Inspiration

Approximately one-in-five pediatric patients are either overdosed or under-dosed each year. The goal of SnapDose is to reduce medication errors and under/overdosing pediatric patients, particularly in outpatient areas where accurate weights may not be readily available and emergencies are uncommon.

What it does

SnapDose is a device agnostic application designed for pediatric providers. SnapDose makes weight estimation and weight based dosing a snap. The app has a very simple interface, which users can choose to either insert a known patient weight, age or they can choose to calculate the weight with the app. In order to calculate the weight, users can either use the app to take a full body picture of the patient or they can upload an image. A calibration item must be set ahead of time, we recommend a credit card, license or id (something that has a very standard height), this item is also placed in the image with the patient. The app uses the tracking.js library to detect the calibrated item and the patient's body. SnapDose then uses the known item height and scales to find the patient's height. The estimated weight is then calculated based on an algorithm derived from CDC growth charts.

Users can search for a medication or select medications from common categories. All medications list the proper dosage for the current patient. Selected medications are stored in a patient summary with time markers, allowing care givers to thoroughly document the patient's administrations.

This all can be done extremely quickly to help care givers in emergency situations. We would love to add the ability for care givers to customize the layout of their medication list to improve ease of use.

How we built it

We used the MEAN stack to develop SnapDose. The tracking.js library was utilized to detect the calibration item and the patient in the provided image. SnapDose was intentionally designed with a mobile-first approach, allowing for use on a variety of platforms.

Challenges we ran into

Performing the image recognition was by far our greatest challenge. Recognizing both the calibration object and the patient in the image was difficult; however, it was even more difficult to ensure that the recognition functionality was extremely accurate and quick. Furthermore, ensuring that SnapDose is device agnostic with an intuitive and seamless layout required careful attention to detail.

Accomplishments that we're proud of

Very proud to have successfully estimated the height of a real baby and extrapolated it to an accurate weight!

What we learned

We learned how to manipulate an image to detect different features and objects in the photograph. Additionally, the two non-medical members learned a lot about health informatics.

What's next for SnapDose

We plan to extend support for SnapDose by developing native applications on other platforms (phone, ipad, etc.). We also plan to add the ability for care givers to customize their medication lists to show drugs that they use most frequently. Finally, we also plan to expand SnapDose's database of medications by linking to NYP's medication database.

Share this project:

Updates