Multicentre evaluation of the IL Densiscan

The electrophoretic separation of human proteins (for example serum proteins and lipoproteins) is a very popular test in clinical chemistry; separation is usually followed by a densitometric reading. In this paper the characteristics and performance of a new automatic densitometer are described (the IL Densiscan, Instrumentation Laboratory SpA, Italy), together with the results of an evaluation. It was not possible to follow any international guideline or recommendation-there is no accepted international standard for the description and evaluation of densitometers. Therefore the experiments were developed in comparison with other kinds of instrument; some general statements in the literature were also followed [1 and 2].


Introduction
The electrophoretic separation of human proteins (for example serum proteins and lipoproteins) is a very popular test in clinical chemistry; separation is usually followed by a densitometric reading. In this paper the characteristics and performance of a new automatic densitometer are described (the IL Densiscan, Instrumentation Laboratory SpA, Italy), together with the results of an evaluation. It was not possible to follow any international guideline or recommendation-there is no accepted international standard for the description and evaluation of densitometers. Therefore the experiments were developed in comparison with other kinds of instrument; some general statements in the literature were also followed [1 and 2].
Precision and resolution-which define the accuracy of minima point detectionwere specifically investigated.
Any interference from the quality of the electrophoretic migration on the performance ofthe instrument was ruled out.

Materials and methods
Instrument design Densitometer hardware: the IL Densiscan is a fully automatic instrument designed for the analysis ofa single slide at a time. The instrument is programmable for scanning a variety ofelectrophoretic separations in micro, semimicro and macro size on a number of supports, which include cellulose acetate, agarose, poliacrylamide. Both transparent and partially transparent supports can be analysed.
The optical system of the instrument consists of an halogen lamp for scanning in transmission only and two filters, 525 and 620 nm. No fluorescence is possible. The analysis time is 20s per sample (including print-out of results). A photodetector transforms the emerging light beam into an electrical signal, which is then amplified, digitized and converted in accordance with a fast Fourier transform algorithm. After the results have been calculated, the signal is reconverted to a digital form for printing. An alphanumeric printer and a simple display system provide for a comprehensive dialogue between the operator and the instrument. A schematic diagram of the instrument is shown in figure 1.
Densitometer software: a general program is available for any kind of electrophoretic pattern; a graph is supplied with minima points identified (no identification is given when an intlexion in the curve is found). The percentages of the fractions are indicated underneath each graph. There is also a dedicated program for serum protein pherograms. The scanning movement covers the gamma area to albumin and all data are stored. Serum protein fraction identification is carried out as follows" the first fraction, if greater than 20%, is identified as albumin (pre-albumin, if any, is included in the albumin fraction).
The following two areas are considered as o and Then two areas are identified as [1 and [2, if necessary allocating a value to , if [ and [ are not separated.
The remaining area is indicated as %,; any monoclonal component, migrating in gamma region, is included in the gamma fraction. No further identification" is made if more than six fractions are read--a list of unidentified fractions with their percentages is given. A procedure is available for manual correction of minima points without the need to rescan the sample.
The software also includes a quality-control program to check instrument precision avoiding any interference connected with electrophoretic migration. A single pherogram is automatically scanned 32 times; the mean, the given for each fraction. This program can only be applied to serum protein pherograms.

Methematical treatment of data
Minima detection is a crucial feature of densitometers because of the high noise interference of the signal due to the characteristics of the different electrophoretic supports. To avoid this, the analogue signal is usually filtered at the output of the amplifier. In most densitometers the filtering function is fixed and does not take into account the different backgrounds which occur with different supports. This results in a limited capability of minima detection with manual correction often required. In the IL Densiscan the digital filtering function is optimized by adopting a fast Fourier transform algorithm. In this way filtration is modified by the background produced by the support, resulting in a better identification of minima points (figure 2).

Main evaluation
Two laboratories (Ospedale S. Raffaele, Milano, and Ospedale Civile Stradella, Pavia) were given one instrument each from the production line. In the first laboratory, instrument precision, accuracy of the slidepositioning system and the capability of minima detection were evaluated. The second laboratory checked the sensitivity and the criteria of serum protein fraction identification.
Electrophoretic separations were obtained on routine samples for serum proteins, lipoproteins and haemoglobins following the procedures described in table 1.  Figure 2 (a). Example of a pherogram scan before filtering using using the fast Fourier transform. (b) Example of a pherogram after filtering using the fast Fourier transform.

Evaluation
Pre-marketing evaluation Three Italian clinical laboratories were supplied with three prototypes of the instrument and were asked (after a short training period) to use them for a month with their own equipment. No strict evaluation protocol was followed; operators were asked to repeat their routine work with the prototype and to report on performance, ease of operation, drawbacks and problems. A generally positive conclusion was drawn by all participants.

Experimental and results
First laboratory Electrophoretic separations were obtained on 11 specimen for serum proteins (three containing monoclonal components, two with increased gamma fraction and six normals), 10 for lipoproteins and five for haemoglobins. Each pherogram was scanned 32 times a day for five consecutive days on the IL Densiscan and on a reference instrument.
The overall mean value for each fraction obtained on both instruments is reported in table 2 (serum proteins), To evaluate the imprecision of each instrument, CVs within series, between series and overall were calculated for each electrophoretic fraction in each sample.
To obtain a synthetic expression of the imprecision on different kinds of samples, the mean CV was calculated with the following formula: /= CV/n Table 2. Overall mean value for each fraction obtained using the IL Densiscan (IL) and the reference instrument (R). Mean CVwithin day (W), between days (b) and overall (o) for serum proteins. Student t-test as follows: ns (not significant), * * *p < 0.01, p < 0.02, *p < 0.05. The mean CV was calculated by the formula: 2'27*** 3"78 1"79" 3"09 1.09"* 4"56 3"06** 6"34 1"92"* 5"25 2"12 ns 9'86 2"76 ns 11"01 1"37"** 4"73 3"34*** 7"73 3"23 ns 8'40 3"07 ns 10"54 3"18 ns 12"84  No statistically significant differences were found for lipoproteins and haemoglobins. Repetitive readings were normally made without repositioning pherograms, the influence on the precision of refitting the same slide into the instrument was tested using two serum protein pherograms. They were scanned 32 times on each of three consecutive days, repositioning them for each scan both on the IL Densiscan and on the reference instrument. The results are shown in table 5. No statistical evaluation was performed because of the limited numbers in these experiments. The capability of minima detection was tested by scanning abnormal pherograms with poor resolution between fractions (tables 6-8).
Each pherogram was scanned 32 times and the performance of the instrument was evaluated by recording how many times the operator had to correct the results; comparing the findings with those on the reference instrument.

Second laboratory
To investigated instrument sensitivity, 1:2 1:4 and 1:8 dilutions of a normal serum were prepared and the pherograms obtained with each solution were scanned. The results are shown in table 9. Additionally, the limits of the dedicated program for serum proteins was studied in 30 pherograms characterized by various abnormalities" eterozygosity, splitting of 02 zone, and presence of monoclonal components. Unless there is a sufficient change of slope between an abnormal and normal fraction, no identification of the abnormal fraction is possible. Usually, when more than six fractions are present they are not automatically identified, instead they are listed together with their percentage value.

Discussion
The evaluation concentrated on instrument precision, capability of minima detection, sensitivity and validity of the program for serum proteins. The results shown in tables 2-4 indicate a high instrument precision in comparison with the reference instrument, both for serum proteins, lipoproteins and haemoglobins.
It appears that the high precision of the instrument is due to the data-processing method (fast Fourier transform), to the capability of minima detection and also to the system of slide positioning. Even refitting slides for each scan does not reduce precision (see table 5). Tables 6, 7 and 8 show that automatic minima detection, in comparison with the reference instrument, is excellent. Table 9 demonstrates that at low protein concentration the most heterogenous zones of the pherograms ( and y) give low values. This is probably due to the quantity of linked dye being lower than the instrument sensitivity limit, so that, to achieve success electrophoresis of specimens with low protein concentrations, for example urine, require a preliminary concentration to at least 4 g/dl total protein.
Automatic fraction identification by a dedicated software system is easy and saves time, but does not do away with the necessity for the operator to inspect each pherogram [3]. The software does not identify abnormal zones (eterozigosity and monoclonal components for example).
However, the instrument can be recommended as easy to operate and appropriate for a medium-size laboratory's Work-load.