-
-
Moodfluent - Logo
Inspiration
The key inspirations for our projects came from our lived experiences of living with mental ill-health as well as frustration regarding the state of current offerings for wellbeing apps. Alongside this, we were also acutely aware of the fiscal costs of mental ill-health, as explored below.
Mental ill-health is a growing cause for concern in the UK and beyond, especially given the predicted rise in mental health concerns following the COVID-19 pandemic. Indeed, statistics from 2019 show that stress and mental ill-health accounted for 66% of all short-term work absences, and that in the same year, mental ill-health was the number one cause of long-term work absences.
Alongside negative consequences for individuals, poor mental wellbeing has significant effects on both the economy and the health system at large. Sticking with our figures from 2019, stress-related illnesses (including other mental health conditions) accounted for a whopping 68,897,000 GP appointments across the UK. In total, healthcare expenditure related to stress and mental ill-health in 2019 reached $14.76 billion USD, which is equivalent to £10.66 billion.
For those who wish to take steps to improve their mental wellbeing using mobile applications, however, there are some significant issues. Firstly, the market for general mental wellbeing apps is saturated with sub-par solutions which often only have a single feature per application. There are few apps that effectively combine existing technologies into one, simple to use solution that does not require the user to self-import extensive information. Indeed, there are even fewer apps that utilise augmented reality, despite the large (and growing) body of evidence as to its benefits in facilitating mental health interventions.
These ideas are best contextualised through the eyes of Sandra - a persona created based on our past experiences as a team. Sandra has been feeling down lately, and the pressure at work certainly isn’t helping. She knows there are plenty of wellbeing apps out there, so she downloads a few on her phone to try. However, Sandra soon gets frustrated with each app as she can’t find one with the right combination of features for her. She either has to journal in one app and then do breathing exercises in another, or the other way around. On top of this, she doesn’t like the mood rating system that many apps force her to use as it makes her feel uncomfortable to examine her own mental state in such detail. She’s so overwhelmed, she’s not sure could effectively rate her mood anyway!
What it does
Moodfluent is a patient-led app aimed at users with no serious mental health conditions, who simply want to keep on top of their wellbeing - whether that be long-term, or just during a rough patch. It uses natural language processing to determine the user’s mood from their journal entries. When the system detects possible low mood, it guides users to complete mindfulness activities in an AR environment.
Quite simply, there is nothing like our solution already on the market. Yes, digital mood diaries and journals are available - but these force the user to input their mood using either numbers, scales, or pictograms. Similarly, there are many mindfulness apps, but only one - Healium - utilises AR technology. Even then, Healium is simply impractical for many users as it relies on external input from devices such as an EEG headband or Apple Smartwatch.
A video demonstration can be found here: link
How we built it
Our team is fortunate to incorporate a diverse range of skills from coding to design, and business. Admittedly, with such a wide range of skills (and ideas), it took us some time to decide on a final concept, but this delay was worth it as it meant we gained a thorough understanding of the problem area. To do this we utilised tools such as PICO, SCAMPER, personas, and stakeholder analysis.
Once we had agreed on a final concept, as a team, we very quickly assigned tasks according to our skillset, in order to maximise our time during the 24 hour window. The division of labour was as follows:
- @ketteridge-lauren: Research and business case development
- @feirog - Implementation of AR mindfulness activities
- @pthara - UI/UX design and HTML/CSS implementation
- @nayan100 - NLP design and integration with webpage
Challenges we ran into
Though our team included members that have a thorough understanding of various coding languages, we found that tasks which we estimated to be relatively code free, were not so. In particular, when building our AR solution we very quickly found that a knowledge of Blender and C# was required - this posed a steep learning curve for @feirog as it meant she had to learn a new programming language in a matter of hours! Furthermore, there were some practical difficulties regarding the use of EchoAR, as it appears the system was at times overloaded with users.
Throughout the development of the WebApp, there were many instances where @pthara struggled to transform her code into the desired layout. Merging the sentiment form into the HTML and CSS files also proved difficult as it required @pthara to setup Visual Studio Code, pip, Python, and TensorFlow through the command line, which was a new experience. There were numerous results of “invlaid syntax”, but with perseverance and teamwork we continued to troubleshoot and successfully integrated the files!
When it comes to the NLP model, @nayan100 faced similar issues, as at the beginning, the model inappropriately predicted “joy” and “sadness”. After some tuning of parameters, this problem was mostly resolved but there are still some improvements to be made such as adding in more predictor variables such as keywords and surrounding word tests.
For @ketteridge-lauren, however, the challenges she faced were less about developing new skills or carrying out her tasks, and more related to finding her role within the team. Surprisingly, this is the first hackathon @ketteridge-lauren has attended that actually required technical ability, and she notes that this is something she was unprepared for! That being said, this challenge was one that was easy to conquer, given the kind-hearted nature of everyone on the team.
Accomplishments that we're proud of
Alongside our final outcome, as a team we are particularly proud of our persistence through this project, even when things got tough. At times, there was certainly some banging of heads against walls - but we still carried on! This meant that we could deploy a relatively polished outcome within 24 hours (which was a crunch). Who couldn’t be proud of that? Who couldn’t be proud of that? Indeed, @nayan100 notes that the level of persistence required to complete the NLP model is one of his proudest achievements in this hackathon.
For @pthara in particular, the fact that she managed to independently create a webpage after only a few weeks experience of HTML and CSS marks a major achievement, and we are all incredibly proud. And as a result of @nayan100’s advice, she can now communicate via the terminal to access repositories, install add-ons, and communicate with programmes. Another amazing achievement!
Similarly, @feirog is particularly proud of how she adapted under pressure to learn animation, echoAR, and Blender. These are all technologies that can be hard to get a grip on regardless of time pressure!
What we learned
Where do we start on this one - there has been so much!
At the heart of things, however, we cannot avoid how much our technical skills as a team have developed over the last 24 hours, as we rose to the challenge of unfamiliar technology. As a result, we have developed initial skills in Blender, C#, and developed are already existing skills in Python, HTML, NLP, design, and business development. Perhaps most importantly, these skills are ones that were not only picked up from developing Moodfluent - as a team, we made a concerted effort to take time to teach each other, even if it were not directly related to our project. In particular, @nayan100 has been a wonderful teacher, allowing the rest of us to get used to git, when it would have probably been easier for him to do it himself.
Alongside these technical skills, our soft skills such as teamwork and collaboration have certainly increased. At the start of this hackathon, we were quite literally strangers to each other, and we now have a solid, working prototype which is testament to our development!
What's next for Moodfluent
Our initial market evaluation revealed a large and volatile space, but most importantly for us, it showed a market that is still open to be cornered. We strongly believe that given our unique combination of features, our solution will be a success - especially if we engage with organisations such as The AHSN Network who can assist with both grant funding and accelerated market access, as well as ORCHA, and NHS digital at an early stage. These relationships will also be vitally important as we look to commercialise our product via a subscription model to both consumers and the wider NHS.
Going forward we would like to engage with key experts in NLP in order to refine our codebase, as well as consultancies such as Nexer Digital who are experts in accessible user interface design. Furthermore, whilst our app is legally classed as a health device, rather than a medical device (which means we do not have to adhere to the Medical Devices Regulations of 2002) we would be keen to obtain both a NICE evaluation and CE marking.
Log in or sign up for Devpost to join the conversation.