We recognized the importance of the 5-minute time frame in emergency response situations and how proper documentation leads to better accountability and organization in a hospital environment. Doctor Christina tackles the challenges of the stressful environments of medical professionals by giving them a way to communicate completely hands-free! Previous technologies such as pagers and public announcement systems either have long response times or require written work. Even the smallest decrease in response time in mission-critical situations will decrease the mortality rate. Additionally, reducing errors will decrease hospital spending, wait times and adverse effects on patients.
What it does
Doctor Christina is a completely voice-controlled wireless medical assistant, who will stay in the pocket of medical professionals as they treat patients, perform surgeries, and write documentation. Upon the input: “Christina!”, a short message will be recorded that is automatically converted from speech to text, and sent to a collective task-list that is ordered by urgency. This information is automatically compiled for a hospital administrator and makes it easier for them to allocate duties. For example, a doctor with their hands full could say, “Christina, Urgent: nurse to room 1 for assistance with childbirth”, to quickly call a standby nurse.
How we built it
Using the Raspberry Pi3 B+ and Adafruit i2s mem microphone to record, a 3-D printed case, Python for back-end, and Java for front-end, Doctor Christina relies on Picovoice for keyword detection and IBM Watson APIs for transcription. The Raspberry Pi is then rigged to send to a local PC through ssh where Java parses the transcript and displays it through a list GUI.
Challenges we ran into
None of us have a lot of experience with Raspberry Pi or Python. The largest issues we had were installing and running the APIs and libraries. Python install commands were inconsistent and likely had path errors. There were original plans to connect a PiZero to exhibit the low cost and size required but it had many installation difficulties. The entire cloud/server setup is only a temporary solution since accessing a PC through ssh is a severe security risk to the PC in use. The front-end also went through some difficulties with updating the list “database”.
Accomplishments that I'm proud of
Through a difficult and arduous process, we were able to get all of the back-end and front-end work. Though there was some trouble having all the components communicate properly, our final result functions well. In a short time period, we completed more than we originally anticipated we could, which allowed us to make new goals to achieve.
What we learned
We learned how to use a voice library, wake words, how to connect a Python back-end with a Java front-end. We learned how to attach a I2S mems microphone to a Pi and recording on the device.
What's next for Doctor Christina
We plan to reduce the size of the device by 75% using a Pi zero w and attaching a battery pack to make it fully wireless and portable. Additionally, we will be looking to make the front-end distinguish between rooms, to increase organization.