Globally, senior population and the proportion with cognitive issues are increasing rapidly while healthcare/eldercare is lacking resources. The cost of Alzheimer’s and dementia can reach $1 trillion by 2018 (estimated in 2015), including healthcare, long-term care and hospice. Current methods for cognitive assessment are old-fashioned and not efficient. Only 25% of people with Alzheimer’s disease have been diagnosed. In most countries, there are actually those common challenges in this field:
- Only via the seniors’ medical visits, most assessments are currently performed in a time-consuming and expensive way by healthcare professionals, due to the purpose for detection and diagnosis of cognitive diseases.
- It is even not good enough for assessment data traditionally to be just collected as questionnaires or self-reports by using pen and paper.
- It is also similar in seniors’ welfare or wellness sectors how to do the assessment (by using checklists mostly now) for assisted living, skilled nursing, in-home care assistance, all the levels of elder care or potential care needs in a more quantified and evidence based way.
As our opportunity to bring new innovations, the markets are requiring digital-assessment and AI (Artificial Intelligence) more effectively to be front-line screening or interacting, as well as giving the seniors more time in good health and/or even a cure!
What to do as our unique know-how?
Cognition is how a person understands the world and acts in it, for tasks ranging from the simplest to the most complex. We assess cognition for elderly people such as learning, remembering, problem-solving, and paying attention. Our SIMO Cognitive-Assessment Product is an instrument of neurocognitive assessment with simulation-based software developed by Scandicode Oy (Ltd.) and installed to a multimedia device: such as Samsung Galaxy View tablet with or without a stylus pen (as an option for some of elder users to use due to dry finger skin).
SIMO product family can be aimed in long-term via three variants step by step continuously leading to be an AI (Artificial Intelligence) / ML (Machine Learning) enabled solution:
- Standalone b) clustered c) cloud-based
The 1st variant is in the scope of intended uses focused on a digitized version of neurocognitive assessment in comparable or contrast with traditional pen/paper-based method, such as Mini–Mental State Examination (MMSE) and The Consortium to Establish a Registry for Alzheimer’s Disease (CERAD).
Intended uses of SIMO product as medical device:
To briefly describe in short or long term, those are our intended target groups, medical indications, and contraindications: The particular technical characteristics of our product are intended to digitize, visualize and assess older people’s neurocognitive conditions. A deviation from normal i.e. an abnormal status of the user can be qualified, analysed and tracked with feasible repetitions of the assessment. During clinical evaluation, the product as a reusable tool must also show its excellent safety / usability to elderly end-users for neurocognitive status recognition in healthcare purposes. The tool can work as regular application comparable and/or together with other means. When applying AI / ML technologies fully in the future, such cognitive condition will be able revealed ideally as the data in clinical form, stage, severity, symptoms or aspects with the evidences from multiple sources (such as eldercare or elders’ home).
The legal and regulatory implications of applying medical or non-medical AI
As regulated businesses, the company should be mandated to follow an ever-increasing set of regulations. Scandicode Oy (Ltd.) designs, manufactures, and distributes medical devices and/or non-medical devices for solving ageing-population challenges in healthcare / eldercare / senior-wellness sectors. It is for our R&D not only guided as such (from MDD to MDR in EU as an example) towards a new era of AI, but also inspired as great opportunities if positively shaping them to work. Due to AI to capture various forms of personal data, data protection and cyber-security will also become very important to ensure sustainability of the technology (such as the General Data Protection Regulation applied from May 25, 2018).
AI Development Process:
In Scandicode Oy (Ltd.), the particular technical characteristics of our solution are explored at first as a MVP (Minimum Viable Product) demo version, and then productized by intended uses or customer needs accordingly to clinical-trial or market-release version. However, build-and-freeze model of CE or FDA process isn’t an easily comparable way for AI innovations, especially to unsupervised machine-learning (such as Alpha GO Zero) technologies. At Scandicode Oy (Ltd.), a hybrid supervised/unsupervised machine learning approach has been in the development by three projects under such an “agilean” way as follows:
- Based on the principles of design control for medical device, as well as popular agile or lean methodologies, own “3in1” KANBAN System has been created as research outcome of AI Business-Model Innovation.
- It is reasonable to leave the milestones away due to progresses of development work much faster and more dynamic in start-up environment. Or, it could also mean those virtual milestones are going through the process much faster and more often.
- All of the work with relevant deliverables (medical devices and/or non-medical devices) should be still kept, or checked interactively and dynamically, to ensure same quality required by market needs and regulatory guidance.
With such creative know-how, we are determined and convinced in developing AI-enabled medical devices and/or non-medical devices simultaneously to markets / end-users via such an “agilean” process, with all elements independently and easily reusable for multiple purposes.
Competitive Advantage: A hybrid supervised/unsupervised ML (Machine Learning) approach
- Supervised ML (S-ML) applying originally to medical device – data is labelled as class and/or value with direct feedback to predict outcome or future (to produce new class or value labels)
- Unsupervised ML (U-ML) applying intentionally to non-medical device – data is unlabeled or as unknown value without feedback for its outcome to find hidden structure (to determine data patterns or groupings).
- A hybrid approach as Semi-supervised and/or reinforcement ML with both task-driven S-ML and data-driven U-ML in our case to be a preferable combination.
- It means clinical trialled data purposely labelled to build decision process and reward system (the rewards act as spare or time-delayed labels) in the solution.
- New assessing or training data is unlabelled beforehand mainly for finding unrealized patterns or underlying structure of data, from series of actions and statistical properties (such as frequency) in or out of clinical environment.
- Such a hybrid approach can act as artificial neural-networks possible for deep learning algorithms, multiple processing layers, real-time optimal control or adjustment, dynamically performing parallel and series tasks – all integrated into our solution.
With our deep insight to AI technologies, open disruptive innovation is thus targeted to break the boundaries of medical devices and/or non-medical devices, via comprehensively refreshing overall-care of the elderly, still undoubtedly safe, effective, and fit for intended uses!
Scientific References for AI-enabled Solution Development
We are aiming to a digital reinvention for distinguishing between “normal” cognitive decline with increasing age and the decline preceding the onset of dementia.
Our Passion in Scandicode
The best cognitive assessment should be from AI-enabled solution to fight or repair age-related problems, not just current product as an interface of simulation testing. We will implement such a vision with healthcare or eldercare sectors, as well as the seniors’ wellness, creating win-win value together to solve their challenges. This is the mission of successful ageing as a unique perspective that will change our tomorrow!