? Scalable Health Initiative

AutoFeat

Automatic Diabetic Retinopathy Detection

Team Members

Nishant Doctor (ECE)

Sihan Zheng (ECE)

Caleb Lu (ECE)

Problem

According to the Vision 2020 report, there were 45 million cases of blindness by 1996, out of which 15% of those were due to diabetic retinopathy or glaucoma. Diabetic retinopathy is a condition that can cause damage to the blood vessels inside the eye thus leading to blindness. It is a critical eye disease, which can be regarded as manifestation of diabetes on the retina. Early diagnosis, timely treatment and proper method of screening of this disease have been shown to prevent visual loss and blindness in patients with diabetes. The basic fundamental problem faced with the current method of detection is that the process is completely manual and time consuming. The purpose of this project is to design an automated and efficient solution that could detect the symptoms of DR from a retinal image within seconds and simplify the process of reviewing and examination of images.

Report

Read our report titled "AUTO-FEAT: Automatic Diabetic Retinopathy Detection".

What we achieved in one semester

  • Wrote an algoritm that can detect "Cotton-Wool Spots" (CWS) on retinal images

Future goals

  • Improve algorithm accuracy
  • Receive and integrate feedback from clinic beta testing
  • Enable the link between the Android application and the feature detection algorithm

DocTouch

Post-surgical wound recovery

Team Members

Tianyi Yao (ECE)

Stephen Xia (ECE)

Momona Yamagami (BIOE)

Problem

With increasing quality of healthcare and better technology, the mean length of stay at a hospital after surgery has been decreasing over the years worldwide. As such, it is much more difficult for doctors to follow up and ensure that the patient does not have surgical site infections, or SSIs. Current interventions include phone calls made to the patient, routine checkups, and questionnaires, but all methods take time for both the patient and the doctor. Additionally, they assume that the patient can tell whether their wound is infected or not. However, research shows that patients do not have the ability to determine whether their wound is infected or not. By providing patients with an application through which they can complete a questionnaire about their wellbeing as well as take a video of themselves pressing around the post- surgical wound, DocTouch will allow the doctor to make the final conclusion about whether the wound needs further attention and can remotely track the wound healing process.

Report

Read our report titled "SPeach: Automatic Classroom Captioning System for Hearing Impaired"

What we achieved in one semester

  • Built the mobile app and partner web app, which allows for video capture and easy review
  • Created a clear stand to enable hands-free video capture

What we hope to do next

  • Improved integration of cloud storage and other features
  • Create an artificial neural network (ANN) to perform a preliminary examination of the wound, alerting the doctor if a recorded wound shows clear signs of infection.

PhaseGear

Transform to a new phase

Team Members

Adriana B. Flores (ECE)

Siam Hussain (ECE)

Cunzhu Xu (ECE)

Problem

Weightlifting injury can be faced by all levels of experience, from beginner to advance weightlifters and atheletes. The key issue leading to injury is performing weightlifting exercises with the wrong form. Achieving the correct form of an exercise requires careful observation of movement and instantaneous feedback. PhaseGear workout apparel records data about your body and motion, interpreting and personalizing in real time. It responds to your form and provides instantaneous vibration feedback to identify areas needing correction, and records all data to review.

Report

Read our report titled "Phase Gear: Transform to a new phase"

What we achieved in one semester

  • Built a partial sleeve with sensors
  • Developed an associated Android App
  • Device was able to detect the exercise, give vibration feedback if the speed was too great, and track the repetitions.
  • User is able to enter a Training Mode to calibrate the device to their physiology

What we hope to do next

  • Add 3D incorrect positioning feedback
  • Build a complete sleeve with multiple sensors, and eventually a full shirt
  • Add additional motion sensing applications around the data provided by the full PhaseGear shirt

RehabMe

Interactive In-home Rehab Exercise Solution

Team Members

Dan Volz (ECE)

Jack Wang (ECE)

Yize Zhao (ECE)

Problem

Physical rehabilitation is essential to the recovery of injuries and diseases such as sport injuries, stroke, and Parkinson's' disease, that affect millions of people every year. During the rehab period, which usually lasts from 4 weeks to 3 months, patients are expected to perform a set of simple exercises daily. Due to insurance constraints, most patients only visit rehab facilities once or twice a week. This forces patients to perform these exercises at home; however, many patients do not perform their in-home rehab exercises. Currently, therapists have no way to track patient's in-home exercise at all and could end up being blamed for the lack of recovery progress even if they have done everything right.

Report

Read our report titled "RehabMe: Interactive In-home Rehab Exercise Solution"

What we achieved in one semester

  • Developed a mobile app and cloud app to connect patients with therapists
  • Developed exercise 'cards'
  • Allowed user to record themselves performing the exercise and associate the recording with that 'card'
  • Chart performance over a 5 day span in the mobile app

What we hope to do next

  • Build a reward system (Digital Badges, Messaging, Social Network Integration)
  • Integrate with Apple's HealthKit and ResearchKit

Scalable Health Initiative