CHAPTER 4

CHAPTER 4 APPLICATION OF INSTRUMENTATION

Detection of Biophoton Emission From

Partial Transparent Sample

As application of the instrumentation, we use partial transparent samples to measure the biophoton emitted from these samples.

Experiment

The background is measured first, then put samples inside container sitting in between two multiplier tubes. The counting curves are shown below:

Fig.4.1 Counting for Sample 1 Background

Fig.4.1 Counting for Sample 1

Fig.4.3 Counting for Sample 2 Background

Fig.4.4 Counting for Sample 2

Fig.4.5 Counting for Sample 3 Background

Fig.4.6 Counting for Sample 3

Fig.4.7 Representation of Axis

As Fig.4.7 indicates, x axis represents time in nanosecond, perpendicular axis represents number of counts.

Analysis

Experiment data are listed in following tables:

Table 4.1 Background

	Total counts	Peak counts	Channel no. at peak	Counts/minute
2 PMTS not blocked in between	3580	59	672	2.486
2 PMTS totally blocked	275	9	619	0.191
2 PMTS not blocked in between scintillator on	2712	40	608	1.928

Table 4.2 Sample 1 (lettus leaf)

	Total counts	Peak counts	Channel no. at peak	Counts/minute
Empty cuvette	1284	17	672	0.915
Cuvette + water	1589	28	670	1.103
Sample in cuvette	1100	23	674	0.938
Sample + water in cuvette	1350	29	677	0.766

Table 4.3 Sample 2 (fish eggs)

	Total counts	Peak counts	Channel no. at peak	Counts/minute
Cuvette holding water	3459	58	679	2.402
Sample in water in cuvette	3233	62	663	2.245

Table 4.4 Sample 3 (half transparent fish)

	Total counts	Peak counts	Channel no. at peak	Counts/minute
Plastic bag with water	15288	274	623	16.987
Sample in plastic bag hlding water	15250	238	636	16.944

For sample 1, we use lettus leaf to measure its biophoton emission and background.

For sample 2, we use fish eggs to measure their biophoton emission. Compare the background counts and count rate with sample, we can see, the count rate with sample is larger than background at peak but total counts are less than background.

For sample 3, we use live half transparent fish to measure its biophoton emission. Compare the two measurements in Fig.4.3 and Fig.4.4, the count rate with sample is smaller than background. Possible reasons can be:

Absorption of sample
Plastic bag holding water and fish also has some effects on measurement.
Scattering and absorption effect of water.

Randomness Analysis of Random Sequences Recorded From Instrumentation Using Neural Network Method

Computers and humans are good at doing different kinds of things. Neural networks are human attempts to simulate and understand what goes on in nervous systems, with the hope of capturing some of the power of these biological systems.

Neural network computations on RNA sequences can be used to demonstrate that data compression is possible in these sequences.^[22] According to reference 22, a random sequence would be incompressible, while structured sequences could show varying degrees of compressibility.

Method

As an application of the instrumentation of this paper, we use physical method to get "random" numbers and use recirculate neural network to analyze these numbers for the compressibility. If the numbers are periodically random, they can be compressed. If the numbers are really random, they can not be compressed.

Comparison is made between computer generated random numbers and our physical instrumentation generated random numbers.

In experiment, the physical random numbers are gathered by recording time sequences of each count appearing at different channel numbers on Multichannel Analyzer. The channel range then is divided into four equal regions. Each region is assigned one of T, C, A, G which are standard nomenclature for RNA.

The channel numbers are converted into T / C / A /G determined by their residence at different regions. We have following region assignment:

Channel 100--649 T

Channel 650--1199 C

Channel 1200--1749 A

Channel 1750--2300 G

Upon this time a sequence of T C A G has been obtained.

Using computer program to convert each character into binary number, for example, G--001, C--010, T--110, A--011,

We get a random sequence of 0 and 1 with 12 inputs and 12 outputs. (Total input 1512 numbers. Total output 1512 numbers in experiment)

Number of hidden layers and times of learning for physical random sequence and computer generated random sequence at the same reconstruction error are listed in table 4.5:

Table 4.5 Number of hidden layers and times of learning for computer generated random sequence and instrumentation generated random sequence at Reconstruction Error = 0.01.

	number of hidden layers
times of learning	12	11	10	9	8	7	6	5	4	3
computer generated random sequence	23810	590752	2138239	3234054	Critical 524624
physical random sequence by instrumentation	113254	159730	147278	162335	201309	316335	488851	486083	593101	Critical 62511

Before getting to critical point, the reconstruction error is kept at 0.01. When reaches critical point, the reconstruction error is 0.2580 for instrumentation random sequence and 0.2600 for computer random sequence.

Conclusion

Based on our data in section 4.2.1, we can see that instrumentation random sequence is more compressible than computer generated random sequence. As we see, instrumentation random sequence can be compressed to 3 hidden layers to reach critical point while computer random sequence can only be compressed to 8 hidden layers to reach critical point. The randomness of physical random sequence depends on how the data are gathered and also on how many numbers are gathered.