Reproducibility/repeatability spectral

Figure 8.8. Percent transmission measured in regions 1 through 8 using three different FT-IR spectrometers. In each case, the middle curve was the result of adding the values of 512 different single-scan spectra, and the curves above and below this spectrum represent this line cj, where a is the standard deviation of the 512 data points. No instrument performs well in region 1. For the other regions, the spectra measured by instrument A showed repeatable spectral artifacts, while the spectra measured by instruments B and C were relatively flat, as they should be since the transmission of PET in these regions was less than 0.01%. The noise level of instrument C, as estimated by the standard deviation, was clearly lower than that of the other two instruments. (Reproduced from [5], by permission of the Society for Applied Spectroscopy copyright 2000.)...

Provided El spectra have been measured under some sort of standard conditions (70 eV, ion source at 150-250 °C, pressure in the order of 10 " Pa), they exhibit very good reproducibility. This is not only the case for repeated measurements on the same instrument, but also between mass spectrometers having different types of mass analyzers, and/or coming from different manufacturers. This property soon led to the collection of large El mass spectral libraries, either printed [76-78] or computerized. [79] The best established El mass spectral databases are the NIST/EPA/NIH Mass Spectral Database and the Wiley/NBS Mass Spectral Database, each of them giving access to about 120,000 evaluated spectra. [80-83]... 

On current spectrometers, spectral line frequencies can be determined with a precision of 0.1 Hz or better and a difference in frequencies better than 0.2 Hz. The same reproducibility can be achieved for repeated measurements of a sample without need for excessive care when the measurement is run with temperature control and stable lock. 

Following identification at color Doppler ultrasound, the cavernosal vessels are interrogated. Pulse-wave (PW) duplex Doppler is turned on putting the sample volume on the cavernosal arteries. The spectral analysis is preferably performed at the base of the penis where the Doppler angle is particularly favorable (between 30° and 50°) and the flow velocity shows major reproducibility and correctness (Mills and Sethia 1996). The flow velocity must be measured repeatedly (at least three times) at the same level and the mean value reported. Functional studies have shown a progressive decrease of blood velocity in the cavernosal arteries from the base to the glans penis... 

Identification of Taxon-Specific Biomarkers Experiences from our work and those of others strongly suggest that efficient strategies for identification of taxon-specific biomarkers should ideally involve measurements of a representative number of strains per taxonomic unit (Lasch and Naumann 2011). This is particularly important for accurate identification at and below the species level. To differentiate microorganisms below the species level, MALDI-TOF MS characterization should be ideally carried out by repetitive measurements from an adequate number of independent microbial cultures per strain. Therefore, the cmrent experimental workflow at the Robert Koch-Institute (RKI) encompasses measurements of four technical repUcate spectra per each individual culture, that is, per biological replicate. Since we normally generate three independent crrltrrres per strain subsequent statistical analyses can be carried out on the basis of at least 12 spectra per strain. In this way it is asstrred that the mass spectral database contains sufficient information with respect to repeatability and reproducibility. The expanded ntrmber of spectra serves as an improved basis for systematic statistical analyses to identify taxon-specific microbial biomarkers. 

Databases of spectra of normal and diseased tissue need to be established, which will amply represent the spectral variance that may be encountered in practice. Because these databases will be established over long periods of time (and will most likely be updated on a regular basis), highly repeatable and reproducible instrument calibration (both wave number and intensity axis) is required. This will also facilitate the ability of data transfer from one instrument to another. On-line signal analysis techniques are needed to immediately characterize and/or classify tissue on the basis of its Raman spectrum. Various methods and techniques that can be applied for these purposes have been reviewed recently [8]. Therefore, their discussion will be omitted here. 

In an ideal set-up the repeated measurement of a sample would provide exactly the same spectrum. There are, however, two fundamentally different constraints on the one hand, there is always a contribution caused by random error, i.e. noise, which can originate from, for example, the statistics of photons or from the electronics of signal amplification. Noise can be reduced, e.g. by cooling the detector, but based on fundamental laws of physics it cannot be zero. On the other hand, apart from noise there are small remaining spectral variations among spectra of identical samples, which are summarised under the expression systematic errors. They are caused, for example, by sample carrier inhomogeneity or differences in the drying process, and they also affect the reproducibility of the spectroscopy. 

A variety of further tests concerning the repeatability, reproducibility and quality of IR spectra obtained can be performed. In fact, tests for these three important spectral parameters were first introduced to biomedical IR spectroscopy in the late 1980 s within the context of a project for FT-IR analyses of intact micro-organisms. In these tests, the spectral quality is checked with regard to absorbance values of the raw spectral data, the signal-to-noise ratio (SNR), spectral contributions from water vapour, optical fringes and more. Quality tests have been adapted and expanded for IR micro-spectroscopic imaging and are comprised now of five independent quality checks ... 

---