The Second International In-Vivo Monitoring Intercomparison Program for Whole Body Counting Facilities by Canadian and United

two fission product radionuclides (¹³⁷Cs and ⁶⁰Co ). The activity in the phantom was homogeneously distributed throughout all sections and was proportional to the volume of that section. The phantom also contained ⁴⁰K homogeneously distributed in an amount similar to a Reference Female, to produce a representative Compton background in the resulting spectra.

The amounts of the radionuclides on 08-May-96, 20:00 GMT (Greenwich Mean Time) were: ⁴⁰K: 2.99 " 0.15 kBq , ⁶⁰Co: 19.89 " 0.19 kBq , ¹³⁷Cs: 20.05 " 0.17 kBq. The participants were unaware of which nuclides were in the phantom. Each facility was asked to determine the number, identity and amount of the radionuclides in the phantom. Participants were advised to examine the energy range of 200 - 2000 keV. Each facility was asked to make an estimate of the precision (P) and estimate their Minimum Detectable Activity (MDA).

Determination of activity: Participants were requested to determine the activity of the identified nuclides in the phantom by two methods. The first method was to use their normal "man-sized" calibration factor (method-1). The comparison of activity estimated in this manner to the actual activity in the phantom gives the phantom-size dependency for the facility’s whole body counter. The second method was to use an efficiency corrected for the size of the female phantom (method-2); if the facility did not have any correction factors, then no results were reported. How this size-corrected factor was developed was not known to the organisers of the intercomparison and was left to the participants’ best judgement.

The accuracy of counting was obtained by evaluating the bias. The bias, B, is given by the following expression (HPS 1996, AECB 1997):

B ~ = ~ 100 ~ * ~ {(A sub i - A)} over A~~~~~(%)

Where A_i is the observed value obtained by either method-1 or method-2 and A is the true value.

Estimation of counting precision: Participants were requested to count the phantom repeatedly; without moving it. Participants were asked to supply gross counts, net counts and the counting time for each of the replicate measurements. This test was to estimate stability of the counting system (gross counts) and the effect of analysis on precision (net counts). Observed precision, OP, was then estimated by the following expression:

Where OP is the observed precision, A_i is the observed value (from gross counts or net counts), M is the mean of that data set, and N is the number of measurements (usually 5).

Radioactive decay is governed by Poisson statistics and the ideal whole body counter should have a precision close to the Poisson statistic. Poisson precision, PP, was estimated from the following expression:

Where PP is the Poisson precision, and M is the mean of the data set (gross count or net count). It follows that the ideal whole body counter should have a ratio of OP:PP of unity.

Determination of the Minimum Detectable Activity (MDA): Each facility was asked to calculate an MDA for all the radionuclides, except ⁴⁰K, detected in the phantom by measuring either an uncontaminated person or use a phantom containing an amount of ⁴⁰K expected to be found in Reference Female. The following expression was used to calculate MDA values:

MDA ~ = ~ 4.65 ~ * ~ SQRT {N} over {E ~ * ~ T} ~ + ~ 3 over {E ~ * ~ T}

Where MDA is the minimum detectable activity (Bq), N is the counts in the background region under investigation, E is the counting efficiency (cps/Bq), and T is the counting time (sec).

The MDA of a counting system is a function of the ambient background, the amount that background has been reduced by the shielding, the detector type, and the detector size. Although background count rate data was not supplied by the participants it was possible to extract it from the information supplied. The background counts were obtained from the facility’s MDA value for ¹³⁷Cs. The MDA was divided by the counting efficiency, E (calculated from the net counts and the facilities estimated activity), and the counting time, T, as shown below.

The background count rate is then obtained by dividing the background counts, N, by the counting time. For example, the background counts in the ¹³⁷Cs photopeak region varies from 0.002 counts per second (scanning bed with two low efficiency Ge detectors in a steel room) to 73 counts per second (Large NaI detector in a steel room). One facility had a very high background count rate of 2350 counts per second (large unshielded plastic phosphor scintillation detector).

MDA’s were time-normalised to 3000 sec by applying a correction defined by:

Where MDA_norm is the time-normalised value, MDA is the reported value, T₁ was the counting time used to determine MDA, and T₂ is the normalisation time (3000 sec).

RESULTS AND DISCUSSION

Identification of unknowns: All facilities correctly identified the unknown radionuclides as ⁶⁰Co and¹³⁷Cs. All counting systems were able to identify the unknown radionuclides except AV, presumably due to the poor resolution of the detector system.

Determination of activity: Results for the facilities are shown in Table 1. Fewer facilities supplied activities using method-2. A normal probability plot for ¹³⁷Cs and ⁶⁰Co(method-1) bias data after the outliers have been removed and show that the data are now normally distributed and each can be assumed to belong to a single set.

All things being equal, one would expect a distribution of results about a mean of zero bias for ¹³⁷Cs and ⁶⁰Co bias (method-1); however, this will depend on the calibration protocol of each participating facility. For example, a systematic bias would be introduced by a facility that calibrated its counting system with a phantom of different size or geometry to the one being measured.

The data show that most facilities have little difficulty in meeting the US and Canadian performance targets for bias (HPS 1996, AECB 1997). Data (method 1) are significantly different from zero bias (t-test, "=0.05) and have means of 6.7 " 10.1 (1.3 F_mean) and 5.6 " 7.2 (1.1 F_mean), respectively. Data (method 2) are not significantly different from zero bias (t-test, "=0.05) and have a mean of 3.5 " 8.4 (1.8 F_mean) showing that size correction factors have improved the results; however, the ⁶⁰Co data (method 2) in Fig. 4 are significantly different from zero bias (t-test, "=0.05) and have a mean of 6.9 " 9.4 (2.1 F_mean) showing that size correction factors have not improved these results.

Outliers excepted, a first inspection shows little difference between method-1 and method-2 for either radionuclide. This is proved by t-testing the ¹³⁷Cs method-1 and method-2 data, and the ⁶⁰Co method-1 and method-2 data. In both cases the null hypothesis is accepted, i.e., there is no difference between the method-1 and method-2 for either radionuclide. In other words, size dependency is not an issue for the Reference Female BOMAB phantom.

Analysing the bias data as a function of counting geometry, outliers excepted, there seems to be little difference between the different counting geometries. Performing Analysis of Variance on the data (less the Arc geometry and the Scanning Detectors where there are too few values) one finds that the null hypothesis, that all counting systems give the same bias values, is accepted for all cases. ¹³⁷Cs (method-1), (F = 2.00, p = 0.124); ⁶⁰Co (method-1), (F = 1.11, p = 0.352); ¹³⁷Cs (method-2), (F = 1.64, p = 0.225); ⁶⁰Co (method-2),(F = 0.25, p = 0.782).

Precision of counting: Results for the facilities are shown in Table 1. Testing the gross count data one finds that the null hypothesis, that the mean is no different from unity, is accepted for the Ge detector based systems (t-test, "=0.05, P=0.17), i.e., the variation is due to Poisson statistics. However, the null hypothesis is rejected for the NaI detector based systems (t-test, "=0.05, P=0.01) and the mean is different from unity, i.e., the variation is not due to Poisson statistics.. Other factors must have influenced the values of the data identified as outliers. The cause is not known.

Testing the net count data one finds that the null hypothesis, that the mean is no different from unity, is accepted for the Ge detector based systems (t-test, "=0.05, P=0.67). Similarly, the null hypothesis is rejected for the NaI detector based systems (t-test, "=0.05, P=0.00) and the mean is different from unity. It must be concluded that the use of NaI detectors is introducing a systematic uncertainty compared with Ge detector based systems.

Determination of the MDA: Analysis of ¹³⁷Cs MDA values versus counting time only suggests that lengthening the counting time improves MDA. Similarly, the analysis of time-normalised ¹³⁷Cs MDA values versus total shield thickness, only suggests that increasing shield thickness decreases MDA.

The time-normalised ¹³⁷Cs MDA data analysed as a function of total detector volume for NaI detector counting systems is also scattered and although no statistical testing has been performed the trend seems clearer. Larger detection systems tend to have lower MDA’s. Similarly, the analysis of time-normalised ¹³⁷Cs MDA values versus total relative efficiency for Ge counting systems shows that larger efficiency systems have lower MDA values.

CONCLUSIONS

The intercomparison has shown that the whole body counters that participated in this intercomparisons are not phantom-size dependent when measuring a Reference Female phantom. Photon energy, at least within the ranges tested, does not affect this dependency. The use of phantom-size correction factors improved facilities’ performance for ¹³⁷Cs, but not ⁶⁰Co. The whole body counting configurations statistically tested were shown to perform equally well. Comparison of measured counting precision with predicted Poisson counting statistics showed that the NaI detector based systems seemed to have a systematic uncertainty in addition to Poisson variability. Contrarily, this was not found for Ge detector based systems. MDA data supplied by the participants was scattered (14 - 3500 Bq for ¹³⁷Cs and 9 - 460Bq for ⁶⁰Co). No relationship was found between the facilities’s MDA values and shielding thickness or counting time. It was clear that counting systems with more detector volume (or higher relative efficiency) had lower MDA values than systems with smaller detector volume (or lower relative efficiency).

REFERENCES

American National Standards Institute. Performance Criteria for Radiobioassy. McLean: Health Physics Society; HPS N13.30-1996; 1996.

Atomic Energy Control Board. Technical and Quality Assurance Standards for Dosimetry Services in Canada. Ottawa: Atomic Energy Control Board; Atomic Energy Control Board Standard, S-106, 1997.

Kramer G. H.; Noel L.; Burns L. The BRMD BOMAB Phantom Family. Health Phys. 61(6): 895-902; 1991

Kramer G. H.; Limson Zamora M. The Canadian National Calibration Reference Centre for Bioassay and In-Vivo Monitoring: A programme summary. Health Phys. 67(2): 192-196; 1994.

Kramer G.H.; Loesch R.M.; Olsen P.C. The 1993 Intercomparison of the measurement of in vivo radioactivity. Rad Prot. Dosim. 86(3): 197-206; 1999.

Table 1: Identification, bias and precision, and MDA results. Blank space indicates no results received.

	¹³⁷Cs Bias (%)		⁶⁰Co Bias (%)		Ratio OP/PP
ID Code	Method-1	Method-2	Method-1	Method-2	Gross counts	Net Counts	MDA_Cs (Bq)	MDA_Co (Bq)
AA	14.9	0.5	10.5	-3.5	1.8	2.2	35	41
AB	21.4		25.6		0.9	1.2	343	90
AC	1.7		0.1				41	28
AD	6.6		-1.7		1.4	1.5	104	70
AE	-30.4		-35.4		0.7	1.0	148	96
AF	-27.9		-37.4		1.5	1.5	167	114
AG	-26.9		-33.8		1.6	1.9	155	108
AH	11.2		11.9		3.1	3.0	27	44
AJ	4.0		1.6		0.6	1.4	76	166
AK	19.6		18.1		1.2	1.1	152	218
AL	-6.8		8.5		0.4	6.7	56	54
AM	-9.1		-5.3		1.3	0.9	57	15
AN	0.1		-2.9		1.0	1.0	14	9
AP	8.4	7.1	12.6	11.2		1.3	67	67
AR	16.5		0.5		1.5	0.9	28	16
AS	9.6		8.5		1.0		185	185
AT	8.5		7.6			2.1	94	78
AV	-19.1				1.6	2.3	80	0
AW	11.3		11.6			1.2	3500	0
AX		-8.5			1.3	1.4	300
AY	-2.3		-2.8		1.4	1.5	223	113
AZ	-1.7		1.9		1.2	1.4	104	54
BA	22.6	29.6	-7.3	-2.0	1.3	4.4	306	273
BB	2.7		7.7			1.6	70	50
BC	-3.7		-37.5		5.9	1.4	374	348
BD	-3.8		-32.8		3.7	1.6	409	351
BE	-6.3		-26.7		2.6	3.9	24	28
BF	4.3	-7.5	26.9	18.8	0.7	0.7	105	69
BG	23.2		28.1		1.6	1.9	70	80
BH	-0.8	-5.4	6.5		0.8	1.2	59	53
BJ	7.6	2.5	1.1	-4.8	1.0	1.8	70	50
BK	20.0	7.7	16.8	15.9	0.2	1.1	85	67
BL	-4.7	-0.6	-3.1	-0.7	1.3	1.1	50	55
BM	8.9		3.0		1.3	1.2	360	460
BN	19.8		6.8		5.1	5.1	145	142
BP	-5.9		14.0		1.1	1.6	95	80
BR	12.1	4.3	9.7	6.7	1.7	1.0	130	115
BS	28.0	18.5	28.1	22.0	0.5	0.7	20	15
BT	8.8	3.1	-2.2	-4.1	0.6	0.4	300	230

Table 1 (cont): Identification, bias and precision, and MDA results. Blank space indicates no results received.

	¹³⁷Cs Bias (%)		⁶⁰Co Bias (%)		Ratio OP/PP
ID Code	Method-1	Method-2	Method-1	Method-2	Gross counts	Net Counts	MDA_Cs (Bq)	MDA_Co (Bq)
BV	1.0	-4.1	1.4	1.4	1.2	3.6	100	80
BW	-1.2		5.4		5.7	5.8	30	18
BX	-0.6	-5.7	-17.0	-20.1	5.7	3.4	61	43
BY	10.8		9.3		0.9	0.7	72	49
BZ	14.2		10.1		0.7	1.2	48	21
CA	-13.1		-10.1		0.9	0.9	49	34
CB	0.6		2.4		2.7	1.2	126	95
CC	8.3	-1.4	7.6	0.1	2.1	2.6	40	40
CD	9.2	0.4	14.9	5.3	11.7	2.1	125	125
CE	-8.0		-16.2		1.4	0.6	77	62
CF	18.8	22.0	8.5	11.0	1.6	3.3	158	178
CG	27.0		22.2		1.3	1.3	50	53
CH	7.2	16.0	10.8	24.9		2.3	60	60
CJ	13.9	9.9	12.6	6.2	1.0	0.4	150	150
CK	8.8	5.7	9.0	6.7	0.8	5.1	43	32
CL	3.1	-2.6	2.9	-1.3	11.3	9.4	44	28
CM	13.2		13.9		0.8	0.8	70	72
CN	80.1	78.6			1.9	4.2	35
CP	14.5		-17.1		0.9	2.1	78	94
CR	6.1		1.9		2.6	1.6	320	330
CS	6.4		4.3		1.5	1.3	23	40
CT		1.7		21.3	0.6	1.1	208	135
CU	13.2	12.8	3.1	1.7	2.5	2.6	45	57
CV	2.6		-6.9				119	76

^[1] Administered by the Human Monitoring Laboratory (HML), Radiation Protection Bureau, Health Canada