Medical Diagnosis using Raman Spectrographs and Machine Learning

点击翻译按钮获取中文摘要

MedMira is the developer and owner of Rapid Vertical Flow (RVF) Technology. The Company's rapid testapplications built on RVF Technology provide hospitals, labs, clinics and individuals with instant diagnosis fordiseases such as HIV and hepatitis C in just three easy steps. Together with a Tier II Canada Research Chair,Dr. Christa Brosseau, Medmira has identified rules based on sharpest peaks and preceding troughs in Ramanspectrographs of surrogate blood samples. Data analysis via application of these rules can accuratelydistinguish positive samples from negative samples. Such an analysis requires an experienced chemist toprocess the results from a Raman spectrometer; up to half an hour. The proposed project plans to automate thedata processing and explore the possibility of using machine learning to automatically generate the rules thatwere identified by the expert chemists. The machine learning approach is especially important for extendingtest from laboratory generated surrogate samples to complex human blood samples. Unlike the laboratorygenerated surrogate samples, Raman spectrographs of human blood samples can vary significantly from personto person. The heterogeneous blood samples may have additional peaks and troughs which could complicateinterpretation. More sophisticated machine learning techniques are more likely to replicate the diagnosticprocess facilitated by experts. The outputs of the project will enable MedMira to initiate development of toolsallowing non-experts to obtain diagnostics results in non-traditional health care settings without sacrificingquality.