Lilly Medline

Inspiration

The growing medical costs around the world for the healthcare industry. According to the Centers for Medicare and Medicaid Services, the national health expenditure in the United States in 2021 was estimated at $4.3 trillion dollars (18.3% of our Gross Domestic Production) and expected to grow at a rate exceeding 5% each year for the foreseeable future. Compounding this issue, the U.S. Surgeon General has indicated there is a growing healthcare workforce shortage with an estimate of between 54,100 and 139,000 needed physicians by the year 2033, impacting primary care and rural areas the most. Additionally, current hospitals on average emit 4.4% of greenhouse gas emissions alone for the world with a waste estimate of 5 million tons per year (from a 2019 study, by the Association of American Medical Colleges) Even with our best-intended efforts, the quality of care can vary drastically from location to location throughout cities, states, and countries. By creating platforms such as Lilly Medline it is possible to alleviate some of this issue the global healthcare system is faced with

What it does

The app is designed to give patients more control over their own health by providing them with a tool to input and track their symptoms and receive a potential diagnosis. Patients can utilize speech-to-text to dictate their medical concerns before their doctor's visit, allowing for faster and better care as physicians will have more efficient use of time to ask more specific questions. This will in theory allow for higher accuracy in diagnosis and greater hospital throughput. Patients overall will be provided with a convenient user-friendly tool to take charge of their own health.

This tool also allows pharmacists, doctors, and other medical personnel with extensive case files to quickly separate a complex medical case into predefined categories of interest such as isolating cardiovascular issues or neurology-based medical concerns. Since diseases usually have clusters of presentation doctors can more readily rule in or out conditions or have a better idea of tests to run to identify conditions.

This also has the added benefit of reducing the stress on medical personnel who can better allocate resources.

How we built it

Front End

The frontend app was developed using Android Studio and the Java programming language. With Android Studio, we were able interfaces, add functionality and interactivity, and test our app on emulators or physical devices. By using Java in conjunction with Android Studio, we were able to create robust and feature-rich front-end applications that provide a seamless user experience.

Backend

The Backend app was also developed using Android Studio and the Java programming language. In addition, we leveraged the Back4App platform for user authentication where users can log in and log out as well as the ChatGPT API for illness suggestions based on user prompts. The ChatGPT API, on the other hand, is a machine-learning model that provides suggestions of possible illnesses based on user inputs.

We gathered and categorized all of the words in a medical dictionary into an Excel with 15 common categories (some words with multiple categorizations) that medical students study when learning about disease presentation (all of the medicine can fall into the 15 in some form). These words would then be identified in sentence structure which would highlight important flags.

1. Immunology and Virology, 2. Microbiology, 3. Cardiovascular, 4. Endocrine, 5. Gastrointestinal, 6. Hematology and Oncology, 7. Musculoskeletal, 8. Neurology and Special Senses, 9. Psychiatry, 10. Renal, 11. Reproductive and Genetics, 12. Respiratory, 13. Pharmacology, 14. Biochemistry, 15. General.

Challenges we ran into

*Categorizing each medical term from the medical dictionary into 15 medically related categories required an understanding of disease presentation. The success of the classification process depends on the accuracy of the categories. If the categories are not well-defined, it could lead to misclassification of the symptoms. *learning to build a mobile application while having limited coding experience *Converting voice to speech took a while to incorporate as the emulator wasn't given permission to record until tried on a physical device.

Incorporating the Excel spreadsheet to be accessible in the environment, had problems accessing a CSV file from raw. *Incorporating a function that creates a new HashMap object to store the data, then reads the CSV file Identifying the right categories.

Accomplishments that we're proud of

Integrating multiple features into one app that is fully functional, such as speech-to-text
Finishing the categorization of medical terminologies to best suit our app

What we learned

*Learnt how to Incorporate external tools or features that may require troubleshooting (such as emulator permissions or file access) *A deeper understanding of the data being classified

What's next for Lilly Medline

To incorporate basic preliminary questions for patients to answer. To incorporate a chat feature with an AI with complete trained data with different keywords to output one's diagnosis and continue with follow-up questions. (ongoing) To incorporate A.I. in giving recommended tests for isolating conditions among potential disease cases. To incorporate multiple language accessibility