Code to generate simulated blood biomarker profiles, generate corresponding non-invasive measurements, build and train neural network and reconstruct data with trained network

Code provided in accordance with open data policy for publication (currently under review), 'Reconstructing Blood Biomarker Profile from Sweat, Tears and Other Non-Invasive Media: a Neural Network Approach', Submitted to MDPI Biosensors Oct 18th 2018. Part 1 of this code simulates concentration vs time data of a notional biomarker in the blood and creates a simulated 'measured' data set based on a series of mathematical shifts. Part 2 of this code builds a neural network, splits the simulated data into training examples, trains the neural network and then reconstructs the data sets from the split windows. New test data may be split into windows and run through the trained neural network in order to generate a prediction of the underlying blood-biomarker data. See publication for context and more information on the working of the neural network. All code written and run in GNU Octave version 4.2.1.

Donora, Matthew. (2018). Code to generate simulated blood biomarker profiles, generate corresponding non-invasive measurements, build and train neural network and reconstruct data with trained network, [software]. University of Edinburgh. School of Engineering. Institute for Micro and Nano Systems. https://doi.org/10.7488/ds/2462.