Fuzzy C-means clustering and principal component analysis of time series from near-infrared imaging of forearm ischemia

Hyperspectral Imaging: A New Approach to the Diagnosis of Hemorrhagic Shock

“…The data were then converted to optical density units by dividing the sample data by data acquired from a white reflectance standard using a Beer's Law algorithm. A complete description of this methodology can be found elsewhere 20 and in Appendix 1. End of study Bleed, periods during which 10 ml/kg blood was removed (10 min each); postbleed, periods of observation following bleed periods; LR, lactated ringer's solution; blood transfusion, reinfusion of shed blood.…”

“…20 Linear regression fit coefficients c1, c2, and c3 were calculated for each spectrum in an image cube from in vivo reference spectra 21 for oxyhemoglobin, deoxyhemoglobin, and an offset spectrum, respectively, by: S ij ϭ ʈc 1 OxyHb ϩ c 2 DeoxyHb ϩ c 3 Offsetʈ 2 (2) These coefficients allowed the determination of total hemoglobin concentration (THb):…”

Infrared imaging of subcutaneous veins

“…Despite a long history of medical IR photography, this technique is not widely used in medicine because of time considerations [3][4][5][6][7][8][9][10]. Prior experiments provided some interesting data using other near-IR detectors [11][12][13][14][15][16][17][18][19][20][28][29][30][31]. However, the device parameters were empirically selected without detailed experimental studies leaving the clinical prospects unclear.…”

Clustering multivariate time‐series data

“…In analyses of forearm ischemia [26] and clustering gene expression data [27], researchers found that applying PCA to the features does not necessarily improve clustering performance and in some cases may even degrade performance [27]. Although these researchers do not recommend using PCA with clustering, their results appear to be application specific.…”