Artificial Intelligence, Deep Learning, and Medicine

• Read original article here

As physicians, nurses, dentists, or any healthcare expert, we all have experienced the earshot of floating buzzwords about the themes of Artificial intelligence (AI), machine learning (ML), and deep learning (DL). But not all of us are mindful of their potential consequences. On the contrary, yet generally speaking, most people, particularly the millennials, seem to be sparkly optimistic about the role of Artificial intelligent technology as being collectively encouraging.

Deep learning is a component of a much more comprehensive group of technology termed machine learning. DL defines the spectrum of artificial neural networks amidst imitation learning. Accurately why deep learning is also referred to as deep structured learning or differential programming, which can adopt a form of supervised, semi-supervised, or unsupervised modalities. Deep neural networks, deep belief networks, recurrent neural networks, and convolutional neural networks have been mainly applied to domains such as speech recognition, natural language processing, computer vision, audio recognition, social network filtering, machine translation, bioinformatics, drug design, medical image analysis, material inspection, and board game programs. Each and every component of the DL techniques have produced results analogous to human expertise, and even better.

In general, the concept of machine learning follows; that the contraption should be able to learn and adapt through experience and execute the tasks “smartly.”

Artificial Intelligence implements whatever learned by way of machine learning, deep learning, and other systems to solve substantive predicaments. In the computer science realm, artificial intelligence (AI), also referred to as machine intelligence, is nothing but machines’ capacity to demonstrate, what is typical for natural intellect exhibited by humans and animals.

With Artificial intelligence, today, one can perform an extraordinary spectrum of tasks. Using AI, one can ask questions by voice and get answers about a multitude of issues not stereotypically known to everyone. Or The computer can find data that could never come to a person’s mind. Artificial Intelligence, utilizing Deep Learning, will offer a narrative summary of someone’s data and suggest other ways to probe into collected information. Similarly, AI will furthermore distribute information narrated to earlier inquiries from others who asked the same questions. You’ll get the answers on a screen or directly through conversation.

The utility of artificial intelligence and Deep Neural Learning may seem potentially legit and promising, particularly concerning the extension of quality human life. Nonetheless, in realism, the messages portrayed are varied. Indeed, In health care, treatment efficacy can be determined instantly, whereas, in retail, inventories suggested quickly, or in finance, fraud prevented instead of just spotted. In each and every latter scenario, the computer efficiently recognizes what information is necessitated, looks at relationships between all the factors, forms an answer, and automatically communicates it to the users. It provides options for follow-up queries and even carries out additional pre-determined tasks with little human intervention yet even better.

Every AD, ML, DL technology relies on a set of finite sequences of explicit, computer-implementable instruction or algorithms, which frequently not disclosed to the public. Consequent to everything mentioned, the notion of Artificial Intelligence utility is a bittersweet experience, as the risk versus benefit of the technology lies within its particular algorithm.

Artificial intelligence delivers the promise of genuine human-to-machine interaction. It literally magnifies human potential with cumulative precision. The intelligent machines, over time, utilizing various machine learning techniques, can understand requests irrespective of a good deed or evil feat.

Artificial Intelligence help connect data points and draw conclusions irrespective of moral consequence, while they can learn to reason, observe, and plan.

All the advancements from Amazon Alexa to Apple Siri brought artificial intelligence closer to its original goal of creating intelligent machines, which we’re starting to see more and more in our everyday lives. From recommendations on our favorite retail sites to auto-generated photo tags on social media, many ordinary online amenities are powered by artificial intelligence. Further, we see thru advances in AI technologies, the more the privacy goes out the door, and the farther trivial turn out to be, our individual liberty.

Artificial intelligence is becoming a transformational force in the healthcare arena, as expected to disrupt healthcare in many ways.

Establishing a direct connection between technology and the human brain without the use of keyboards, mouse, and monitors is a state-of-the-art research theme that has abundant applications towards patient care. It will, for example, take up some of the responsibilities for kinds of functions that could be potentially taken away by some Neurological diseases and trauma to the nervous system. Or AI will be able to speak for the patient when impaired otherwise move his arm if paralyzed.

Radiological images captured by MRI machines, CT scanners, and x-rays offer non-invasive visibility into the inner workings of the human anatomy. Though several diagnostic processes still rely on direct tissue sampling or tissue biopsy to carry risks of infection and bleeding, AI will actually enable the next generation of radiology machines thorough enough to omit the need for diagnostic biopsy in selected instances.

Artificial intelligence is enabling “virtual biopsies” by advancing the innovative field of “radiomics.” The following science emphases on harnessing image-based algorithms to portray the phenotypes and genetic properties of tumors.

Shortages of qualified physicians, including radiology technicians and radiologists, can potentially curb admittance to life-saving care in developing communities around the globe.

Artificial intelligence could help alleviate the repercussions of a severe deficit of qualified clinical staff by taking over some of the responsibilities typically earmarked to humans.

Electronic health records are playing a more active part progressively in the healthcare industry’s drive towards documentation and The Health Information Technology for Economic and Clinical Health (HITECH). However, the transition to the digitalization of health records has faced innumerable problems, from cognitive overload, continual documentation, to physician burnout.

HITECH industry is now using AI and deep learning to create more spontaneous interfaces by automating some of the formal rules that occupy most of the physician’s time. Most likely than not, machine learning and AI may further support preparing conventional requests from the inbox, like medication refills and results from notifications. It may additionally assist in prioritizing tasks that truly require the clinician’s awareness.

Smart medical devices are filling up the user scene, allowing everything from real-time video from the inside of an intestine to sensing facial expression for early diagnosis of Autism.

In the medical setting, smart machines are decisive for monitoring patients across various spectrums of sceneries, from ICU to home care. Using AI physicians will benefit from enhanced ability to identify various pathologies deterioration, such as if sepsis is imminent, or detect the development of complications before it happens, hence significantly improving clinical outcomes and may reduce costs related to hospital-acquired condition forfeits.

Antibiotic resistance is a growing peril to populations around the ecosphere as the overuse of these essential medications fosters the evolution of certain strains of bacteria that fail to respond to future therapies.

Today, pathological specimens provide over 70% of the sources of diagnostic data for physicians across the spectrum of care delivery. And almost all the extracted data is widely available within the electronic health record systems. So the more precise we become, and the sooner we get to the right diagnosis, the better we’re going to be, making digital pathology, data and the AI an invaluable opportunity to deliver better medical care.

Deep learning algorithms and Artificial Intelligence analytics that can drill down to the minute precision on large digital images, thus allowing physicians to pinpoint subtleties that may skip the human eye. AI can further enhance productivity through the identification of features of concern in pathological preparations before human clinician studies the data.

Immunotherapy is one of the most astonishing achievements in cancer remedy. It teaches and uses the body’s own immune response to attack malignancies. Deep learning algorithms and their artificial intelligence upshots promote the synthesis of highly sophisticated datasets that formulates precise decisions for targeted therapies in the direction of individual cancer’s sole genetic structure.

Patient’s medical records are a goldmine of personal data, however extracting and analyzing such a wealth of information in a precise, timely, and consistent manner has been a continual challenge for physicians and data analysts.

Data quality and probity problems, as well as a mixture of data setups, makes the task complicated. Moreover, whether inputs are structured or not, along with incompleteness records, make understanding of exactly how to engage in meaningful risk stratification, predict analytics, and support clinical decision making extremely difficult.

EHR analytics have produced many thriving risk scoring and stratification tools. Yet, Amidst all, researchers apply DL methods to classify unique associations between seemingly irrelevant datasets.

With the increasing accessibility to wearable devices that use sensors to collect valuable consumer health data and transmit over smartphones, their utility is becoming more than ever inevitable. For instance, with step trackers, one can continuously track a heart pulse. In short, By implementing such technology, a growing portion of health-related data is generated on the go.

Collecting and analyzing medical information and supplementing it with data obtained from patients through apps and other home monitoring devices can contribute a matchless viewpoint into individual and population well-being. Therefore, Artificial intelligence can play a significant part in extracting actionable insights from this massive and endless treasure of data.

Harnessing the potential of portable devices, experts believe that images taken from smartphones and other consumer-grade sources will be an essential supplement to clinical quality imaging, particularly in underserved populations or developing nations.

The quality of cell phone cameras is growing by the year, as they can yield images that are viable for analysis utilizing artificial intelligence algorithms. Such technologies are very well known to modern Dermatology and ophthalmology.

British Researchers have even developed a means that identifies developmental anomalies by analyzing images of a child’s face in the womb.

As the healthcare industry is drifting away from fee-for-service reimbursement system, towards a merit-based compensation model, so, is it moving further and further from “reactive care” to treating the already manifesting disease to addressing the problem before symptoms appear, hence “proactive care.” Artificial intelligence will lay the grounds for that diagnostic revolution by powering predictive analytics and clinical judgment guide instruments that will alert physicians with obstacles long before they might otherwise recognize the need to tackle.

Deep learning carries valuable potential for real-world applications. Traditionally, machine learning described the training methods by which pictures used to train the program were tagged with the name of the thing in the picture.

The traditional machine learning scheme typically uses the photo and matches it with the “Tag” included within the image. The latter ML technique is referred to as “supervised learning.”

Do you recall tagging your photo or your friend’s photo with his or her name on Facebook posts? — That is how the ML would be able to learn a person’s face, identify it amid all others, and match it with other identifying factors on the internet for future authentication and identification.

Supervised learning is fast and demands comparatively less computational power than some other training techniques used in machine learning. However, It has a significant drawback for real-world applications. Every day, an immense amount of information about people is gathered from social media, hardware, and software service contracts, app authorizations, and website cookies.

The ML aided data mining or personal information collection is precious to businesses at all levels with a variety of agendas. The problem is that all of the said data is untagged and can’t be used to teach machine learning programs that depend on supervised learning. Because it still requires a person’s help to label or tag the data, which is not only time-consuming but also a costly process.

Deep learning networks can bypass traditional ML shortcomings because they utilize the so-called “unsupervised learning.” DL, do not require data labeling or tagging. Even though the pictures don’t come with the name “Tag,” Instead, Deep neural networks will still learn to identify the person.

The ability to learn from un-tagged or unorganized data is a tremendous advantage for those interested in real-world applications. Deep learning unlocks the treasure trove of big unstructured data for those with the imagination to use it.

21st-century’s millennial vision of physicians and healthcareis still about maintaining Hippocratical personalized medicine while sustaining the quality medical care using state of the art technology. Concomitantly the medical community is losing sanity by rapid putsch of sacred clinical judgment to a protocol based unyielding algorithmic patient care. The old fashioned population health model is one of the reasons to blame for such a course. But Artificial intelligence, if obtained throughtransparent and accountable methods, can lay the foundation for the personalized healthcare system. Deep learning technology can learn everything about a patient from birth onward and preserve in a decentralized fashion (Using Blockchain technology) without exposing personal information to alternate use. The data collected and held by the individual patient, physician, or any other user as the sole owner of their data will be able to take advantage of the unsupervised DL technology to help them take advantage of the kind of personalized care they want and need. Centralized big dataprocessing will only benefit other industries and further contaminate the already flawed population health model.

The population health principle is not receptive to a patient’s individual wants. Deep learning will learn every Individual need according to his or her expectations and needs, thus by way of AI, will advise the physician and patient alike, the best.

Modern society has been liberal in handling public digital information. But citizens are and will still pick up the consequences of such naivety of their attitude, yet, only the hard way. For instance, they will eventually figure out how valuable is what they are putting out there on harm’s way and how it is abused or used against them. Most of all, people will, in the end, realize- despite public reassurance by the giant social media and tech moguls, their data is not only a covert weapon against them but also is the digital cash they could put back in their pocket. Instead, personal data are indirectly being weaponized and laundered in the global scenery. Nevertheless, lets only hope it will not be too late; we all recognize those as mentioned earlier.

They say, on Facebook, users are becoming more and more discreet about who they share what kinds of data with, but with the use of Un-supervised Deep learning methods, even being discrete will be superseded unless people stop using the Facebook altogether.

Likewise, True, patients tend to trust their physicians more than they might believe in a big company like Facebook, which may help alleviate discomfort with contributing data to large-scale research initiatives but what good it will do if the data is centrally stored and the giant HITECH company is the sole holder of the “Big Data”?!

Although the subject of Artificial Intelligence replacing human jobs is a matter of great controversy, yet it is the least of all concerns. AI, indeed, will replace particular types of jobs in many industries, and not exclusively those jobs requiring predictable and repetitive tasks. Nevertheless, no doubt, disruption has already begun.

Ill-disposed use ofAI could threaten digital securityby various modalities is an imminent threat. Training machines to hack or socially engineered victims is a matter of great concern. Also, non-state actors weaponizing consumer drones, or privacy-eliminating surveillance, profiling, repression, automated, and targeted disinformation campaigns are a few of many fits of abuse we can face by refusing to see the trend.

Similarly, audio and video created by manipulating voices and likenesses. Deepfakes is already making waves. Using machine learning and deep learning will potentially involve natural language processing; an audio clip of any particular politician could be tainted to make it seem as if that person spurted racist views when in reality, they uttered nothing of the sort.

The widening socioeconomic disparity can be well thought using AI-driven job loss is a major concern. Along with education, work has long been also a driver of social mobility. However, when it’s a certain kind of work, the predictable, repetitive nature that’s prone to AI takeover research has given away that those who find themselves out in the cold are much less apt to get or seek retraining compared to those in higher-level positions which have more money.

We always need to remember; Artificial Intelligence is the product of human beings, and humans are innately influenced. AI researchers merely come from certain racial demographics, who grew up in high socioeconomic areas. Scientists are primarily people without disabilities from a fairly homogeneous population. Therefore, it’s difficult for those individuals to efficiently connect with the diversity of the society and their assorted concerns.

The root of all biases in the process from Data Mining to deep learning and ultimately, Artificial intelligence is socially and economically motivated. Because technology is the derivative of what humans design, hence making scientists and executives some of the most treacherous people in the world by way of the illusion of objectivity and greed.

Artificial Intelligence can be extra dangerous than bombs. The important is whether it is a good idea to start a global Artificial Intelligence arms race or to prevent its future proliferation. If any major military power acquires AI weapon development, a global arms race would be virtually foreseeable, and the endpoint of this technological trajectory is obvious, as autonomous weapons will become the guns of tomorrow.

Unlike nuclear weapons, AI requires no costly or hard-to-obtain raw materials. They will become ubiquitous and inexpensive for all significant military supremacies to mass-produce. It is going to only be a matter of time before the smart robotic weapons surface on the black market and in the hands of radicals, dictators wishing to better control their masses, tyrants wishing to commit ethnic cleansing, etc.

Self-governing armaments are flawless for jobs like subverting a country, committing murders, mollifying populations. A military AI arms war would not be propitious for humanity. There are many ways in which Artificial Intelligence can make battlegrounds safer for humans, especially civilians, without designing new tools to kill people. But then again, The US Military’s proposed budget for 2020 is $718 billion. Of the mentioned amount, nearly $1 billion would support AI and machine learning for things like logistics, intelligence analysis, and weaponry. AI can farther enhance the selective assassination of a certain ethnic group.

Historically corporations have enjoyed the munificence of personhood, collective influence of its stakeholder’s money, and technology. Today, using AI, corporate cartels are exercising the ability to read the human mind, access their personal information without breaking a single law. Nevertheless, the forfeits of people to the entire collective action of the industries are real.

It is the prevailing conception more so by the HITECH industry that machines will eventually replace physicians. Although this may be true, it is farther from wise. The indiscriminate utility of ML and AI is not only overwhelming to physicians’ practices but also influences the quality of care a patient receives from their provider. Building a technology that will utilize a prewritten algorithm through business intelligence or machine learning that is primarily designed to collect data from various sources is a growing and scary trend. Not just from a business perspective that it would be valuable just like a gold rush of our century but also from the quality and utilization perspective that is directly involved in the care of the patient.

Data industry, Big data, and more so, Health information has turned into a money-printing engine for every single industry. Health information lone has become trillions of dollars market. Software companies persuade citizens that the data is encrypted thus not accessible, even to their own employees. But AI has provided them with the capability to utilize public data any way they please. Parallel to advancements with Deep Learning technologies, the concept ofInternet Freedom and Net neutrality is becoming more and more obsolete.

The human being is fascinated by finding ways to teach the machine to express full empathy just like the human. The concept of Empathetic Transference is the ugly image of the human being on its way to satisfy the longtime battered ego.

The imbalance between strategy and tactical mission of many industries has turned out to be the upshot of tempting Big Data gold rush, pivoting the industry away from what their vision and mission originally conveyed. The latter has been further enhanced by advances made in Deep learning schemes.

Artificial Intelligence, big data, are presently used in artificial insemination, donor eggs, and genetic profilingwith particular pros and cons. The latter said it- is with particular reference to the science, technology, and cultural as well as ethical applications. When applied collectively, their impact on the constancy of social norms is exponentially deleterious. We’re simply entering a space where the upkeep of anonymity, respecting individual privacy as well as preventing major social, psychological, ethical, and legal dispute will be a strenuous undertaking. Unless fundamental solutions are realized with regards to AI and DL algorithms, the era of paternal anonymity will be soon coming to an end. Between the Sperm banks being forced to break rules of a confidentiality agreement with their donors, the growing Genetic Testing Market, along with lucrative corporate financial gain; the upkeep of donor confidentiality and offspring identity is fated to become an unrelenting task.

Compassion, sentiment, empathies are all significant parts of the healing processand medical treatments. But although you can teach computers to act empathic, it will always fall short of real human emotion, that is exactly why Artificial Intelligence will never replace the physician’s role. Nevertheless, it doesn’t override the fact that physicians should not adapt themselves to the perpetual changes happening around them.

Artificial intelligence is here and is most likely to stay. Physicians can choose to pull the blind eye on technological advancement, particularly over the DL algorithms, or they take ownership of their domain. If they select the former attitude, physicians will lose their job and sacred obligation to patients to industries and people who have little or no knowledge of patient care.

The physicians must reform the way they practice medicine. They must harmonize and direct the way they care for a patient on the right path using the most up to date tools that are validated and tested by physicians for the healthcare community.

It is alsocrucial to maintain transparency of the algorithmsif physicians and the medical community ought to ensure quality medical practice. Since trusting technology is nothing like trusting its designers, that sure stands true for Artificial Intelligence and machine learning.

Accountability is a must; however, to assert culpability proceeding Artificial intelligence, first proper transparency initiatives must be implemented. Most importantly, physicians must demand such transparency and mandate the accountability if they are not to be held liable themselves. The latter is the epitome of change expected of the physician community.

The legal community, particularly attornies, are facing similar challenges as the physicians; nonetheless, they seem to be effectively retaining the ownership of their artificial intelligence algorithms. All attornies have collectively established that the conflict of interestthreatening the legal system by way of financial benefit over admitting “non-lawyers” to own or invest in law firms. Technically speaking, the opponents of the latter rule are particularly concerned about the method of validation of their technology. Improper validation and oversight would give rise to more effortless undertaking certain legal activities by non-attorneys.

The Healthcare technology rush is the major factorbehind the disconnect between Physicians and their Domain and vice versa. Companies other than the health industry have alternate motives, hence mining for valuable patient and physician. The utility of Deep learning to snatch patient information has already begun. As mentioned earlier, Big data mining is vital to make available the vast pool of information required for robotic medicine and artificial intelligence. Besides, all the above is required to replace the human factor in the prospect.

Imperatively, Algorithms should deliver as they intended for tactical medical care, devoid of any strategic undertaking to pivot corporate interest to financial gain.

Physicians are the only ones who can warrant the adaptability of Deep learning algorithms to individual circumstances while designing them to act submissive to physicians versus acting as an independent provider.

The Medico-legal Perils of Artificial Intelligence and Deep Learning can be deleterious to physicians, if not recognized. So a valuable AI must identify the particular reference point for the standard of care on a precise scenario, time, place, and individual.

The physician, with the aid of patients, ought to redefine every case and have the legal, ethical, and methodical power to mutually override verdicts, by making a personalized approach. If failed to address patient problems within the medical standard of care spectrum, the treating physician will potentially face legal responsibility if something goes wrong.

Unfortunately, the physician’s profession relies on self-created habits. Doctors’ habitual practice has formed cultures and staff hiring practices that align with those personal habits. But routines need to change, something that will further disconnect physicians from the contemporary world, if not turned around.

Along the spectrum of attitude reform, physicians must embrace Artificial intelligence, deep learning technology, just like they did embrace stethoscope and X-ray during the past centuries. Doctors must understand its utility and perils. Only then can they adapt the best practice, maintain independence, ensure patient safety, and promote modern personalized healthcare.

This article was originally published by Data Driven Investor