Delivering a new standard in speech and language development

We want to help parents understand the needs of their child, in terms of speech and language development, as early as possible. 

At the moment it's hard for parents, at home, to understand if their child needs help with their speech and language development, and there are no clinically validated tools that parents can use, at home, to help them help their child in the early stages of development.

We are using data and AI to build an assessment technology that can let parents and other stakeholders know if a child has the correct speech skills for their age. Our voice enabled games can facilitate personalised interventions as early as possible to improve the chances of staying out of the clinical care pathway. 

Our goal is to ensure that no child starts school with a communications disadvantage

Families are not being adequately served when it comes to children's speech and language development. There is a huge variance in prevalence rates for speech and language delays in children, these can range from 5-8% in some studies up to 20%. 

Parents are at a loss about whether their child is on track and often take a "wait and see" approach to speech development, which can be a problem; It takes four times as long to intervene, for learning difficulties, in fourth grade as it does in late kindergarten. 

Part of the problem is that parents have no way of benchmarking their child's speech development, they ask their physician, pre-school teacher, parenting forums, but realistically the only way to get information is to attend a Speech Pathologist. 

There is lack of data on the equality or quality of services to different children, however Morgan et al, found "the likelihood that Black children would receive these (SLP) services was about 45%–60% lower than for otherwise similar White children". The industry itself is struggling to meet demands, in 2017, 35% of SLPs responding to the ASHA SLP Health Care Survey reported that job openings exceeded job seekers.

We are starting with parents, the CDC national children's health survey and our own research found that over 20-40% of parents have concerns with their child's speech development. Parents are aware of how critical this developmental phase is for their child, particularly when they see reports that children with poor vocabulary skills at age 5 are four times more likely to have reading difficulties, three times as likely to have mental health problems, and twice as likely to be unemployed when they reached adulthood (Law). 

We have spoken to parents across several countries where we identified the unmet need being in this pre-clinical phase. We have a group of testers from 4 different countries that help us develop our applications. After speaking to pre-school teachers, we know that parents are asking them to evaluate their child's speech skills but they don't feel qualified to make these evaluations so they normally just tell the parent to go to a Speech Pathologist. Our solution will allow parents and primary care professionals to make informed decisions about a child's speech development needs. 

Our solution is to use speech data and machine learning algorithms to assess a child's speech skills from speech samples collected through voice enabled games. We are working with a company that has developed a speech recognition engine specifically for children, we have developed and tested machine learning algorithms on the output of this speech recognition. Using AI we plan to develop models for what normal development looks like, mapping individual development on to this we will provide parents or other professionals with quantifiable data to understand if a child's speech is on track, think of a speech version of the height and weight chart parents already use. 

Our speech and language intervention games are linked to the output of this ongoing assessment and expected milestones to seamlessly present personalised interventions. The success of these can be continuously monitored and adjusted overtime. 

Parents can use these tools at home to monitor their child's speech development and decide if their child is not meeting normal development milestones to go and get a clinical assessment. Our tools can provide the clinician with a comprehensive evidence base for what has been happening over time, not just a one off snap shot. The clinician can use our tools to monitor exercises at home. 

A large part of our mission is to take away the ambiguity around speech development, we are developing our solutions under a Precision Digital Therapy framework. This means supporting our claim with a rigorous evidence base as well as embracing the regulatory requirements. 

Other providers are led by speech pathologists, while this seems ideal it actually means they end up building digital versions of what happens in the speech pathologist's office. These solutions are not suitable for parents to use in an unsupervised way, and are not suitable for large scale deployment for use at home. By including assessment and monitoring capabilities we are able to provide parents with a new insight into their child's development. Parents can see how their child is developing, how well interventions/training are working and let them make informed decisions about their child's needs. If they have to enter the clinical care pathway our tools can be used by professionals to get a window into what's been happening over time and make better assessments and implement specific intervention protocols. 

Our innovation centers around putting the parent as the centre of the decision making process and building solutions that use the latest technology to enable this. 

Another innovative step in our solution is our approach to validation, we want to provide parents and other stakeholders evidence of efficacy and improvement over the status quo. This means we will make, and support through evidence, claims on the intended use for our solutions.

Our technology is built on speech recognition and predictive analytics. Using a children specific speech recognition engine and machine learning techniques we are able to score specific speech samples and compare this data to published normative data.  

We are currently building an MVP, a pronunciation scoring app that can tell if a child has the right articulation skills for their age. This is the first step in building a comprehensive monitoring and assessment application.

We are developing new voice user interface methodologies to ensure we can collect the speech data in the most constrained manner in order to provide accurate results. 

The big data side of our solutions has not been developed yet but we are managing our data with the view to build models of what "normal" speech development looks like over time. Predictive analytics can then identify when an individual children's profile is within expected developmental norms, based on these models, and when it is likely they will not meet subsequent milestones. 

We plan to develop new intervention training games that harness the power of technology to improve home based training by being more engaging for children and easier for parents to understand. Improved adherence to training have the expectations of shortening the time to resolution for speech delays and impairments. 

Taking a technology and data approach to building solutions means we can rapidly prototype and test new, more targeted intervention techniques. 

Our research has found that parents are aware of the importance of speech development, they have real concerns about their child entering school without proper communication skills and the repercussions of under diagnosed speech delays/impairments have been discussed before.  

In phase 1 we want to empower parents to be able to understand with ease how their child is developing. Giving them objective, reliable data that shows if their child is developmentally on track. The goal is to let parents act as the gate keepers for their child's needs and have reliable informative data to get help if and when it's needed. 

We want to keep families out of the clinical care pathway but providing tools that can address common speech delays and impairments that can be remediated in the home with the correct tools and early identification. 

In Phase 2 we want to improve the efficacy of delivering speech and language development services at the healthcare/education provider level. Validated digital tools mean that families can be served remotely, and more cheaply ensuring that location and wealth are not a barrier for any family.

Currently we are not serving any users (0), we are still developing our MVP which we plan to release in Sept '19. 

With the first release of our product, planned for late 2020, we expect to serve 3000-5000 people per month for the first year as we build our validation and profile. However we plan to achieve the 10k users a month within 2 years and doubling this within 3 years as the market matures. 

In year 5: With 36 million English and Spanish speaking children between 3 and 5 years of age and conservatively 10% experiencing speech delays the number of children would could benefit from our technology is 3.6 million children. 

If we take the least serious of these, between 10-20% that is 360k-720k users of our products per year that we can keep out of the clinical care pathway, freeing up clinical resources for more difficult cases. Of course we can still help a large section of children within speech therapy. 

The prevalence of speech disorders in Asia is less well known but China and India have only 5000 speech pathologists combined for populations over 3bn. There is clearly going to be disparity in these countries for resources and we plan to address them in the future

Our goals within the next year is to deliver the first validated speech development assessment tool that can be used outside of the clinical care pathway. We are building a tool that addresses speech and language impairments for young children as efficiently as possible. Whether this is through: 

• Reliable prediction of the need to attend speech therapy
• Early identification and intervention
• Reducing the demand on clinical services through improved home-based management
• Improved efficiency of clinical speech and language services delivered within healthcare and education

Our larger goal is to change how this industry works moving the early stages of speech development management into the home. We want to establish the most reliable tool to benchmark childhood speech and language development across the world, building tools that are deployable to anyone with a smart phone. 

All our development has the goal of improving outcomes for all children around the world, whether through smarter technology or improving the efficiency of traditional services. 

We have 2 main barriers:

Skepticism and push-back from the speech and language community. Many speech pathologists have the following concerns:

• Technology is not a substitute for face to face therapy sessions
• Screen time is already out of control
• The technology can't do what you say it can

It has been difficult to attract a champion in this space to help us get the message out there. 

Traditional fund raising: Our solutions need a significant amount of funding and time in order to collect the data to build the required data models. Adding the time required for validation and regulatory alignment means traditional funding mechanisms are difficult to secure. We are based in Ireland and the appetite for this type of investment is small. Traditional funding mechanism prefer B-B/enterprise solutions and are not as keen to fund direct to consumer applications. 

New languages will be a challenge for us in the future, particularly for non- European languages, where there are very few speech pathologists and increasing demand for services. 

We are overcoming our first barrier by initially going direct to parents, this is where we see the largest unmet need. We are building our solutions to tap into the type of insight parents want to see about their child's development. Our strategy centers on providing parents with this new level of insight at home, and from this they will expect the same level of data and insight from clinicians if they need to see them. We believe demand from parents will push the clinical community to change.

Clinical evidence is our second strategy for overcoming skepticism from speech pathologists. Generating a robust evidence base for our clinical claims will help us engage clinicians and get their buy-in. In terms of attracting a champion we are building relationships with early intervention groups as they see the need more acutely and we hope to overcome this barrier through them. 

In terms of funding we are building a prototype application to demonstrate the market need. There is the assumption that parents will pay for this service but funders want to see hard evidence of this. We are planning to release our MVP in Sept this year to demonstrate traction, after which we will search for funding again. 

For non European languages we will have to partner with organisation in the target countries to understand the clinical landscape, needs of the healthcare systems and challenges of replicating our applications for these languages. 

Our team consists of 2 founders, one working fulltime and one working part time. We have a technology lead who is working full time. 

We have a technical lead, who works on a as-needed basis and will join the company when we have full funding. 

Our team are all PhD researchers with experience in technology startups. The founders are both experts in speech recognition and all team members have a background in signal processing. Shona D'Arcy was Chief Scientific Officer for a digital health startup building devices to treat Tinnitus and Celine DeLooze is the architect for a speech processing application for the aviation industry. 

We are a unique team profile for building speech development solutions, we understand data but also appreciate the patient journey and how important this is for adoption of any new technology. Our team understands data, we have all worked on health research projects collecting and analysing large datasets from different patient groups. One of our founders is a trained phonetician and linguist and the other has validated 2 multi-site digital health clinical trials for smart inhalers and a device to treat tinnitus (both largest of their kind at the time). 

We are in a unique position to deliver the right solution for this problem because of our diverse set of skills and the fact that we are looking at this problem from a technology perspective that can be scalable. As mentioned before speech pathologists tend to reproduce what they already do while we are redefining how we assess speech development. 

We are still early stage and have some way to go before fully partnering with an organisation. However as part of our market validation we have spoken to Early Intervention Initiative in Dublin and a pre-school software company called Kidspath. We are in talks with both of these around partnering for the validation study we are planning next year.

They both wrote us letters of support for our recent application for European research funding. 

Our business model is a mix of B-B and B-C but we are starting with B-C, i.e. parents, as this is where we see the biggest unmet need. We are building a subscription based application that is recommended to be used for a minimum of 6 months in order to get an accurate picture of speech development. Other providers in this space charge approximately $10 per month, we are using this as the lower benchmark for what we can charge but this can increase as our capabilities improve

There is the possibility of hosting tele-therapy on our platform, we have not fully investigated this but know that it is a potential service parents will pay for. 

For the B-B market are solutions have 2 applications. The first is a preventative/screening tool to reduce the number of children unnecessarily seeking therapy. Screening can reduce the number of underdiagnosed issues that can be easily addressed if caught early reducing the time to resolution. Finally our training games can be used within intervention protocols to improve home based training and reduce the overall cost of care. Instead of a child needed to attend a therapists office 3 days a week for 10 weeks they would only need to attend 2 day a week reducing the cost from approximately $3000 to $2250. Our B-B strategy is a licensing model with insurance companies, hospital groups, national health services and education providers as potential customers.

Through equity funding and Irish/European grant money we hope to secure approximately $2.5m over 2 years. Our conservative revenue projections for 75k users in year 2 generates approximately $3m. In year 3 as this number expands to 120 and our retention rates improve we are predicting revenues of over $8m dollars. 

B-C products are difficult to sustain, however by going B-C first we get to build a community of parents and children who support us but more importantly we are continually collecting data to improve and develop new functionality. This route also enables us to complete ROI analysis to help us in our B-B commercial strategy. We haven't been able to understand the value of the B-B market, mainly due to difficulty finding out the cost of providing speech therapy, however we have found one reference to $9000 to provide therapy annually. We will work on demonstrating the reduction in cost that can be achieved by licensing our solution. 

This not only provides us with a more reliable business model but it democratizes access to our technology. If our solutions can be delivered through CHIP/ health insurance (paying us a licence fee) more children will benefit from it. 

We are an Irish company and while the network of startups and investors is quite close knit we are having trouble getting through the door of funders outside of Ireland. The Irish investment network is quite new to the longer life cycles for AI in digital health/life sciences and tend to fund companies at much later stages. Support from Solve for our solution would be invaluable for our funding rounds both here in Ireland and the US. Participating in Solve would expose us to more US based networks that we know are investing in companies building technology complimentary to ours. 

Partnerships are critical for us as well, Solve could help us build partnerships with organisations in the United States who are looking for innovation in this space. Again being from Ireland it can be hard to find the right person and cold calling is never successful. As mentioned we get pushback from the speech therapy community but this can be because they are not necessarily motivated to change the status quo. Partners involved in Solve are clearly looking to change and innovate and these partnerships could be invaluable for the future. 

Finally, we know that we are building solutions for just one part of childhood development, meeting other companies building other solutions for other aspects could result in partnerships that could accelerate deliver of solutions for both partners. 

We would like to partner with insurance companies, healthcare systems delivering/covering speech therapy costs. We would like to understand the payer landscape for speech and language services, what kind of savings would be required in order to . 

We would like to partner with any parenting organisations, either charities or companies such as Babycenter, Ovia Health or similar digital health solutions monitoring early childhood development. We would like to understand how we can partner them as the next developmental milestone in their suite of tools

The AI innovation prize would fund the development and testing of a AI models of speech development. This money would fund a team with several data scientists and researchers and a speech pathologist.

We would recruit a set of children through schools or pre-schools and collect speech data over 6-9 months. The first task is to improve the automated scoring algorithm, through manual labeling and iterative testing of machine learning rules. This data set would also enable us to develop the longitudinal profile of properly developing speech against which we would test new data in order to validate the scoring/prediction algorithms. 

This project would include a clinical speech assessment at the beginning and end, against which our claims of reliability will be matched. 

The AI innovation prize would fund the development and testing of a AI models of speech development. This money would fund a team with several data scientists and researchers and a speech pathologist.

We would recruit a set of children through schools or pre-schools and collect speech data over 6-9 months. The first task is to improve the automated scoring algorithm, through manual labeling and iterative testing of machine learning rules. This data set would also enable us to develop the longitudinal profile of properly developing speech against which we would test new data in order to validate the scoring/prediction algorithms. 

Our team is made up of researchers in bio-engineering and as such we have always worked using rigorous research methodologies and this project is no different. In the past we have collected data from vulnerable patients and this has informed our management of children's data which is critical to protect. We collect data under a Good Clinical Practice framework, which includes: ethics applications, patient information leaflets, informed consent from parents, coding/anonymisation of data etc. Ireland has strict data management laws (GDPR) that we adhere to, this cover where and how data is stored and protected and what to do in the case of data breaches etc.