Technical Risks

T1. Security breach

T2. Battery drainage due to the high use of sensors in smartwatch

T3. Ability to determine an appropriate amount of data needed to validate our initial model, and to then capture enough example data to effectively train the personalized algorithm

T4. Correlating data between sensors to detect a seizure

T5. Machine learning might degrade performance

T6. The detection process will go through a “learning” phase which may increase the number of false positives

Technical Mitigation

T1. Encrypt all data in transit and at rest

T2. Only activate sensors that will be used and turn off all sensor listeners when not being used

T3. Speak with experienced neurologists who may be able to suggest or guide us to the frequency of certain paths of the AI decision tree. Initialize trained network using public research data biometric readings

T4. Set the same sampling rate of all sensor updates

T5. There will be an option to revert threshold values back in time, or completely reset threshold values back to initial values

T6. Let the patient know ahead of time that there will be a learning phase for the application. Starting out, we will apply typical biometrics of a user experiencing a seizure until the application “learns” the individual

Customer Risks

C1. The detection process will go through a “learning” phase which may cause the user to not trust the application due to an increase in false results

C2. May require a certain level of end user technical competency and commitment/dedication

C3. Patient data could be leaked in the event of a potential data breach

C4. If no network is available, important data can be lost

C5. If no network is available, emergency contacts can not be notified.

Customer Mitigation

C1. Let the patient know ahead of time that there will be a learning phase for the application, give a basic explanation to reassure the user regarding why it is necessary

C2. Provide detailed instructions of how to properly use the application

C3. Encrypt all collected data about the patient

C4. We will preserve the most important data, which is seizure data, in the Smartwatch itself.

C5. An audio alert will sound and instructions will display on watch so anyone nearby can aid effectively.