Camera and lens
An iXon 888EM+ (back-illuminated) EMCCD camera (Andor Technology Ltd, Belfast, Northern Ireland) with the following specifications was used: CCD201 (e2v technologies Ltd., Essex, UK), pixel format 1024×1024, pixel size 13 µm2, 16-bits at 1 MHz readout rate, system readout noise < 1e- at 16-bit 1 MHz readout rate, peak quantum efficiency (QE) 92.5% at 575 nm, and a dark current of <<1 electrons/pixel/s. The spectral sensitivity of the CCD is in the range of about 200–1000 nm.
The detector was cooled to -100°C with thermoelectric cooling including extra cooling by a water-cooled chiller (OasisTM 160LT, Solid State Cooling Systems, NY, USA) which was set to 15°C to avoid condensation in the camera. When using water cooling, the internal fan of the camera can be turned off and exceptional cooling performance is maintained indefinitely while absolutely no errors will be introduced due to vibrations of the internal fan.
A Xenon 0.95/25 mm C-mount lens (Jos. Schneider Optische Werke GmbH, Bad Kreuznach, Germany) was used. It covers a 1” format sensor with a 16 mm image circle. The high aperture (f/0.95) provides good image quality even under low light conditions. Xenon lenses are designed to cover the visible spectrum (400–700 nm) whereby the overall spectral sensitivity of the camera and lens system was reduced in the UV and the IR region. A lens with a slightly smaller image circle than the diagonal of the EMCCD detector (18.8 mm) was chosen as a trade-off for higher aperture since there are not many lenses with high aperture available on the market for a C-mount of 25 mm.
The acquisition parameters for UPE imaging of two human hands were empirically determined by imaging a representative set of subjects with different acquisition settings. There is a trade-off between imaging time and resolution (as well as signal-to-noise ratio) when choosing acquisition parameters. As a long imaging time ensures a high signal-to-noise ratio and a good resolution, it is demanding for the subject that needs to sit motionless in a completely darkroom for the respective time. For our UPE images the following settings were found to be optimal and were used: 1800 s (= 30 min) exposure time, 4×4 binning, 1.7 µs vertical shift speed, 1 MHz pixel readout rate, ‘normal’ clock amplitude, Electron Multiplying ‘on’, EM gain level= 300, pre-amplifier gain= 5.2. Pilot tests revealed that shorter exposure times than 30 min did not give enough signal intensity for some subjects, resulting in a poor signal-to-noise ratio. This could not be satisfactorily compensated by larger binning (e.g. 8×8) because cosmic ray effects and other noise on the detector would, in case of larger binning, affect a larger area of the image at the same time. The readout rate was the slowest possible to introduce less noise; a fast readout was not required for this experiment. For the two-hand imaging setup, the camera was placed at a distance of 65 cm from the hands to achieve an adequate field of view encompassing both left and right hand.
Measurement room
A darkroom for applications up to whole body UPE imaging was utilized. It was designed to be absolutely dark for reliable and high-quality UPE imaging of different kinds of biosystems. The darkroom itself was positioned in a temperature controlled room. It was ensured that the darkroom was free from any phosphorescent objects (walls, surfaces, etc.) that would interfere with the UPE imaging. A light lock with 2 doors was installed for entering the room without introducing stray light. The inner walls of the darkroom were covered with a black graphite paint (RAL 9005) having no phosphorescence. Electricity cables were introduced into the room through a sophisticated light-tight access. The cables were zigzagged from the bottom to the top of a double wall with alternating light blocking beams such that no light was able to come through the hole in the wall. In the inner wall, a fan was installed to introduce fresh air in the room.
For UPE imaging of two hands simultaneously, the hands were situated on a wooden plate with 2 wooden stubs designating the points between middle and ring finger of left and right hand. The plate and stubs were painted with black graphite. To avoid possible light reflections from one hand to the other, a wooden plate or black cardboard was introduced between the hands.
The light-tightness of the room was checked utilizing 2 procedures: (i) As conventionally done, a photomultiplier tube (PMT) was placed in the darkroom at various locations and with different orientations to estimate the signal as compared to the dark count of the PMT. This method provided first insights into light leakages. (ii) Since a PMT may have different spectral sensitivity than the EMCCD camera, a second procedure was introduced. A (dark adapted) white ball that reflects light was imaged with the EMCCD camera itself to see if a part of the ball was illuminated by stray light. Light leakage into the room relevant to the field of view could thus be located using the orientation of light on the ball. The same acquisition parameters of the camera were used for this purpose. The room was said to be completely dark when the image had similar intensity to an image with the camera shutter closed. For the UPE images of the hands shown in this manuscript this was ensured, i.e. the room was completely dark.
Experimental protocol
2 subjects participated in this experiment. The study was conducted in accordance with the principles of the Declaration of Helsinki. Subject #1 was a female (age: 32 years) and subject #2 a male (age: 44 years). All subjects adequately understood the study procedure and gave their oral informed consent.
On 5 different days, UPE hand images were taken twice (one in the morning and one in the afternoon), on 4 days with subject #1 (16th, 19th, 30th and August 31st, 2011) and on one day (August 27th, 2011) with subject #2. The time difference between the UPE images taken in the morning and the afternoon was 4.2 ± 2.5 h.
Right and left hand were imaged simultaneously with an exposure duration of 30 min. The dorsum side of the hand was imaged. Before starting the imaging, the subject was out of bright daylight for 2 h and the hands were dark-adapted for 30 min using hand gloves and once inside the darkroom, a dark-adapted black cloak was used to cover clothing to ensure that no delayed luminescence but only spontaneous UPE was imaged.
Image processing and statistical analysis
The UPE images were processed with Matlab (version 2013b; Natick, MA, USA). Noise from cosmic rays was removed by thresholding the images at 2000 total photon counts. Adaptive noise-removal filtering was applied to each image using a pixel-wise adaptive low-pass Wiener filter with a 5×5 pixel kernel to estimate the local image mean and standard deviation. Mean values of regions of interest (ROI) from the dorsum side of each hand (right and left) were determined by a 150×150 pixel ROI. These ROI values were used for a correlation analysis with respect to the time of day the measurements were taken. A second-order polynomial function was fitted to the data and the goodness of the fit was determined as well as the statistical significance of the correlation (t-test).