Slide 1

Slide 1 text

Whachacallit med Contribution to DayOne HealthHack 2020 «Medical Jargon Buster» Challenge

Slide 2

Slide 2 text

The Team Meet our core team

Slide 3

Slide 3 text

 Larisa – experience as breast cancer patient and definer/refiner of the challenge  Egle – experience as breast cancer patient, former nurse and experienced hacker  Simone – different perspectives as trained pharmacist, in contact with patients and through disability after an accident patient herself  Florian – software development and A.I. expertise, integration of technical cloud services, and clear explanations to the less technically endowed  Matthias – storyboards, user journey, visual experience and presentation skills  Stefanie – perspectives on and shaping of focus and pitch  Catherine – experience as lexicographer and translator, UI, project contact

Slide 4

Slide 4 text

The Challenge Making medical information more accessible and understandable to patients https://2020.healthhack.solutions/project/9

Slide 5

Slide 5 text

“I truly believe that patients have the power to influence the future of health care - The moment that patients really understand their important and vital role in this healthcare transformation, they become the leading force.” Larisa Aragon, Patient Champion Health Hack 2020

Slide 6

Slide 6 text

Feeling lost and confused about health related information can lead patients into despair and hopelessness

Slide 7

Slide 7 text

Vision Statement The solution to help patients navigate through the often complex and confusing world of medical information. Help patients to be empowered to manage their disease and make informed decisions with their care teams On top, caregivers, family and friends supporting their loved ones on their journey have a tool at hand that they can use or recommend as trusted source

Slide 8

Slide 8 text

The challenges Patients often struggle with specialized terminology and the sheer amount of information available on their health conditions – e.g. medical reports, research papers, studies, treatment brochures ... Key challenges patients commonly encounter:  Text is complicated and too long  Special terminology is not understandable  The text is not easily accessible (readability, unknown language, small fonts, too narrow etc.)  Source and credibility of information is unknown

Slide 9

Slide 9 text

Key features of the solution With our solution we will be able to: • Automatically summarize and analyse medical texts • Highlight and explain complicated terms with the help of and connection to an established lexicon • Make the text more accessible and reader friendly • Assess the credibility of the information

Slide 10

Slide 10 text

Design Process https://miro.com/app/board/o9J_kgzZthA=/

Slide 11

Slide 11 text

The Prototype Building a working first version (a.k.a. «Minimal Viable Product»)

Slide 12

Slide 12 text

User Journey: «Understand this document» «I wonder what this document is about. Let’s upload...» «I see it’s a research report that’s discussing my diagnosis and some treatment method I heard about. Seems quite trustworthy, apparently. Let’s read the summary – but I’m having a hard time without my glasses. And it would be nice if it were available in German...» «That was interesting. However, I don’t know this term. Can you explain it to me in easy words?»

Slide 13

Slide 13 text

Demo Try yourself: https://aka.ms/healthhack- jargon

Slide 14

Slide 14 text

Solution Architecture

Slide 15

Slide 15 text

Automatic Text Summarization Latent Semantic Analysis (LSA): Extracts the most «meaningful» sentences from a text  Alpaslan, Cicekli (2011): Text summarization using Latent Semantic Analysis https://www.researchgate.net/publication/220195824_Text_sum marization_using_Latent_Semantic_Analysis  Steinberger, Ježek (2004): Using Latent Semantic Analysis in Text Summarization and Summary Evaluation http://www.kiv.zcu.cz/~jstein/publikace/isim2004.pdf  Luís Gonçalves (2020): Automatic Text Summarization with Machine Learning — An overview https://medium.com/luisfredgs/automatic-text-summarization- with-machine-learning-an-overview-68ded5717a25  OSS Python code: https://github.com/luisfredgs/LSA-Text-Summarization

Slide 16

Slide 16 text

Detect health related concepts and entities in text Azure Text Analytics for Health With Text Analytics for health, users can detect words and phrases mentioned in unstructured text as entities that can be associated with semantic types in the healthcare and biomedical domain, such as diagnosis, medication name, symptom/sign, examinations, treatments, dosage, and route of administration. • See: https://techcommunity.microsoft.com/t5/azure- ai/introducing-text-analytics-for-health/ba-p/1505152

Slide 17

Slide 17 text

Immersive Reading for better accessibility Azure Immersive Reader Immersive Reader is an Azure Cognitive Service that lets you embed text reading and comprehension capabilities into applications. Helps users of any age and reading ability with features like reading aloud, translating languages, and focusing attention through highlighting and other design elements. Immersive Reader supports people of all abilities, including readers with dyslexia, ADHD, autism, cerebral palsy, emerging readers, and non-native speakers.  See: https://azure.microsoft.com/en-us/services/cognitive- services/immersive-reader/

Slide 18

Slide 18 text

Integrate a Medical Dictionary Merriam-Webster Medical Dictionary This up-to-date dictionary of medical terms and definitions is designed for health-care professionals or anyone who needs explanations of current medical vocabulary. More than 60,000 entries. Pronunciations provided for most entries. Covers the latest brand names and generic equivalents of common drugs. See: https://www.dictionaryapi.com/products/api-medical-dictionary

Slide 19

Slide 19 text

Future Improvements  Auto-summarization should improve with more advanced algorithms and/or medical language models. There are specialized commercial services, e.g. https://www.agolo.com/  Better semantic understanding and presentation using Text Analytics for Health  Trust Score: Assess the credibility of information sources  Immersive reading could be extended with in-context medical term definitions, picture etc.  Channels and document types: Transcribe a call with your doctor, scan & analyze your prescription or MRI report, support multiple languages, file types …  Gamifications, Social features, …

Slide 20

Slide 20 text

Thank You ☺ Our Project page: https://2020.healthhack.solutions/project/62