| Summary: | Advancements in machine learning over the past several decades have provoked a sharp increase in humanity's ability to gather and analyze bulk data at a deep level of understanding - and use that data to make informed decisions and solve critical issues. Artificial Neural Networks (ANN's) have been used to handle analysis of various high-dimensional data types - including audio, video, visual, and numerical data. Medical Infrared Thermography (IT) utilizes infrared light to produce surface-temperature measurements of thermal images which may be used for medical diagnosis and analysis. Surface temperature thermograms give insight into the subsurface blood flow and tissue properties in a medically non-invasive and cost-effective manner. Image analysis of thermograms with ANN's have already been used to diagnose conditions and identify patterns in medical imagery for humans and many other species - but in particular - there are not many examples in the literature of its use in dogs (Canis lupus familiaris). Dogs have varying thermal surface temperature patterns based on breed/mix, coat type, relative size, and other characteristics. Our research seeks to identify a normal thermal pattern/range of dogs and determine how machine learning technologies may be applied to diagnose medical phenomena therein. We are developing a Convolutional Neural Network that seeks to analyze paired thermal/visual images of dogs alongside numerical data and determine if the dog has a normal or abnormal thermal pattern. Further classification, training, and network optimization may lead to increased diagnostic capability.
|
|---|