Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Whachacallit med - Presentation

Florian Georg
November 08, 2020

Whachacallit med - Presentation

Presentation on our contribution to the DayOne HealthHack 2020. Project "Whachacallit med" on the "Medical Jargon Buster" Challenge

Florian Georg

November 08, 2020
Tweet

More Decks by Florian Georg

Other Decks in Technology

Transcript

  1.  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
  2. The Challenge Making medical information more accessible and understandable to

    patients https://2020.healthhack.solutions/project/9
  3. “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
  4. 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
  5. 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
  6. 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
  7. 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?»
  8. 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
  9. 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
  10. 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/
  11. 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
  12. 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, …