The proposed research project is intended to develop an educational tool for the general public to be able to perform self-screening of their skin lesions. For this, a melanoma-detection mobile application will be developed to analyze lesion images acquired using a low-cost dermoscope attached to a smartphone. This product helps public or students to learn more about skin cancer and increase their self-awareness for early skin cancer prevention and treatment. The application automatically finds a lesion in a captured image. After lesion analysis, it displays different irregularities may exist in the lesion. It then explains for the user what these irregularities are and how these concepts are related to skin cancer. The partner organization will benefit from the product as there is huge interest in buying such educational, self-health and self-awareness application. Furthermore, the project can pave the way to higher goals and more expert systems in the future.

In medicine, imagining-based exams are used to support diagnostics, treatment planning and/or the treatment itself. Considering the number, integration and nature of devices in a multi-surface environment (e.g. smartphones, tablets, multi-touch surfaces and wall displays), this project proposes a system for imaging visualization and manipulation that takes advantages from each device’s capabilities in the environment. The goal is to integrate the framework MSEAPI developed at the University of Calgary – responsible for the creation of multi-surface environments and applications, as well as interactions between users and devices – with ResolutionMD from Calgary Scientific Inc. – an imaging visualization tool for medical diagnosis. This integration aims to show (i) how the future of medical imaging manipulation and visualization tools is envisioned and (ii) present a case for Calgary Scientific to utilize their tools in a novel environment.

This project aims to investigate machine-learning approaches for extracting speech-of-interest from competing background noise. Given a training set and a testing set of noise-only and voice-only recordings, and the ability to mix them in different ratios, find an algorithm that gives a likelihood estimate of the test signal being speech at a different times and frequencies. The first stage of almost any acoustic signal analysis is a pre-processing stage that performs a time-frequency decomposition of the signal. Features are then extracted from the time-frequency representation. A machine-learning algorithm will be applied to these features. Objective measures of speech distortions and noise attenuations will be used to measure the performance of the proposed algorithms. The partner organization will benefit from this intership by adding to the offering of solutions it can offer to the market. This could lead to increase sales and the creation of more employment in the company.

Trauma occurs in many forms at various instances of our daily life and some subtle ones go unnoticed until sever symptoms are observed. The affects on life vary from impairments to severe incapacitation. Current diagnosis methods are based on questionairs and subjective assessment with inherent variability associated to the doctors. In this internship we will be developing an automated diagnosis framework based on MRI imaging. The frist task will lead to automated image analysis pipeline providing accurate delineation of structures in the brain and provide their pertinent measurements. The next task will develop a framework to automatically identify data as being from patients or healthy individuals. This will then incorporate the current clinical paradigms and lead into an integrated framework. This pipeline and framework will then be further integrated into the current NeuroKinetics clinical assessment protocol.