Human Intelligence Archives - InnoHEALTH magazine

AI: Synthetic Intelligence with Organic Source

InnoHEALTH Magazine — Thu, 11 Jul 2019 09:23:32 +0000

Artificial intelligence: The know-how

Artificial Intelligence (AI) refers to “the intelligence generated through synthetic sources”. However, the irony here is, the source is a product of an organic (human)brain. Thus, it is a misnomer. Earlier it was speculated that AI progresses as per the evolutionary rate of human brain and if it happens otherwise, it can soon turn into an extinction tool. But today human evolution has reached a stage in which data processing technology stored in smartphones is far more superior than those used in sending the first man in the space. This miraculous journey of human evolution took nearly five to six decades.

The current century not only witnessed a revolution in the areas of biotechnology and genomics; it also produced enormous data (annotated sets). Since the last two decades, we have reached a complete shift from human-based to machine-based data analysis, filtering, classification, and interpretation. Deep learning algorithms are replacing 1940’s neural networks and providing data which was invisible to human perception earlier. Biologists now face this challenge of prediction analysis based on this in-depth data analysis.

Tools and Tactics

Some of these machine-based data analysis tools such as Cell Profiler have now evolved from individualistic feature-based characterisation to analysing multiple features like DNA staining, organelle texture and quality of empty cases-based cell sorting. In a similar way, Deep Variant has evolved from genomic information processing to image-based analysis. Google’s TensorFlow (open platform for deep learning algorithms); Atom wise (visualise molecules into 3D pixels referred to as voxels, provides atom- based interactions); answer ALS (a consortium-based approach to combine genomics, transcriptomics, epigenomics, proteomics, imaging and pluripotent stem cell population to target neuro degenerative diseases) and Contour (clustering of cellular imaging on the basis of trends rather than mortality factor alone) etc., are some remarkable newly evolving Deep Algorithm tools.

The over fitting of the model to its training data, which in turn is huge in size, poses a challenge to find ways to classify data to train it more efficiently. Computers have learned to find the needle (required information) in a haystack (huge database)which we as humans fail to identify and process.

During data analysis earlier, we were keeping experimental variations constant, but now with input tools like adding environment design, multiple controls etc., the perspective of data processing has changed completely. Present elaborate databases are analogous to our DNA in which most part looks repetitive and to know exactly why it is there, critical data screening and processing is required by new tools which further get improved through the acquired information and are better prepared for next generation of data.

Also Read: The Vulnerability of Medical Institutions to Cyber Attacks

Human intelligence is a complex simulation outcome of twelve characteristic features that include perception, resolving problems, learn, reason, abstract, plan,decide, understand, feel, act, create, and finally communicate. The human brain can be compared to a processor which processes these features using eight mental mechanisms which include imagery, concepts, rules, analogy, emotions, actions based on intentions, language, and consciousness. These twenty features (twelve characteristic features and eight mechanisms), taken together present the evaluating benchmarks for current AI. The major flaw in our current AI system compared to human intelligence is its limited ability in terms of abstracting and understanding with no feelings. However, AI is expanding its knowledge and processing in an exponential manner. Human-made AI now understands more than its earlier versions. AI is forecasted to pass through phases of knowledge engineering, machine-learning, and contextual marginalisation. It is like acting by rules, making new rules, learning to act again, to evolve new rules.

A simple AI tool such as a calculator can enable a person to use arithmetic for him/her. AI is now providing solutions in natural language processing, complex data sets for agriculture variables, etc.

However, unlike a calculator, it is yet to be accessible both in terms of tools, data set and skill for the public. Networking of multiple disciplines and protected commercial data with limited accessibility are major hurdles in the process. India is getting digitized in a fast manner. Adding Aadhar *unique identification number* was the major step followed by demonetization. As per CIS (India) of 2018, AI will add 957 billion USD by 2035. Health IT is booming with new start-ups every day, but the challenge is integration under streamlined vision. It is important to understand that any solution is sustainable with a participatory approach as exhibited by the evolution of telecommunication globally. Our basic education system is extremely diverse across the country with varied types of public and private players. The variation in state wise curriculum especially other than language is questionable. We can still choose Biology or Mathematics as separate options at our higher secondary education. This itself kicks out the possibility of emerging computational biologists in the future. It took almost ten years to build the world’s largest neuromorphic (neuron-like) computer Dubbed Spiking Neural Network Architecture (SpiNNaker) at the University of Manchester. This machine rethinks the way a conventional computer works. Its major objective is to support existing partial brain models of the cortex, basal ganglia, etc. It can manage 200 quadrillion tasks simultaneously. Still, we have reached only 1 percent of human brain capability that too with many simplified assumptions in the process. SpiNNaker is more comparable to mouse brain which is 1000x times smaller than the human brain. Although we need to explore more to understand the human brain and how it functions but once it is decoded any universal Turing machine can be turned to mimic human cortex actions. Intelligent and super-intelligent machines are our future and will make our life more convenient. Still, there are numerous challenges and ethical issues during these developments. We are moving towards a stage where time will become the ultimate currency governing all the aspects of human life.

About the author

Dr. Sarita Jaiswal, an ex-research officer at University of Saskatchewan, Canada, is an accomplished Plat Scientist having 15+ years of R&D experience with specialization in cereal and pulse crop biochemistry and genomics. She has been awarded twice for the category of Young Scientist (Indian Society of Plant Physiology and amp; KK Nanda Foundation for Advancement of Plant Sciences).

The post AI: Synthetic Intelligence with Organic Source appeared first on InnoHEALTH magazine.

Software as Medical Device

InnoHEALTH Magazine — Wed, 09 May 2018 05:23:22 +0000

The tectonic shifts in technology are transforming human life in ways unfathomable just a few years ago. Health-tech and med-tech are touching our lives continuously through a number of ways – from simple wearable devices to complex invasive devices; simple AI software which can predict and sense to complex AI software which can diagnose; sensors and other hardware devices including the mobile phone with ever-increasing computing power.

Some of these have made human lives so dependable on these devices, gadgets, software. In some cases, these are dumping human intelligence. We witnessed software wherein by looking at a camera on the mobile phone. The software can predict the heart rate and many other vitals. What if human intelligence gave way in believing the reading as true? The software or the camera is not a medical device. Hence outside the purview of regulations usually applicable to medical devices. Can we ignore the risks? If so, should software be treated as a medical device?

Software as Medical Device (SaMD)

A broadly accepted definition of a SaMD is the one issued by the International Medical Devices Regulation Forum (“IMDRF”); Currently, Australia, Brazil, Canada, China, Europe, Japan, Russia, Singapore and the USA are member countries to this Forum. The FDA has adopted this definition in the US. The Medical Device Directive adopted by the EU in 2010, and in major countries such as Australia, Canada and Japan.

The term “Software as a Medical Device” is defined as software intended to be used for one or more medical purposes that perform these purposes without being part of a hardware medical device. It includes an in vitro diagnostic medical device. SaMD must be capable of running on a computer platform that is not of a medical purpose. And should not need a hardware medical device to achieve its purpose. It can be interfaced or used in a combination with other devices. But cannot be used to drive a hardware device.

The definition, Mobile Application Meeting, also consider as SaMDs. The medical purpose‟ that it must intend to serve can be diagnostic, preventive, investigative, life-sustaining, for treatment of disease or injury, disinfection, control of conception or purely informative. In some jurisdictions, aids for persons with disabilities, devices for assisted reproduction and devices incorporating animal and/or human tissues are also recognized.

A SaMD can also be a means to suggest mitigation of a disease or provide aid to diagnosis. There are further guidelines on the definition of changes to SaMDs- they can be adaptive, corrective or preventive in nature. The manufacturer of SaMD would be a natural or legal person who has the intention for the software to be used under his/its name. It would not include a distributor or the manufacturer of an accessory. The final legal responsibility lies with the manufacturer unless it specifically imposes on another party by the country’s regulatory authority.

Putting them to use

SaMDs are now available in abundance in the Indian market. Both foreign manufacturers, as well as Indian manufacturers, are introducing so many forms of SaMDs. This includes the use of artificial intelligence, IoT, general software etc. Interesting, many SaMDs are enjoying high adoption rates not only for early users but continued users. The glaring point is that there is no legislative framework or guidance policy which works as a guiding principle for the SaMD manufacturers or at least as a self-regulating piece of legislation, in India.

The Medical Devices Rules, 2017 which has come into effect from January 1, 2018. Now defined medical devices and has made a clear distinction between drugs and medical devices. But still, this definition does not include SaMDs or software as a medical device. Interestingly, the draft Medical Devices Rules, 2016 on basis of which the Medical Devices Rules, 2017 have been formalized included software in the definition of medical device. AI, IoT, general software, wearable, and wellness and customized medical devices are flooding with the market. Software as a medical device as a whole should consider with equal importance in the sector. Curiously, the definition of medical devices under Foreign Direct Investment policy includes software.

Regulations

The IMDRF has worked extensively in setting guiding principles for governing SaMD and has put in place a regulatory structure for how the SaMDs shall be governed, regulated, clinically evaluated and how the data shall be evaluated and then used by the SaMD. While India has not yet included any software or apps in its regulatory purview. Countries like U.S.A, Singapore, Australia, EU and Japan has issued guidance documents to make the app developers aware of what might subject to regulation.

The common theme that determines the classification is the level of risk that these apps pose to the consumers. For example, let’s take an app which allows a user to take ECG test by putting their fingers on an external device which is connected wirelessly to the smartphone. It checks the electrical activity of the heart. Such apps may be considered as risky and be subject to regulation since the belief is that any incorrect analysis may hamper a user’s treatment. However, the Government authorities need to strike a balance while assessing these risks so that not all apps need to be certified under the law so that innovation is not hampered.

It is indeed a very fine balance. General wellness apps or products such as apps tracking and assisting in maintaining healthy body weight. Products are generally kept out of the purview of law versus apps which tracks and assists in say monitoring blood sugar or other vitals or treats specific health issues or provides guidance for treatment of specific illnesses. The whole purpose to bring these apps under regulation is that there is a certain amount of rigour before the apps are released to the market. And there is onus and responsibility on these makers. It should enable the app developers to be mindful of how is the product advertises, claims as well.

To protect consumers, certain jurisdictions, like Singapore, have mandated the manufacturers/ app developers to put a clarification statement on their product or on their apps. This statement should clearly state that this app or product is not intended to be used in a diagnosis, monitoring, management or treatment of any disease. Keeping all the innovation in health-tech space, India should provide guidance on SaMDs. The regulatory framework in India for medical devices is by the Central Drugs Standards Organization widely known as CDSCO.

The new Medical Devices Rules, 2017 are comprehensive and now extensively covers almost 351 medical devices and about 247 in-vitro medical devices. It still does not cover SaMDs. Given the increased use of mobile technology and awareness, guidelines on SaMDs could contribute to improving the affordability and availability of healthcare, including rural India, which has a huge user base. Gradual rigour in legislation will allow India to meet increased need, according to when resources for monitoring and enforcement become more available. India already follows the IMDRF regulations with respect to clinical trials. And the clinical evaluation of medical devices, with respect to documents, licensing and safety standards.

It is important for the legislation to allow the industry to grow and achieve its potential. Especially in a country like India where there is a need for the better point of care medical solutions. But at the same time provide unambiguous guidance. A good starting point would be a self-regulating mechanism with a set of standards, methods and procedures, clinical evaluation process. Such guidance would help improve innovation as well and guide the nascent Indian SaMD industry.

(Author is a qualified lawyer who has a keen interest in how tectonic shifts in technology is impacting healthcare delivery. The intersection of law, innovation, interaction with man and machine excites him. Manas works with NovoJuris Legal deeply in AI, IOT, health-tech, med-tech, devices and more in the healthcare segment)

A new paradigm for use of machine intelligence in healthcare

Want to write for InnoHEALTH? send us your article at magazine@innovatiocuris.com

The post Software as Medical Device appeared first on InnoHEALTH magazine.