Department: Computer Science
Department: Mathematics, Statistics and Computer Science (MSCS)
Department: Advanced Information Technology
Specialization: System and Networks (Two years program) Department: Department of Information Engineering and Computer Science
Department: Computer Science and Engineering (Four years program)
I am interested in understanding how we can reliably estimate blood hemoglobin level noninvasively and explore other blood constituent levels applying medical image processing techniques, advanced machine learning algorithms, and complex data analysis procedures. To that end, I'm particularly interested in developing low-cost medical and wellness applications leveraging the sensing and computing power of a smartphone, the best candidate for a point of care tool. Besides hemoglobin project, I have a couple of collaborative research engagements which are briefly illustrated here. (details...)
Hemoglobin level estimation process focuses on the fingertip video and lower eyelid image of an individual, inspired by the new computational imaging paradigm to develop a smartphone-based low-cost and reliable blood diagnosis tool. Our both fingertip and eyelid-based methods effectively estimate the hemoglobin level as well as significantly outperform existing methods applying advanced machine learning algorithm such as the deep convolutional neural network. (BEst and SHeA)
Since patients have limitations to provide self-reported pain symptoms, we have investigated how mobile phone application can play the vital role to help the patients to detect their pain level using their facial images captured by a smart phone. The proposed algorithm has been used to classify faces, which is represented as a weighted combination of Eigenfaces. We apply angular distance and support vector machines (SVMs) on longitudinal and cross-sectional pain images collected from three different countries: Bangladesh, Nepal and the United States. (details...)
For physical activities detection, we have used a smartphone with built-in accelerometers and gyroscopes and developed a multimodal system using pressure sensor data from shoes. We also considered the sensor data as time series shapelets, used our algorithms to differentiate the shapelets, and applied Gaussian Mixture Models with time-delay embedding for detecting different activities. (details...)
In high-risk pregnancies, knowledge of prematurity is important for critical decisions such as the timing and mode of delivery, whether or not to resuscitate the newborn if delivered in the periviable (22 - 24 weeks) period, and whether neonatal intensive care treatment should be stopped if significant complications arise. Here, we approach a smartphone-based educational program that provides anticipatory guidance to parents with prematurity-associated risk factors and gives the opportunity to improve on current prematurity care and parental decision-making. The project was desinged and developed in collaboration with Neonatal-Perinatal experts at the Children Hospital of Wisconsin at Medical College of Wisconsin. (details...)