Although multisensory processing is a ubiquitous sensory effect, computational models of the phenomenon have primarily used quantitative mathematical techniques largely to simulate multisensory enhancement. On the other hand, biologically-close neuron models and networks are extensively used in other fields of computational neuroscience to simulate a broad array of neuronal processing. The goal of this study is use a network of spiking neurons with synaptic plasticity rules to model the full range of multisensory integration (enhancement, depression and subthreshold).
For details, click here.
The goal of this project is to come up with machine-learning generated diagnostic rules to distinguish between normal and abnormal patients based on Bull's eye SPECT imaging of the heart. Abnormal patients are those who were found (in our case) to have either single coronary artery disease (SVD) or double vessel disease (DVD), using coronary arteriography. We have a database containing 160 NORMAL (86 males + 75 females) SPECT images in two views: stress and delayed, and and 24 ABNORMAL (19 males + 4 females) images, also in stress and delayed views. This quite unique (a large number of normal patients) data set was collected by Dr. Cios at the National Institute of Cardiology in Warsaw. To achieve our goal we have created a template and superimposed it on the original Bull's eye SPECT images to indicate boundaries of the coronary vascular territories, namely, left anterior descending (LAD), left circumflex (LCX), and right coronary artery (RCA). Next, we extracted the regions of interest corresponding to LAD, RCA and LCX from the Bull's eye quantitative polar map and discretized it. Finally, the CLIP3 machine learning algorithm was used to come up with diagnostic rules.