Natural language processing AI bot that diagnoses cognitive diseases.
3x hackathon winner and awarded $1000.
- Best Use of Cloud Computing at Hack the Change 2022 (top 4 out of 400+ participants)
- Second Place in Healthcare at MetroHacks 2022 (out of approx. 400 participants)
- Best Use of Google Cloud at Major League Hacking
According to the World Health Organization, 50% of the global population lacks access to essential health services, and 100 million are still pushed into extreme poverty because of health expenses. Currently, 800 million people spend at least 10 percent of their household budgets on health expenses for themselves, a sick child or another family member, forcing them to survive on just $1.90 or less a day. It is incredibly important that a means of accessible medical clinician and diagnosis is available to all, regardless of circumstance. With a plethora of practicalities, Medic.AI solves the growing demand for means of effective remote healthcare. By enabling patients to book doctor's appointments, purchase medication, receive immediate medical diagnosis and seek the advice of medical professionals, Medic.AI is an all-in-one application that empowers everyone, everywhere. Indeed, the infamous Dysthymia served as the impetus for the main concept. Persistent depressive disorder, also called dysthymia, is a continuous long-term chronic form of depression. A diagnosed individual may lose interest in normal daily activities, experience periodic spites of hopelessness, lack productivity, and have low self-esteem. These feelings last for years and may significantly interfere with one’s relationships, school, work and daily life. Not only has it damaged the lives of millions of people around the world, but it has also negatively impacted our collective well-being as a species. Thus, due to the consequences of such mood disorders and medical illnesses, especially within the current COVID-19 epidemic, vulnerable people and sick patients are no longer allowed to see their doctors at a typical hospital. This lack of access to healthcare has prompted Medic.AI to address, challenge, and solve the dire need for effective and reliable means of accessible healthcare.
From the detection of neurological, psychotic and medical disorders, using a variety of diverse means; to the optimization of virtual diagnosis, Medic.AI is an all-encompassing platform that empowers patients everywhere. As an app created for third world countries, rural and developing areas, it is for those who do not have adequate access to medical support or fast and efficient examination. Designed to be used frequently to evaluate the severity of the most common medical conditions and mental illnesses, and to provide users with knowledge of their current health, users are given the option to consult the AI bot through text or speech. This further advances the intuitiveness, user-friendly interface and accessibility of MedicAI.
Using Google Cloud and Google Analytics, Medic.AI was created with Dialogflow AI, Cohere AI and the Google Cloud Console. To maximize the efficiency and optimization of the machine learning model, various datasets were used that encompass extensive ranges of mental illnesses and medical diagnoses from Kaggle and AI intents (see files attached). The frontend of the website is built in HTML, CSS and Javascript, while the backend was built with Javascript, PHP and Google Cloud. Additionally, Google Colab notebooks was used to test the models for efficiency. Cohere AI was integrated for the chatbot's natural language processing. Google Cloud Dialogflow web integration was performed for the integration of the ML model and data analytics products provided by Google Cloud into the website.
A few initial challenges were found throughout this project, especially in terms of the mental illness classification in the ML model. For instance, it had experienced trouble classifying a certain neurological disorder from another. However, these issues were solved by incorporating increasingly diverse and larger datasets in training the model. Through meticulous trial and error, brute force, and deliberate planning, each issue was successfully solved (with the addition of sacrificing sleep).
Significant accomplishments include the implementation of various datasets into the machine learning to increase its accuracy and return the most effective possible diagnosis to users. The average percentage accuracy for the medical diagnoses range from 82-95%, and will continue to exponentially increase as it the model trains!
Due to the extraordinarily versatile and secure infrastructures of Google Cloud and Cohere AI, I am excited to continue my journey building with what the innovative platforms have to offer. Not only did I grow tremendously as a builder, but I have grown intrapersonally as well. Given that it was my first time working with Google Cloud and Cohere AI, I was not initially familiar with the landscapes, however, the only way to grow is to venture beyond one’s comfort zone!
Medic.AI hopes to grow to be able to revolutionize diagnostic care across areas without medical access. To expand its reach, future plans include integrating Google Maps API to allow users to find nearby hospitals and periodically email users so as to remind them to get a diagnosis and seek advice on their present condition. Additionally, the ability to translate between various languages with Google Translate API. At Medic.AI, accuracy is an utmost priority, so there are additional plans to incorporate even more databases into the machine learning algorithm in order to improve the accuracy of all diagnoses.