MoleChecker

Our application, MoleChecker, is a manageable home-based platform for inspecting skin patches that delivers instantaneous results and will promote early care. MoleChecker is a dedicated screening tool with a simple user interface that requires the user to take a picture of the affected area of skin, which is then submitted and compared with the MoleChecker database. Upon opening the app, the user is presented with two choices: 1) use the screening tool to evaluate a skin patch or 2) contribute to the MoleChecker database by providing a picture and diagnostic information about a mole that has been assessed by a dermatologist. If the screening option is chosen, the camera of the phone turns on, the user takes a picture of the affected skin area, and the picture is submitted and compared with the existing database. A screening result is then returned to the user that gives the probability of the skin patch being “normal”, “suspicious”, or “malignant”. If the user chooses the contribute option, the user will be asked to take a picture of the mole, in the same manner as the screening option, and provide diagnostic information obtained from a licensed dermatologist. The picture and the diagnostic data are then sent to a server, which trains a convolutional neural network (CNN) based on user-generated data. CNNs are powerful machine learners due to their ability to recognize images with massive quantities of pixels. This data will help build a more robust model for MoleChecker that will increase the reliability of the screening tool with every new data point added. MoleChecker makes use of the inherent capabilities of a standard smartphone. It is implemented using a client-server architecture. The MoleChecker uses the built-in camera capability to image the skin. The camera’s flash works as a broad spectrum probing device in which the light from the flash emits red, blue, and green wavelengths that are either absorbed by the skin or reflected and scattered. The scattered light is returned to the cellphone’s camera which produces an image. The characteristics of the skin patch will determine the amounts and types of light that are absorbed, which will result in a different “skin profile” in the image. For example, melanomas contain large traces of melanin, which is an absorber in the red wavelengths, which corresponds to a heavy dip in the red channels contribution to the image. These “dips” in image intensity in the red, green and blue channel provided by the RGB picture (Red-Green-Blue) can provide an estimation of the skin’s diseased state. The quantities of absorbed and scattered red, blue, and green wavelengths can be used to quantitatively assess the likelihood of a skin mole being cancerous. By leveraging inexpensive cellphone hardware and providing MoleChecker on both Apple and Android platforms, we significantly reduce the barrier to adoption by users. MoleChecker has the potential to detect skin cancer in the initial stages, when no superficial deformities are visible or when the malignant lesion is still very small and can be easily mistaken for a benign growth. MoleChecker works differently from any other skin application as it uses diffusively reflected light from the flash built into a phone to query the underlying condition of the skin. This process is instantaneous and provides the user with a clear and immediate screening result, as opposed to other applications, which inspect the size, unevenness and color of a mole macroscopically with time. This application can help aid in the early detection of potential malignancies using a home-based screening that can be accessed by anyone with a smartphone. This application is not designed to prevent interactions with a physician, but rather to motivate a patient to seek counsel and potential treatment from a licensed dermatologist. MoleChecker has the potential to reduce the global impact of skin cancer by bringing preventive care into patients’ hands.

The team is composed of one member, Alexande Dumont. I have not had it easy in the past few years, being diagnosed with GBM (glioblastoma Multiforme) in 2014, at the start of my graduate career. For most people, being diagnosed with GBM is a death sentence, and the fear of dying freezes the person like a deer in the headlights. The tumor was removed and no sign of recursion has been found 4 years later. However, the desire for early and non-conservative diagnosis has taken a hold. In worldwide health, there is a generally poor preventative healthcare technology. People are too scared or just unaware of situations and they leave it until its too late, and the diagnosis is sometimes fatal. I want to introduce an app, built with Cordova and Azure platform that will help people to diagnose non-benign skin conditions early on, which will make them seek medical attention when lesions are pre-cancerous and more treatable.