnostic device now in use in Primary Care is the ECG (Electrocardiograph). Although GP’s learn some ECG in med-school, it takes years to gain effective practical experience. Although algorithmic augmentation has creeped into the newest devices, for example, the “Glasgow Algorithm”[1], and is assisting GP’s, the ECG itself has limitations that no amount of algorithmic augmentation can overcome – basically, ECG can only detect and diagnose heart diseases that have or leave an electrical signature. Physiological diseases are not included.

In this case, AI can only be as effective as the underlying technology it depends on.

Relative to all heart disease commonly prevalent in clinical practice, ECG’s effectiveness is limited to a miserly 44%. This ineffectiveness is not enough for GP’s in Primary Care for the widespread screening for heart disease of their patient population – it will lead to many mis-diagnosis or false positives and to missed-diagnosis or False positives.

We’re not saying ECG is bad, no, because what diseases ECG can detect, it detects very accurately and very specifically. A patient with an ECG detectable disease will likely be accurately detected. However, ECG was for years engineered for low sensitivity – to reduce FP’s, it was calibrated for use in Cardiology. This is not optimal for Primary Care as it sacrifices early onset detection for certainty of diagnosis. This is problematic for Primary Care as the first-responder for heart disease, creating a higher rate of FN, resulting in more patients returning by ambulance. 

So what is a Primary Care GP to do?

For AI to truly be a Physician Assistant in Primary Care it has to help with the detection of a wider range of heart diseases, dysfunctions and anomalies, and do it earlier. As a stop gap measure intended to compensate for ECG’s ineffectiveness in Primary Care, the Standard of Care (clinical workflows) requires that an Over-reading Cardiologist (ORC) must review all patient ECG’s. This adds costs and delays to routine examinations discouraging its use on asymptomatic patients.

Although AI can assist by complementing the ORC, ECG limitations, however, cannot be addressed through AI alone. The problem is in the electrical nature of ECG signal – it’s is only good diseases that have or leave and electrical signature, limiting ECG’s effectiveness to less than ½ of all common heart diseases. That means that over ½ of all common heart diseases cannot be detected in Primary Care.

Patients, simply don’t understand this limitation of ECG, putting themselves at greater risk. In this context, AI would not have been of assistance. So only if you’re lucky enough to have a disease that is detectable by ECG, AI can assist the GP, if not, then… the situation is deceiving.

For AI Assistance to be effective, the right bio-signal technology must exist that can provide the data on all of the most common disease conditions, not just less than ½.  

The most glaring omission from ECG signals is morphological and physiological heart information. Other than when that electrical signal is disturbed from its pre-set pathway, the ECG signal cannot identify with physiologically or morphologically natured diseases – those diseases that make up the other ½ of all common heart diseases. In this case, AI must be combined with technological innovations in the form of novel bio-signals to be of assistance.

However, an innovation designed for use in Primary Care is rare, because it’s not “sexy” and there is little money in Primary Care diagnostics. The big money is on the bigger ticket wonder-toys used in cardiology. The problem then becomes one of access, i.e. patients have to get into Cardiology to get access to the wonder-toys.

The access to Cardiology, however, is through Primary Care, or ambulance ride!

There is hope however as there is a startup that has for the last many years been developing a novel bio-signal whose nature is essentially physiological. When combined with ECG and powered by AI, it promises to detect well over 90% of all common heart diseases making it the ideal diagnostic device for use in Primary Care.

Compared to ECG’s 44%, this startup is going to reconfigure the Primary Care landscape.

Currently completing a retrospective clinical study in Serbia, this Canadian startup is aiming to have this device available by the fall of 2019. It’s a game-changing technology for Primary Care heartcare and extend a level of heartcare assistance to GP’s not otherwise available in Primary Care.

Called Cardio-HART™ or CHART, it will provide the GP with substantial assistance that would in essence, do the heavy lifting related to disease diagnosis. By applying AI to this innovation in bio-signal, the diagnostic learning curve is mitigated and the GP has the tool to help them do what they should – effectively diagnose the patient. The retrospective Clinical Study confirms that this the case.

AI assistance provides the GP with effective assistance at the point-of-care, where patients expect it.

In a secondary benefit, AI assistance can be just as beneficial for the ORC. At the halfway mark in the study, when comparing ECG only with CHART, ORC’s revised their diagnosis in 17% of the cases. They also expressed higher diagnostic confidence in 67% of their cases, because there was better medical justification. The only exception was in purely ECG detectable diseases, where there was no difference, as would be expected.   

In this study, GPs were paired such that one reviewed one of two tests, the ECG test results and the CHART test results for the same patient. At the ½ point in the study, the results show that the when relying only on ECG, the GP’s tended to refer a patient to the cardiologist 33% more often than when compared to the CHART reading physician for the same patient. They claimed that there were two reasons for these results. First, they were generally more unsure when viewing the ECG only results, and second, but related, they were more sure after seeing the patient's CHART results, because CHART provided more complete medical information leading to higher confidence in their diagnosis. These results were consistent for both FP and FN results, even when compared to those of the ORC results.

After Primary Care, each participant received an Echo examination to validate the initial diagnosis. This was then followed-up, by a consensus phase, when 3 independent cardiologists reviewed the ECHO to determine the ground truth. This ensured that the actual FP and FN could be truly identified. 

AI assistance to the GP when combined with the right bio-signal technology, is clearly leading to better patient outcomes in Primary Care.

As this prospective clinical study continues, it is clear that AI on its own is not a sufficiently effective assistant – AI needs to be combined with the right type of diagnostic technologies.

In this study, AI did not replace the GPs or the ORC but instead helped “up” their game. The assistance provided by the AI, especially through CHART, greatly improved the diagnostic outcomes for all patients but particularly in the small cities and towns where the most junior physicians were located. This front-line use of AI to provide GPs in Primary Care diagnostic assistance is clearly leading to better patient outcomes.

---

[1] The Glasgow Algorithm is an ECG algorithm developed by the University of Glasgow and is one of the best and most prevalent algorithms licensed for use in many ECG devices, since the mid-1980s.  

AI Powered Heartcare. Root-cause cost cutting in an intelligent but not artificial way.

Recently, research findings suggested that a meaningful reduction in variations of care is more effective than traditional cost-cutting. As an American president once said “its all about the economy stupid”. Any meaningful reduction in variations of care is really about focusing on the essential services the hospital needs to deliver in the community it serves, including the primary care community.

Heartcare represents a significant portion of any hospital budget. With heart disease responsible for more death, trauma, and morbidity than all the cancers combined, any meaningful reduction in variations of care must focus on the variations of heart care.

Yet a root-cause analysis of when and how heartcare costs are triggered reveals that they are triggered well before the patient arrives at the hospital, in primary care.

That means that an effective strategy for a meaningful reduction in hospital-related heartcare costs should really start in primary care. Until then patients will continue to be wheeled into the hospital through the ED doors, the most expensive doors to open in healthcare - not a red carpet but a carpet of red. 

But little has changed in primary care. It is still the same basic, archaic, system that “feeds” the hospital with the same basic centuries-old technologies.

A strategic new vision to cost-cutting is needed to address the issues at source. Two innovations can easily, and cost-effectively, enable that vision.

The first innovation should enable more direct collaboration between the hospital and primary care. Hospitals have the expertise that if applied at an earlier stage of disease onset, when it is first discovered in primary care, would lead to more effective and meaningful care, thereby reducing downstream, a.k.a. hospital, costs.

The earlier the onset of heart disease is detected, the more cost-effective the resulting treatment options and the more meaningful the cost savings. And patients remain in Primary Care. 

The bottleneck to this interaction is the technology between the hospital and the primary care clinics – the focus has to be on simpler diagnostic level collaboration.

Too often EHR interoperability is invoked as the panacea for collaboration when in fact it is really a red herring, a minefield of complexity, hidden costs and user problems – a shot-gun wedding is more fun.

This interaction should focus on a simple and effective solution that both fosters and enables “collaborative-triage”, where the hospital’s expertise is made available as part of the Standard of Care to help with patient triage at the primary point of care. The goal, more solid medical-justification as the basis of feeding patients to the hospital.  

This leads us to our second point. To enable effective collaborative-triage requires better Primary Care telemedicine-based diagnostic devices that can provide that provide more robust medical-justification. The right telemedicine diagnostic device is the key, it’s also the hard part.

Currently, ECG is the main “go-to” medical device for diagnosing heart disease in Primary Care (and this only with cardiologists over-read). Yet, despite its iconic status, ECG remains woefully inadequate and ineffective as a widespread screening device for use in Primary Care as the range of diseases it can diagnose is limited to 44% of common heart diseases[1]. This is a serious limitation “gap”, that patients are rarely aware of.

Currently, that gap can only be bridged when and if the patient has access to hospital-level heartcare services, which are not cheap and not readily available - you need a referral based on medical justification. A catch-22?

You need access to higher levels of heart care, to get the medical justification needed to get access to higher levels of heartcare! Yup, Catch-22!

To address this limitations gap requires advanced technological innovations that will both augment and complement ECG with new bio-signal technology powered by Artificial Intelligence (AI). 

Where are the innovators?

Conclusion

For hospitals, effective root-cause cost cutting starts with greater primary care collaboration. Novel innovations in AI-powered bio-signals will help bridge the diagnostic capabilities gap in current technologies enabling more meaningful and effective primary care level screening.

Improvements in Collaborative-triage innovations help bring together primary care and hospital level care that will help hospitals address root cause cost cutting whilst ironically providing more effective and meaningful care.

Cost reductions don’t have to mean care reductions.

---

[1] Common heart diseases are defined as all heart diseases that can be identified either by Echocardiography, ECG and are not rare. Rare diseases are defined as having a prevalence of less than 1:10,000+ and do not require specialized tests or specialized equipment to detect it.  

Not for Primary Care - The Promise of AI.

By Marc Torok,

Healthcare AI specialist.

Feb 13^th, 2019.

The RSNA 2018 Annual Meeting, demonstrated the rapid rise of AI and ML in Healthcare. Around the world, the application of artificial intelligence (AI) and machine learning (ML) in medical imaging is a hot topic. While research work will undoubtedly continue and expand, there are also growing examples of AI and ML being used in real-world, clinical settings. It also targets where it is not being applied – Primary Care!

Thus far, two main areas of focus have emerged as the most promising for AI and ML applicability, the first is healthcare information and second is medical imaging. Healthcare produces a wealth of disconnected patient-related information where AI and ML promise to find trends and patterns that could lead to better patient management and better clinical decisions. Whereas in medical imaging, AI and ML are leading to better detection of disease conditions in diagnostic images that promise to assist the specialists.

The progress has been rapid, as the goliaths of tech including Googles, Facebook, Amazon, Microsoft, IBM, … and a host of others, rush-in to capture the pole position in the new race into an emerging branch of value healthcare. As AI heads into clinical use, these early use cases are helping both technology vendors and healthcare providers understand how to integrate these systems into its clinical workflows. So too, healthcare institutions are learning how to value and monetize their data, the life-blood of AI in imaging and decision support systems.

And thus far, the promise of AI is all but absent from Primary Care.

The problem - the lack of first-use medical devices designed for use in Primary Care that can be AI augmented, or AI Powered. And the data needed to feed the AI beast. Only two current Primary Care devices are ripe for AI augmentation, the ECG and the stethoscope. Other heart disease-related devices, are not for use in Primary Care, such as the Echocardiograph, CT, MRI, which requires specialized knowledge and training.

The promise of AI in Primary Care should be to increase the devices effectiveness. For ECG this means AI would have to increase specificity, which, at roughly +97%, is already at the top of its game. AI can add little to this. However, ECG is essentially a bio-signal of an electrical nature and therefore limited to diagnosing diseases that only have or leave an electrical trace. ECG cannot identify with the morphological or acoustic aspects of the heart unless the ECG pathways are altered by a disease.

The other measure of ECG is based on sensitivity. Here AI has a modest possibility to increase sensitivity, which typically is between 60-70%, dependent on the device but yet again is limited to the diseases the ECG bio-signal can detect. However, ECG has already been algorithmically enhanced, which means that again AI would unlikely be able to deliver on its promise.

However, ECG’s greatest limitation is that it can only detect about 44% of all common heart diseases typically found in Primary Care. This is understandable as ECG bio-signal is limited by its electrical nature. This also disqualifies it for use as a first-use device for detecting heart disease in general, particularly in Primary Care. It’s simple, with ECG, too many diseases would remain hidden, undetectable. So although AI could make it more sensitive, it does not make more sensitive to a broader and wider range of common heart diseases typically found in Primary Care. AI would fail on its promise.

So current algorithmically augmented ECG devices, like the Glasgow Algorithm produced by the University of Glasgow, although not true AI, already do for ECG what AI promises.   but its effectiveness remains limited to the nature of the device. However, what it can detect, it detects very well – but this is not enough for use as a widespread screening device in Primary Care!

What is needed in Primary Care is an effective first-use device that can detect over 90% of all common heart diseases. 90% is specified because above 90%, the heart diseases become very rare, and very numerous, with very low prevalence in the population. At 90% effectiveness, such a device can be relied upon by Primary Care providers to enable a reliable systematic widespread screening of the at-risk patient population, both symptomatic and asymptomatic. However, even with AI or algorithmically augmented, ECG bio-signals are simply not enough as both the acoustic and morphological aspects of the heart cannot be diagnosed, leaving more than ½ of all common heart diseases undetectable in Primary Care.

New AI powered heart bio-signal enabled first-use devices are needed. Only then, will AI be able to assist Primary Care physicians with their diagnostic responsibilities as it is in Primary Care, where patients have the first contact with the healthcare system, and expect the detection of heart disease onset to be diagnosed first. The promise of AI, is to not betray that trust. 

Dr Google approved AI in Primary Care.

There is much excitement of the role of AI in diagnosing disease, especially as its capabilities are starting to equal and increasingly often surpass the experts, AI is intended to supplement and assist. But that excitement does not extend to AI in Primary Care, where AI is all but absent. Yet it is in Primary care where AI's potential could provide the greatest benefits to Patients, in part by augmenting and extending a clinician's abilities to diagnose.

Recently one of the first AI based device cleared by FDA for use in primary care, was peer criticized for its clinical limitations - it could only diagnose a single disease. Also, peers could not understand how a $18,000 piece of medical equipment that was also needed, that could diagnose a single eye disease condition, be “for-use” in Primary Care. The criticism also extended to the fact that the “autonomous” diagnosis related only to the possibility of this disease beyond its minor presence.

The peers understood that although effective for that one specific disease condition, the patient would be mis-lead into believing everything was “ok”, when they could in reality be suffering from the many other eye related diseases that were not ruled out, not supported by that device.

No matter how clear the disclaimers to that effect, patients simply will never understand that limitation – they want to believe the device has absolute powers!

Specialised single disease diagnostic devices are “dangerous” for use in Primary Care, whether AI based or not. They can be easily misunderstood and misused as they are too specialized, too limited and too expensive to effectively provide widespread screening benefits to potential patients. They don’t provide a comprehensive answer, as expected by patients.

And, in the age of “Dr. Google”, patients believe they are extremely health savvy, sometimes for the good, sometimes not! These internet doctors spread not only generalist health advice, but mainly also fear. It is this fear that drives patients to diagnostic medical devices to provide the medical “truth”. Single or limited first use devices in Primary Care don’t always measure up.

The best but little-known example of a limited diagnosis device which is widely used in Primary is the ECG! Surprised

Don’t be, you’re not alone. Few understand the limitations of ECG, which because of its electrical nature is seriously limited in functionality as it can only detect diseases that have an electrical signature. This means it is limited to about 44% of all common heart diseases, and this only when AI and algorithmically augmented.

This is why most Standards of Care, a cardiologist must “over-read” the ECG before the clinician can discuss it with the patient. Basically, the medical community is admitting that ECG is only effective on less than ½ of all common heart diseases.

No surprise therefore that is has not been adopted for the widespread screening of patients in Primary Care – able to detect only ~44% of all common heart disease, its simply not suitably effective – and that makes it dangerous!

It continues to be used because for the diseases it can detect, it does it really well. But mainly, it continues to be used for lack of a any better device for decades.

Now enter AI in Primary Care, and see a positive example:
Cardio-HART™ or CHART for short, is an AI powered, first-use, non-invasive, diagnostic system that together with the FDA cleared Cardio-TriTest™ device, provides clinicians with a better more complete understanding of a patient’s overall cardiac status - because it can detect and diagnose ~94% of all common heart diseases.

94% is “the” definition of clinical effectiveness, especially for Primary Care. 

Designed essentially for us in Primary Care, Chart outputs a broader range of disease findings and parameters than ECG. CHART provides 64 findings and 167 heart parameters, including all those provided ECG.

No other single diagnostic device in existence today can provide such a wide-ranging scope of cardiac status for use in Primary Care.

With 1 in 3 Americans at risk of heart disease and combined with an aging population, AI can supercharge Primary Care and provide patients with proper CHART.

Dr Google would approve.

Cybersecurity and Ecosystems of Medical Devices - Best approaches and strategies

Ensuring medical devices are safeguarded from cyber intrusions is a shared responsibility across the medical device ecosystem. As medical devices become more digitally interconnected and interoperable, they can improve the care patients receive and create efficiencies in the health care system. However, medical devices, like computer systems, can be vulnerable to security breaches, potentially impacting the safety and effectiveness of the device. Thankfully, manufacturers can adopt a holistic approach towards reducing cybersecurity risks associated with devices and of concern to patients by carefully considering — and building in — cybersecurity during design and development of medical devices, as well as having a robust postmarket plan to both manage emerging cyber vulnerabilities and to respond to intrusions or exploits affecting device performance when they occur.

Applying a best-teams approach

The FDA’s role and commitment to medical device cybersecurity continues to increase in scope and nature as they consider the implications of compromised devices across their total product lifecycle.
In the past 5 years they worked towards the vision of a healthy and resilient cyber ecosystem, that dream is now even closer, this month FDA has taken a ongoing relationship with the U.S. Department of Homeland Security (DHS) to another level. They announced a memorandum of agreement between the FDA and DHS, to implement a new framework for enhanced coordination and information of
cybersecurity vulnerabilities and threats. This could lead to more timely and effective responses to potential threats to patients safety and public health.

DHS serves as the central medical device vulnerability coordinating body and interface with appropriate stakeholders, including consulting with the FDA for technical and clinical expertise regarding medical devices. DHS' National Cybersecurity and Communications Integration Center will continue to coordinate and enable information sharing between medical device manufacturers, researchers and the FDA. The FD will continue to engage in regular, specific problems and emergency coordination calls with DHS and advise them regarding the risk to patient's health and potential for harm posed by the cybersecurity threats and vulnerabilities.

FDA is proactively addressing the risk to medical device in the face of an evolving cyberthreat with a release of the premarket cybersecurity guidance update. This was done to better protect devices from compromise; maintain device functionality in a safe mode even in the event of an attack and reduce potential risk to patients.


Although the FDA issued guidance providing recommendations for device cybersecurity information in premarket submissions in 2014, the rapidly evolving landscape, and the increased understanding of the threats and their potential mitigations necessitated an updated approach. They will hold a public Workshop in January 2019, to discuss the draft guidance “Content of Premarket Submissions for Management of Cybersecurity Medical Devices.”

---

Building strategic alliances

Creating an environment of shared responsibility means seeking out new engagements — formally and informally — with diverse stakeholders, including other government agencies, industry, healthcare delivery organizations, cybersecurity researchers and more. These relationships have provided FDA with insights into complex, device lifecycle challenges; they have also brought forth opportunities to leverage potential new tools and multi-pronged approaches to mitigate current gaps.

In the recent news MITRE, a non-profit corporation with support from the FDA, released a Medical Device Cybersecurity Regional Incident Preparedness and Response Playbook that can serve as a customizable tool for health care delivery organizations to aid in their preparedness and response activities for medical device cyber incidents.

Another example of FDA efforts to advance and support regulatory science for medical device cybersecurity is a participation in the Medical Device Innovation Consortium (MDIC). This non-profit, public-private partnership brings together industry, government, professional societies and advocacy organizations, to add value to the intersecting needs of the medical device industry, to promote the total product lifecycle of a medical device and to improve patient access to innovative products.

On October 1st, MDIC released a report on medical device cybersecurity and advancing coordinated vulnerability disclosure. This report advances an incredibly important topic in medical device cybersecurity  — the adoption of coordinated vulnerability disclosure policies and processes.
Knowing this information is critical, this way FDA can address the cybersecurity risk to medical devices in a timely and coordinated manner.

Another notable partnership is the FDA’s participation in the Healthcare and Public Health Sector Coordinating Council (HSCC). This HSCC task group will be releasing a Joint Security Plan this fall that describes best practices for implementing medical device cybersecurity and resilience recommendations, and further demonstrates the capabilities of medical device manufacturers working together with healthcare provider organizations to articulate a common vision to further safeguard patients.

---

FDA has showed a unwavering commitment to making cybersecurity a top priority for the agency, additional resources have been requested to continue building medical device cybersecurity program. In their Fiscal Year 2019 Budget, they proposed to create a Center of Excellence for Digital Health. This Center of Excellence would help establish more efficient regulatory paradigms, consider building new capacity to evaluate and recognize third-party certifiers, and support a cybersecurity unit to complement the advances in software-based devices.

The importance of establishing and maintaining a robust, collaborative framework for medical device security can be read in the report issued by the Office of Inspector General here.
The FDA has been and continues to work with the medical device industry and other stakeholders to proactively address emerging The FDA has implemented several of the recommendations in the OIG report. Like the evolving nature of the devices regulated — and cybersecurity threats faced — the FDA’s regulatory approach is not static.

This work will continue to drive advances in the increasingly complex medical device ecosystem enabling them as a collective to better anticipate cybersecurity risks and apply mitigation strategies early in the total product lifecycle of a device as well as with increased agility throughout the device lifespan as is necessary.

While National Cybersecurity Awareness Month has just concluded, Presidential Proclamation calls upon government and industry to work together, share information, build greater trust, and lead the national effort to protect and enhance the resilience of the Nation’s cyber infrastructure.

Source of this Article: FDA Medical Device Ecosystem and Cybersecurity