By Rinat Akhmetov, product lead and ML solutions architect, Provectus.
The use of artificial intelligence (AI) is growing across all sectors, and healthcare is no exception. In fact, AI is particularly well-suited to healthcare applications due to the vast amount of data — from electronic health records (EHR) and clinical trials, to disease registries and claims — that is generated in the industry on a daily basis.
Ophthalmology is one area where the application of AI technology is more than justified. Faster and more accurate, at-scale eye screening can help diagnose and prevent such eye conditions as amblyopia, strabismus, diabetic retinopathy, glaucoma, age-related macular degeneration, and many others. AI holds the potential to improve patient diagnosis, reduce cost per screening, and expand the availability of eye screening to all.
This article explores how AI can be used in ophthalmology. We will consider the benefits and challenges of AI, outline prospective use cases, and offer a framework for adopting AI.
Ophthalmology is ready for AI innovation
Artificial intelligence is beginning to be used in ophthalmology for a reason.
A 2020 study researching the use of AI to screen for diabetic retinopathy, a leading cause of blindness, found that AI was able to achieve an accuracy of around 95%, which is comparable to that of expert human graders. Another study used AI to detect glaucoma, also a leading cause of blindness. The AI system was able to achieve an accuracy of over 90% in detecting the disease.
These studies show that the amount of real-world data is enough to develop highly accurate algorithms that can detect disease as well or even better than humans — in all types of eye screens, and at a speed and scale that exceed human potential many times over.
Given the global shortage of ophthalmologists and optometrists, and the widespread availability of technology (from ready-to-use algorithms to cloud computing), introducing AI to augment the work of ophthalmologists seems like a sensible solution.
Despite benefits, AI remains a challenge
The potential benefits of using AI in ophthalmology are significant. The improved accuracy and scale of disease detection lead to earlier diagnosis and treatment, which improves patient outcomes. Automated disease screening frees up time for ophthalmologists to focus on other tasks.
However, there are also some challenges associated with using AI.
AI requires high-quality data for training. And while the volume of data is usually not a problem, finding the right talent to prepare it can be problematic. Only professional ophthalmologists are qualified to label training data in a manner that ensures high accuracy on real-world data in production.
There are risks of false positives or false negatives. Some diseases may be incorrectly diagnosed, while others may be missed altogether. Hence, the importance of prepped data, an infrastructure for AI monitoring and re-training, and human-in-the-loop (HITL) for processing user feedback.
Thankfully, AI technologies are developing so quickly that it becomes easier for practitioners to build eye screening applications from scratch, using open-source tools and cloud services from AWS, Google, or Microsoft.
Practical applications of AI in ophthalmology
There are a number of ways in which AI can be used for disease screening in ophthalmology.
One example is fundus photography, which is a type of medical imaging that captures an image of the back of the eye. For instance, AI can help capture and interpret the retinal vasculature, to determine risk or presence of diabetes. Similarly, AI can preemptively reveal pathologies that cause blindness and vision loss by enabling at-scale screening for fundus and retina abnormalities at birth.
Another example is the use of Optical Coherence Tomography (OCT). This is a non-invasive imaging technique that uses light waves to take pictures of the retina. These pictures are processed and analyzed by AI to detect any signs of anomalies associated with disease.
AI can also be used to augment photoscreening applications. GoCheck Kids, a company assisting primary care networks, implements cost-effective pediatric vision screening, and utilizes AI to supplement image analysis and improve user actions, to help ophthalmologists capture the best image possible for further analysis.
The paradigm for AI adoption in ophthalmology
The power of AI lies in its ability to identify patterns and anomalies in data that may be difficult for humans to spot. Nowhere is this more apparent than in the field of ophthalmology, where AI is used for disease screening — detecting anomalous parts of eye screens that may indicate a specific eye condition.
For AI in ophthalmology to work effectively, however, certain conditions must be met.
Any disease screening system or application has to have an image labeling component. AI is a work in progress, a system that evolves over time on new data, and users should be able to label new screens and verify low accuracy screens that were previously taken.
End-to-end infrastructure for AI has to be in place so that models can be built, trained, deployed, monitored, re-trained, and fine-tuned. Any types of data or model drift, or bias, should be monitored and countered by cyclic model updates.
It is better for the solution to live in the cloud. It helps realize such benefits as automatic scalability, high flexibility, and reduced IT costs. It also ensures collaboration efficiency and business continuity. For instance, an eye screen taken with an app by an optometrist in Chicago can be labeled by a highly trained ophthalmologist in LA, with both of them contributing to the improvement of the application’s AI.
Having the right UI matters. Doctors taking eye screens should have access to a section that explains why AI made certain decisions, to better understand the signs of detected abnormalities. The labelers should be able to sort existing screens, and markup and feed new screens to the system. A customer-centric UI ensures that doctors do not have to spend time examining screens with no signs of pathology, so they can focus on patients who need assistance.
Conclusion
The potential of AI in healthcare is immense. From streamlining administrative tasks to providing insights for clinical decision-making, AI can help to improve patient outcomes, increase productivity and efficiency of care delivery, and make it easier for wider categories of the population to access healthcare services.
In ophthalmology, AI-powered disease screening is the future. By automating pattern identification, AI can help to increase accuracy while saving time. It can identify individuals who are at risk of developing, or who already have a certain disease, as well or better than human doctors.
It is estimated that by 2050, over 1.8 billion people will suffer from some form of vision impairment. This number could be reduced drastically if preventable vision loss was detected and treated early on. The way forward is to scale disease screening with AI, to enable doctors to focus on patient care while leaving routine work to the machine.
Artificial intelligence (AI) is a driving force in the future of technology, often enhancing the speed, precision, accuracy and effectiveness of human efforts. As a result, AI has had a tremendous impact on nearly every working industry, including healthcare and its specialties. In recent years, there has been an increased adoption rate of AI in the healthcare industry.
Although this growth may be related to the need for telehealth and other remote tools as a result of the COVID-19 pandemic, recent estimates predict a more than tenfold growth in the market for AI in medical imaging over the next decade.
Although there is a projected growth for AI in healthcare, if this technology cannot be efficiently implemented into existing daily workflows, then these AI tools will not be practical in a real-life setting. Understanding how healthcare practitioners use AI in a clinical setting and how AI can help solve real-life challenges are crucial to increasing further adoption rates. This will ultimately drive innovation around this technology and will lead to improved quality of patient care.
What clinicians want from AI-powered technology
A recent survey by the American College of Radiology (ACR) indicated that 30% of respondents use AI to enhance image interpretations across all modalities. The modalities most commonly identified were computed tomography (CT) scans and mammography scans. When asked what they would specifically like AI technologies to do to enhance their clinical practices, respondents indicated that lesion detection (73%) and anatomic measurements (71%) were most important.
These responses indicate that clinicians are most interested in reducing the need for manual tracing and measurements of medical imaging, which takes a significant amount of time and effort. The survey’s findings also indicate that respondents would prefer additional support, or a “second opinion”, when detecting and identifying lesions – a task known to be difficult. The clinicians’ responses also indicated that the development of a method to evaluate an AI algorithm within the workplace setting before purchase is of utmost importance.
With these findings in mind, software developers need to prioritize building technology to help provide a practical solution to clinicians’ most important needs. The use of artificial intelligence may be the best way to ensure clinicians’ needs are met in an effective and practical manner. With the use of AI, the software can continuously learn from collective insights of multiple experts, effectively offering practitioners thousands of “second opinions.”
This type of technology can also help to provide interpretations and analyses of medical imaging, significantly reducing the amount of time it takes for clinicians to measure and trace these images. With less time and effort spent on interpretation and analysis, clinicians are able to spend more time with patient facing tasks – ultimately leading to higher quality care for a lower cost while reducing employee burnout.
Although these solutions exist, there are still several barriers that prevent the successful implementation of AI in clinical practices from occurring. For example, clinicians want to ensure that the AI is safe, effective and solves specific needs before the technology is purchased. However, of those providers who currently use AI in their practice, most were satisfied with their overall experience and found that AI provided value to them and their patients. Therefore, it seems that education about the potential benefits of AI in all practice types will continue to be important (Allen et al, 2021).
The healthcare sector has witnessed many transformations and investments over the years, and you can only expect this trend to continue, especially after the difficult lessons from the coronavirus pandemic. Amid the global health crises, the healthcare sector responded by adopting various techniques both in practice and technology. If these encouraging signs are anything to go by, 2022 may usher the American healthcare sector into a new era of healthcare. That said, here are some key trends that are currently transforming the healthcare sector in 2022 and beyond.
One of the first trends you can expect to see more of is the use of artificial technology (AI) in healthcare delivery and general medical practice. While the use of AI isn’t necessarily a novelty in the healthcare sector, you can expect to see it take more center stage in the way healthcare is delivered. One of the main results of AI’s impact on healthcare is the availability of various mobile apps and websites that help people to self-diagnose right in their homes.
AI is now also being used in the form of augmented intelligence to enhance the intelligence of clinicians and improve the quality of service they deliver. AI also plays a crucial role in the use of medical technology like hearing aids. For example, it’s used in a hearing aid evaluation to help predict the best possible hearing or sound settings based on data obtained from the user.
Another huge trend taking the US healthcare sector by storm is telehealth services. You probably already know how popular video calling apps have become since the pandemic. Now the healthcare sector is taking advantage of telehealth to deliver healthcare services remotely. Patients who cannot visit their hospitals physically can now see their doctors literally while at home.
The healthcare industry is under intense pressure to improve its efficiency. However, interoperability between technology and various integrated systems presents many challenges that are hindering health facilities from being fully connected and productive.
We have known for years that healthcare needs solutions that artificial intelligence can provide. But the initial proofs of concept have taken too long to materialize. Without clear boundaries and use cases showing how AI in healthcare can work, leadership teams are unable to horizontally collaborate with each other.
How AI in Healthcare Could Solve Interoperability Problems
Technology has the potential to transform the way healthcare works for patients, but right now, interoperability is difficult to attain. Despite industry guides such as the Fast Healthcare Interoperability Resources, data is still a messy business. Data is stored in different ways and in different silos — and not every facility has the ability to read and understand the information contained within the respective silos and make it actionable.
This has a heavy impact on how practitioners work with technology. A radiologist reading film and a doctor making a diagnosis for a chronic pain patient only have access to their siloed expertise. With AI solutions in healthcare, data can be drawn from different disciplines and diagnosis can become faster and smarter.
When used in conjunction with AI, blockchain technology has the power to help practitioners and organizations work together without security risks. Because the blockchain represents a transparent, single source of information that cannot be changed, it can store data from multiple sources and create a harmonized picture of truth that different users can access without bias. In addition, limits can be put in place as to who has access to the data.
This helps healthcare experts form a central hub where the very best knowledge, therapies, and drug research can be pooled, therefore helping target diseases more effectively while keeping patient and research data absolutely secure and private.
It’s clear that leaders at healthcare organizations need to remove the siloed approach and develop an atmosphere of increased collaboration. But how, exactly?
How Blockchain, AI, and Healthcare Can Work Together
Blockchain technology in healthcare helps fulfill all four kinds of interoperability defined by the Healthcare Information and Management Systems Society: foundational, structural, semantic, and organizational. Blockchain’s uses in healthcare create a basis — a structure — where data can live safely and transparently. Then, blockchain can enable a rendering that helps different kinds of readers see and understand the data.
Two aspects of blockchain technology that are especially interesting to the healthcare industry are permissioned blockchains and smart contracts. A permissioned blockchain maintains the privacy of data, knows all the stakeholders, and makes data viewable by actors on the network who are authorized to see it. Smart contracts are “instructions” on the blockchain that are executed automatically once all necessary conditions or events are met. This means decisions can be made available automatically without human intervention. That’s where the power of AI’s uses in healthcare really materialize. This harmonized dataset — coupled with safe and secure automation — means that AI can be used to make faster, better, and more predictive decisions.
Data is the engine behind AI, but it’s also becoming the engine behind healthcare systems and how doctors diagnose and treat patients. If we can aggregate and translate vast amounts of data into streamlined workflows, AI can be used to efficiently diagnose and monitor patients, detect illness, accelerate drug development, and seamlessly run clinical trials.
The ingredients for interoperability are all there, but it’s now up to operators and developers to find ways to work together. The benefits of AI in healthcare are massively transformative — as long as we can find ways to solve problematic perceptions of blockchain and data privacy and get human beings to open up their silos.
No one technology will save the future of healthcare interoperability. It will take collaboration between developers, operators, academics, drug researchers, and an interwoven stack of technologies to bring together a universe of data and put it to good use.
By Joachim Roski, Principal, Booz Allen Hamilton and Kevin Vigilante, Executive Vice President and Chief Medical Officer, Booz Allen Hamilton.
As artificial intelligence (AI) continues to transform our world, the healthcare sector stands to substantially gain from AI. The ability to compute a massive amount of information quickly has promising implications for delivering better health outcomes, improved healthcare operations, or expanded research capabilities. However, AI is not immune to the risks that accompany the adoption of any new technological advancement – risks that may cause healthcare professionals to doubt its reliability or trustworthiness.
Fortunately, there are steps available that can help build a strong AI foundation, improving the quality of life for patients and healthcare professionals alike. If you’re looking to implement AI in your healthcare organization, here are a few things to keep in mind:
Understand your Data
If a data is not of high quality or is not representative of the intended group of study, the conclusions based on that data will likely be flawed. To accurately anticipate how a flu outbreak will impact hospital visits among a community population, it’s essential to account for variables that can affect all or part of the relevant population. For example, if a significant subset of the population is more susceptible to serious flu symptoms (e.g., based on pre-existing conditions), this needs to be taken into account by the hospital(s) in question. Given that algorithmic models are often designed to grow and expand with compounding data sets, any previous mistakes can snowball and lead to significant long-term bias or inaccuracy in the results.
As such, it’s critically important to build an AI foundation on a robust and comprehensive data acquisition and management operating system, complete with careful and consistent oversight of AI algorithms. Such a system should be supported by compliance and monitoring protocols to ensure that data is securely flowing. Having a clear understanding of where relevant data comes from and how it was collected is critical to ensure AI algorithms create meaningful output.
The main areas of use of biotechnology today are medicine, pharmaceuticals, agriculture, and other industries, where BioTech innovations can reduce the cost of production, speed up the development of vaccines, or model the changes of genomel. But, the development in this area is very expensive, and the necessary research requires a lot of time, and all kind of resources.
The cost of BioTech solutions is also influenced by the availability of technological and human resources. Research centers, special devices, and highly qualified specialists are needed for the development, testing, and implementation of advances in biotechnology and healthcare. The more there are, the lower the cost of producing medtech products due to the ability to scale development. At the same time, it is necessary to demo the statistically representative data, which demands the repetition of experiments and very careful documentation of all the experiments’ work circles, which create additional costs.
It is necessary to increase capacity in research centers through the purchase of modern and productive equipment to reduce the cost of technology. Also, other ways are needed to modernize production facilities, which will help to speed up the Design-Build-Test-Learn cycle. It should be borne in mind that the equipment itself will not help to process and structure huge amounts of data. As one of the way to optimization, artificial intelligence integration can work nowadays. They take the burden off specialists, dismissing the routine work, and in some ways can structurize the documentation flow and research data storage.
Examples of the use of AI and ML in biotechnology
Scientists from Russia and Belarus have developed a special substrate for efficient stem cell growth. This substrate can be used to develop new materials and technologies in the field of regenerative medicine.
According to the representative of the research group, such substrates can be created based on bacterial cellulose, which is produced by the bacteria Acetobacteraceae bacteria, and which is modified with cerium oxide nanoparticles, which give it unique bioactivity and provide accelerated division of stem cells on its surface, which is the main tool of regenerative medicine.
These cells can self-renew, divide through mitosis and differentiate into specialized cells, that is, turn into cells of various organs and tissues. The researchers used special fluorescent stem cells from transgenic mice with a mutation that gives a green glow to all cells in the body. This made it possible to visualize stem cells on a substrate and analyze the process of their accelerated division.
By Amit Garg, vice president of analytics, Gramener.
Amit Garg
The advent of the pandemic and more recent vaccination efforts means there’s never been a time where more people’s Personal Identifiable Information (PII) and Personal Health Information (PHI) in health records and medical documentation are in circulation. It’s vital to protect patients’ PII and PHI, which can include information on their age, race, or medical history, especially given that cyberthreats and fraudulent activity related to the Covid-19 vaccine are increasing.
With so much patient and clinical trial data being stored and shared at any given time, it’s becoming increasingly challenging for pharmaceutical companies to efficiently ensure that patient information is protected. Experts say that medical data is up to 50 times more valuable than credit card data.
Here’s where artificial intelligence comes into play: AI and machine learning (ML) solutions can not only automatically identify what information is classed as PII in a given record, but it can also then automatically redact or anonymize that data to make sure that no adversary can identify the patients.
Joining the dots with AI
AI algorithms use advanced methods such as entity detection, entity extraction, and entity-relationship management to handle patients’ PII and PHI from a given document. This involves identifying and categorizing key information in the text using Named Entity Recognition (NER), a form of Natural Language Processing (NLP).
Useful libraries for teams using NER include Stanford NER, spaCy’s EntityRecognizer, CliNER (a domain-specific NER tool that has been trained on clinical texts), or BioBERT (a domain-specific language representation model pre-trained on large-scale biomedical corpora).
NER works to safeguard PII in the context of healthcare by identifying different elements of a single patient’s PII across multiple health records. In one document the name and age may be present, and in another, age and race may be present but not the name, while in another religion and race, and so on.
Any hacker with access to each document and each data point would be able to join the dots to match all PII to the single person. To prevent this, an intelligent entity detection and extraction solution can identify this information across the documents and redact and anonymize the correct data to prevent reidentification of the patient.
If the solution is not able to de-identify (or redact) the information completely, it will score the subjects based on the probability of re-identification. The publication would then know the risk in advance and take the appropriate action, i.e., to assume the risk and publish or to not publish until specific confidence is reached.
Machine learning is possibly the most disruptive technology of the 21st century. Good machine learning systems will, with limited interaction from a doctor, will be able to analyze almost any kind of information so that doctors can make better decisions.
But since a machine learning system is trained on data from humans, it reflects their internal bias, and can magnify them. Hospitals need to focus not only on what advantages these systems are bringing but also the biases that they encourage. WebText, a solution used to train natural language processing to analyze new articles and documents, was trained on posts from Reddit.
Reddit is almost 70% men with more than half of the users are from the US. The majority of the users are under the age of 35. These biases in the data create machine learning solutions that reflect these biases. One criticism, for instance, when asked “A man is a doctor, as a woman is, too” responded with “nurse.”
AI systems used in hospitals have already shown to be able to do incredible things from being able to diagnose disease from a simple, but also accomplish human mistakes at the speed of a machine; for example, an AI system rejected black medical students because data it was trained on was principally white students.
To prevent these effects companies need to carefully monitor these new artificial employees and make sure they are meeting the standard of governance reflecting the values of the company and the law. This can only be accomplished through specific tools that allow you to enter the mind of these artificial employees and understand how they think.
By 
