SynPhNe (Synergistic Physio-Neuro Platform), a Digital Neurological Therapeutics is a start-up
company that specialises in wearable technology solutions and therapy devices. It highlights a
person’s unconscious muscle and brain responses which may be hampering recovery and
performance of everyday tasks. It is the world’s first medical device that personalizes in post-treatment therapy for faster and fuller
recovery of the patients suffering from any symptom of Stroke.
Abhijeet D. Pandit is the Director & CEO, VP Global Network Dev. & Marketing of SynPhNe. In an interaction with The Tech Pod, Abhijeet speaks about SynPhNe’s AI/ML strategy augments presently being developed. Read more!
How do you gather and refine all the data into useful information?
Over 13 million people will have a stroke this year and around 5.5 million people will die as a result. (World Stroke Org, 2020). The vision of building out the AI/ML strategy augments adherence in patients and supports them in achieving a state of independence.
The SynPhNe technology itself gives us information about what is happening “inside” the body and brain, while we watch what is happening “outside”, ie. visible movement, reactions, etc. Using the lens of a patient centered experience, we ensure that the core challenges are solved in unique ways only possible through ML. This is where Design thinking & Human centered design methods are utilized to identify & collate unmet needs of all the key stakeholders, i.e. patients, therapists and caregivers.
A combination of ethnography, contextual inquiry and empathy maps, journey diagramming was conducted to reframe the desirable goals and unmet needs, targetted towards patient adherence. Secondary research was employed to define the critical dimensions that can drive non-adherence – social or economic aspects, the condition itself, lifestyle and including feedback from caregivers & trainers, patient’s session data parameters like muscle and brain activity, sensor readings and video metadata.
Further filtering of data parameters is executed based on the AI-ML algorithms that best bridge the unstated core challenges of the patients.
Once we collect the raw data, we run it through a transformation process for cleansing, restructuring and aggregation that prepares the data for analysis. We then generate insights in the form of reports, visual graphs, and OLAP (Online Analytical processing).
How do you plan to make use of Big Data Analysis?
Due to the extensive scope and volume of data, it is essential to keep in mind the four V’s of big data: Volume, Velocity, veracity, & variety 
Big data in healthcare involves processing large data sets of patient profile attributes, care outcomes of treatments, condition types and their commercials. New data is being created every second. Due to its vastness, the speed at which the data is analyzed and used becomes critical. Further, a larger challenge is posed when the data itself is missing, incomplete or inaccurate and being structured or unstructured.
Ensuring the 4Vs are met, Big Data Analysis can be utilized for a variety of purposes such as –
- In research – Advancements in streamlining data collection and linkage, without compromising privacy; will not only lead to an exponential growth in the use of data for stroke outcomes research but will also fill gaps in research towards understanding other patient categories. This offers a promising opportunity for evaluations to improve decision making and evidence-based practice.
- Patient categorization – Improve treatment approach and recovery timelines by categorizing similar patients across demographics, behaviors and adherence conditions.
- To provide personalized care – Improve care personalization and efficiency with comprehensive patient profiles.
- To identify pattern and recommend treatment for patients – Provide identification of patterns in health outcomes, patient satisfaction
- To reduce recovery timelines – Improving care efficiency, effectiveness, and personalization
- To reduce healthcare costs – Make better use of technology to increase patient adherence and reduce no-shows for surgeries, procedures and treatments should be part of the solution to improve efficiency and reduce readmissions.
How is AI helping the advancement of Healthcare?
McKinsey evaluated the size of opportunity and determined the potential total annual value of AI and analytics across industries is about $9.5T – $15.4T.
Healthcare systems and services in itself is a huge market with a total potential annual value of up to $906.1B.
Some Predictive Service & Intervention use cases, AI is helping with in the Healthcare space are –
- Predict demand for healthcare services through predictive modelling
- Triage patient cases using patient historical data, reports and additional audio, video data
- Improve diagnosis and identify gaps in patient care
- Personalize approach, and treatment to improve wellness and adherence
- Use chat and voice bots to provide first-level self-care support and answer most frequently asked questions
- Predicting at-risk patients by identifying factors for non-adherence and providing recommendations of interventions to augment retention and maintain outcomes
What is artificial intelligence Neural Networks?
Simulating the behavior of neurons within the brain, neural nets are learning systems which utilizes several inputs to detect, recognize and finally classify patterns as an output predictor. What is interesting is that you don’t need to program it to learn explicitly. Neural nets are best used with large volumes of data sets, and training data to analyze and detect correlations.
Here are a few types of Neural nets:
ANN- Artificial Neural Networks (ANN) or Neural Networks(NN) are composed of elements that are inspired by biological nervous systems. A simple mathematical model to process information that is non-linear in nature.
A powerful subtype of Neural net, based on a study done by University of Texas, El Paso, results were conclusive based on a smaller dataset. This can be utilized to predict treatment and length of rehabilitation with a minimum forecast error. 
CNN – Convolutional neural networks (CNNs), which is a subtype of Deep neural networks (DNNs) involving image classification, is useful in acute stroke management and prognosis. Research is being conducted on prediction of Language Recovery after stroke with Convolutional Neural Networks. 
There are more Neural net classifications like RNN (Recurrent Neural Network), LSTM (Long Short-Term Memory) and many more. Novel approaches & further research is being undertaken in the field of stroke rehab to improve treatment approaches and reduce recovery timelines.
For example, study was conducted for defining rehabilitation treatment programs for stroke patients by applying Neural Network and Decision Trees models. Sensitivity, specificity and accuracy values were computed to define the performance of both algorithms. The results of this study indicated that both techniques can apply for data classification and define the proper treatment programs. However, the results showed that the specificity and accuracy of decision trees model were higher than neural network models.