1992, The British Journal of Radiology, 65, 438^*42
Technical notes Calibration of a 0.08 Tesia magnetic resonance imager for in vivo r, and T2 measurement By S. F. Keevil, M A , MSc, M I P S M , G. Dolke, DCR(R), P. Armstrong, M B , BS, FRCR and * M . A. Smith, MSc, PhD, FIPSM Academic Department of Radiology, St Bartholomew's Hospital, West Smithfield, London EC1A 7BE and 'Department of Medical Physics, University of Leeds, Leeds General Infirmary, Great George Street, Leeds LS1 3EX, UK
(Received 28 November 1990 and in revised form 23 April 1991, accepted 6 August 1991) Keywords: Magnetic resonance imaging, Relaxation times, Calibration
The magnetic resonance unit at St Bartholomew's Hospital is equipped with an upgraded MD800 resistive magnetic resonance imager (Imatec International Limited) operating at a field strength of 0.08 Tesia. In vivo spin-lattice relaxation times (T{) can be measured using a saturation recovery/inversion recovery sequence (SRIR, T* = 1000 ms, TY = 200 ms) and spin-spin relaxation times (T2) using a four echo CarrPurcell-Meiboom-Gill sequence (CPMG, TK = 1000 ms, TE = 32, 64, 96, 128 ms). The SRIR sequence consists of selective 90° pulses every 1000 ms, alternate 90° pulses being preceded by 200 ms adiabatic fast passage (AFP) inversion pulses. It can be considered to be an interleaved saturation recovery and inversion recovery sequence with gradient echoes. In each case a relaxation time map is produced on a pixel-by-pixel basis without the need for elaborate off-line image processing. The aim of the work described was to investigate the accuracy and reproducibility of 7", and T2 values measured using the imager, and to determine suitable calibration factors before commencing a programme of clinical research.
Our T, and T2 measurements were investigated over a period of 3 months, but the accuracy of T{ measurements can only be assessed over a shorter period following correction of an error in the commercial Tx calculation software. T{ measurements showed good reproducibility but poorer accuracy before this correction, as has been demonstrated by Richards et al (1987). In the standard imaging protocol, the test object, together with a thermometer, was placed transversely in a head coil inside the bore of the magnet. After an interval of approximately 30 min to allow the temperature of the gels to stabilize, a single axial slice through the test object was imaged using the SRIR and CPMG sequences. Afieldof view of 256 mm was used, together with a slice thickness of 12 mm and a resolution of 128 x 128 pixels; two signal averages were collected in each case in order to improve signal to noise. These parameters are typical of those used for clinical imaging at our institution. The temperature in the magnet bore was checked before and after each sequence. Typical 71, and T2 maps of the test object are shown in Figs 1 and 2, respectively. The relaxation times of the gels were measured by positioning circular regions of interest as shown and were corrected to 20°C. Methods and materials In addition, the standard protocol was modified in The calibration was carried out using a relaxation time test object developed by the EEC Concerted order to investigate the effects of imaging in the sagittal Research Project Identification and Characterisation of and coronal planes, varying the slice thickness, varying Biological Tissues by NMR (EEC Concerted Research the number of signal averages, using a different imaging Project, 1988; Podo et al, 1988). This object consists of a coil and varying the field of view. Up to four slices were Perspex disc with a series of holes into which glass tubes imaged to guard against possible variation between are inserted. Each tube contains a quantity of agarose slices in multislice imaging. gel doped with gadolinium (III) chloride to produce predictable relaxation properties (Walker et al, 1988, Results 1989). Calibration data and temperature correction Temperature corrected Tx and T2 measurements factors for 0.08 T were available for the gels in the test (mean ±1 standard deviation (SD)) made during the object. period of this study are plotted against the true relaxation times of the gels in Figs 3 and 4, respectively. In the Correspondence should be addressed to Stephen Keevil, case of Tu only data obtained following the software Magnetic Resonance Unit, Academic Department of correction have been included. Linear regression was Radiology, St Bartholomew's Hospital, West Smithfield, used to produce the following fits and correlation London EC1A 7BE. coefficients. 438
The British Journal of Radiology, May 1992
Technical notes
Measured T,/ms 600-1
Gel T,/ms Figure 1. Tx map of the EEC relaxation time test object showing region of interest.
Figure 3. Measured T, values (mean + 1 SD) plotted against the true T, of the EEC gels.
Measured T2/ms 200
-i
1 00
Gel T,/ms Figure 4. Measured T2 values (mean + 1 SD) plotted against the true T2 of the EEC gels. Figure 2. T2 map of the EEC relaxation time test object showing region of interest.
The accuracy of our measurements is indicated by the differences between the true and measured relaxation times of the gels. For Tx, the mean difference across the Tx = 9 . 1 7 + 1.03 r i m e a s (r = 0.997) (1) range of gel relaxation times following the software correction was —6.1% (range —10.4 to —2.9%), and T2 = -3.04+1.08 r2meas (r = 0.996) (2) for T the mean difference was —3.8% (range —9.4 to 2 where Tx and T2 are the true relaxation times of the gels 2.0%). The coefficient of variation of measurements and r lmeas and T2meas the measured values. All relaxation made over the period of the study gives a measure of reproducibility. The mean coefficient of variation was times are in milliseconds. 1.9% (range 1.1 to 3.2%) for Tx and 4.2% (range 3.1 to The variation in the measured relaxation times of each of the gels over the course of the study is shown in 8.0%) for T2. There were no trends in accuracy or Figs 5 and 6. Fig. 5 includes Tx measurements obtained reproducibility across the range of relaxation times before the software correction referred to above, and covered by the gels. It is clear from Fig. 4 that the discontinuities can be seen at the point in time at which accuracy of the measured T2 of one gel appears to be this correction was introduced (Day 78). The discon- significantly worse than all the others. This is more tinuity is worst in the case of gels with relatively long Tx likely to be caused by an error in the stated "true" T2 since one of the consequences of the error was to than to a real discontinuity in the behaviour of the underestimate long Tx significantly. The earlier results imaging system. are valid as an indicator of reproducibility, but were No variation was noted in the measurement of either neither taken into account when calculating accuracy relaxation time when the imaging protocol was modified nor when performing linear regression. in the ways described above. There was no variation Vol. 65, No. 773
439
Technical notes
Measured T,/ms
Measured T,/ms
6OO-1 500-
3
D
•
D
3
5
J
o
400-
300-
3
°
o
O
200 -
1 00 0
Time / days
Time / days
(a)
(b)
Figure 5. Temporal variation in the measured Tx of each of the EEC gels over the course of the study. The discontinuities mark the point at which the calculation software was corrected. The true relaxation times of the gels were as follows: (a) closed circles 156 ms, open circles 240 ms, closed squares 238 ms, open squares 333 ms, closed triangles 419 ms. (b) Closed circles 239 ms, open circles 331 ms, closed squares 416 ms, open squares 519 ms.
Measured T2/ms
Measured T2/ms
1 50 -i
1 50 -i
A
o
10
zA
AA
o
20
30
40
50
60
70
80
90
100
Time / days (a)
10
20
30
40
50
60
70
80
90
100
Time / days (b)
Figure 6. Temporal variation in the measured T2 of each of the EEC gels over the course of the study. The true relaxation times of the gels were as follows: (a) closed circles 86 ms, open circles 106 ms, closed squares 126 ms, open squares 138 ms. (b) Open circles 49 ms, closed circles 62 ms, open squares 81 ms, open triangles 118 ms, closed squares 123 ms.
between the relaxation times measured from different slices when multislice imaging was used. This is perhaps largely because the order in which slices are imaged by our system is such that contiguous slices are never excited consecutively.
measurement errors, such as physiological motion, cannot be overcome by calibration, but it is clearly important that as many sources of error as possible are eliminated in this way. Unfortunately, calibration and quality assurance have not been addressed in most published work on relaxometry, including the extensive reviews by Bottomley et al (1984, 1987). Discussion The work of Richards et al (1987) demonstrated The widespread rejection of relaxometry as a clinical technique results from the belief that the wide range of excellent reproducibility of Tx measurements made with in vivo T{ and T2 values to be found in the literature for our imager using the SRIR sequence. There was slightly a given tissue or pathology (Bottomley et al, 1984; 1987) greater variability for Tx values above about 490 ms, reflects true biological variation. The alternative view and this was ascribed to the use of a simplified algo(Smith & Taylor, 1986) is that these variations may be rithm for calculation of 71,, which assumes that Tx is largely the result of inaccurate and imprecise measure- small compared with the repetition time (Redpath, ment. Some of the factors which may give rise to 1982). An upgrade in 1988 introduced a more accurate
440
The British Journal of Radiology, May 1992
Technical notes
algorithm which allows for partial saturation. The CPMG sequence for measurement of T2 was implemented at the same time. The work reported in this note demonstrates the improvement in Tx reproducibility due to the new algorithm and also the reproducibility of T2 measurements. Furthermore, the use of the European Community test object has enabled us to demonstrate the accuracy of our techniques relative to an internationally recognized standard. Our discovery and correction of an error in the r, calculation software demonstrates the importance of calibration and quality assurance of this kind to ensure accurate measurement of r, and T2 using commercial imagers. Since there is an approximately linear relationship between the measured and true T{ and T2 of the gels, simple calibration factors (eqns 1 & 2) can be used to bring in vivo measurement into agreement with the test object calibration. These calibration factors have been shown to remain valid when the standard imaging protocol is varied in the ways described above. The accuracy of Tx and T2 measurements made using our system is in marked contrast to the poor results obtained from some other systems using the test object (Lerski et al, 1988). Some of the possible sources of error in relaxation time evaluation have been discussed by Johnson et al (1987) and, for T2, by Crawley and Henkelman (1987). The SRIR sequence, the original "Aberdeen" imaging technique (Edelstein et al, 1980), is thought to be a particularly accurate way of measuring r, because the majority of 71, information is incorporated into the signal by means of non-selective adiabatic fast passage, providing efficient inversion throughout the sample despite inhomogeneities in the radiofrequency field (Abragam, 1961). The improved algorithm used to calculate Tx from SRIR data is shown in eqn (3), SI and 52 being the signals obtained following the saturation recovery and inversion recovery portions of the sequence, respectively. This normalized signal difference is compared with a "look-up" table in order to evaluate Tx. 51 - 5 2 = 2{exp(- r 1 / r , ) - e x p ( - TR/T{)} 51 l-exp(-rR/r1) (3) The expression takes account of partial saturation, but it is assumed that inversion is 100% efficient and that the slice selective excitation pulse nutate spins through precisely 90° throughout the slice of interest. Errors due to slice profile effects can arise unless TR » Tx. In fact, the slice profile generated by our system is satisfactory, since the radiofrequency pulse used for selective excitation is a three-cycle sync pulse apodized by a Hanning window function. It is therefore likely that incomplete inversion by the adiabatic fast passage (AFP) pulse is the dominant source of errors. The performance of the SRIR sequence for T, measurement, particularly with regard to the behaviour of the AFP pulse in the presence of Bo and Bx inhomogeneities, has been simulated and investigated experimentally by Hardy et al (1985). These authors demonstrate the dependence of the inversion Vol. 65, No. 773
produced by a linearly swept AFP on the sweep duration and derive a non-linear sweep function with superior performance. Our AFP pulses consist of a linear frequency sweep from —4 kHz to +4 kHz through resonance with a duration of 10 ms. In the case of T2 measurement, it is likely that the systematic error is largely due to imperfections in the excitation and refocusing pulses and approximations in the calculation algorithm. Acknowledgments This work was carried out within the context of the European Community Concerted Research Project Tissue Characterisation by MRS and MRI (COMAC-BMEII. 1.3). We are grateful to the project leader, Dr Franca Podo, and our other colleagues in this multinational collaboration for their support and encouragement.. Dr J. P. Ridgway, Dr R. A. Lerski and Dr A. Lunt gave valuable advice on the use of the European Community test object. S. F. Keevil is supported by the Bunzl Gift Fund. References ABRAGAM, A., 1961. The Principles of Nuclear Magnetism (Clarendon Press, Oxford). BOTTOMLEY, P. A . , FOSTER, T. H . , ARGERSINGER, R. E. &
PFEIFER, L. M., 1984. A review of normal hydrogen NMR relaxation times and relaxation mechanisms from 1-100 MHz: dependence on tissue type, NMR frequency, temperature, species, excision, and age. Medical Physics, 11, 425-448. BOTTOMLEY, P. A., HARDY, C. J., ARGERSINGER, R. E. &
ALLEN-MOORE, G., 1987. A review of 'H nuclear magnetic resonance relaxation in pathology: are 7", and T2 diagnostic? Medical Physics, 14, 1-37. CRAWLEY, A. P. & HENKELMAN, R. M., 1987. Errors in T2
estimation using multislice, multiple echo imaging. Magnetic Resonance in Medicine, 4, 34-47. EEC CONCERTED RESEARCH PROJECT, 1988. Identification and
characterisation of biological tissues by NMR. Concerted research project of the European Economic Community. IV. Protocols and test objects for the assessment of MRI equipment. Magnetic Resonance Imaging, 6, 195-199. EDELSTEIN, W. A., HUTCHISON, J. M. S., JOHNSON, G. &
REDPATH, T. W., 1980. Spin warp NMR imaging and applications to human whole body imaging. Physics in Medicine and Biology, 25, 751-756. HARDY, C. J., EDELSTEIN, W. A., VATIS, D., HARMS, R. &
ADAMS, W. J., 1985. Calculated Tx images derived from a partial saturation-inversion recovery pulse sequence with adiabatic fast passage. Magnetic Resonance Imaging, 3, 107-116. JOHNSON, G., ORMEROD, I. E. C , BARNES, D., TOFTS, P. S. &
MACMANUS, D., 1987. Accuracy and precision in the measurement of relaxation times from nuclear magnetic resonance images. British Journal of Radiology, 60, 143-153. LERSKI, R. A., MCROBBIE, D. W., STRAUGHAN, K., WALKER, P. M., DE CERTAINES, J. D. & BERNARD, A. M., 1988.
Identification and characterisation of biological tissues by NMR. Concerted research project of the European Economic Community. V. Multi-centre trial with protocols and prototype test objects for the assessment of MRI equipment. Magnetic Resonance Imaging, 6, 201-214. PODO, F., ORR, J. S., SCHMIDT, K. & BOVEE, W. M. M. J.,
1988. Identification and characterisation of biological tissues by NMR. Concerted research project of the European 441
1992, The British Journal of Radiology, 65, 442^44 Economic Community. I. Introduction, objectives and activities. Magnetic Resonance Imaging, 6, 175-178. REDPATH, T. W., 1982. Calibration of the Aberdeen NMR imager for proton spin-lattice relaxation time measurements in vivo. Physics in Medicine and Biology, 27, 1057-1065. RICHARDS, M. A., GREGORY, W. M., WEBB, J. A. W., JEWELL,
S. E. & REZNEK, R. H., 1987. Reproducibility of spin-lattice relaxation time (T,) measurement using an 0.08 tesla MD800 magnetic resonance imager. British Journal of Radiology, 60, 241-244. SMITH, M. A. & TAYLOR, D. G., 1986. The absence of tissue
specificity in MRI using in vivo Tx or T2 determination: true
Technical notes biological variation or technical artefact? British Journal of Radiology, 59, 82-83. WALKER, P. M., LERSKI, R. A., MATHUR-DE VRE, R., BINET, J.
& YANE, F., 1988. Identification and characterisation of biological tissues by NMR. Concerted research project of the European Economic Community. VI. Preparation of agarose gels as reference materials for NMR relaxation time measurement. Magnetic Resonance Imaging, 6, 215-222. WALKER, P. M., BALMER, C , ABLETT, S. & LERSKI, R. A., 1989.
A test material for tissue characterisation and system calibration in MRI. Physics in Medicine and Biology, 34,
5-22.
The disappearing bladder—modifying imaging techniques after rectal excision By W. C. G. Peh, FRCR, N. C. Chokshi, FRCR and *C. H. Young, FRCS Departments of Diagnostic Radiology and "Urology, Selly Oak Hospital, Birmingham B29 6JD, UK (Received 5 July 1991, accepted 3 October 1991) Keywords: Bladder diseases, Pelvic surgery, Rectal surgery, Urinary outflow obstruction
Urological complications after abdomino-perineal excision of the rectum are well recognized. One complication is chronic urinary retention, which occurs in 14-50% of patients (Watson & Williams, 1952; Buckwalter et al, 1955; Eickenberg et al, 1976; Neal et al, 1982; Watters et al, 1983). The three main theories explaining causes of bladder dysfunction are direct nerve injury, loss of bladder support and pericystitis (Buckwalter et al, 1955; Eickenberg et al, 1976). Vesical neuropathy is now generally accepted as the main cause of urinary retention and this has been reflected in the emphasis on urodynamic studies, conducted mostly by urologists (Rankin, 1969; Fowler, 1973; Fowler et al, 1978; Neal et al, 1982; Chang & Fan, 1983; Lupton, 1986). We present two male patients who developed posterior—inferior bladder prolapse subsequent to rectal excision and illustrate the modifications of standard imaging technique required to demonstrate the altered anatomy.
sacrum. An intravenous urogram (IVU) examination was done and on the frontal projection, the bladder was noted to have an abnormal shape. The bladder also appeared to be in a low position (Fig. 1). A lateral projection done with the patient
Case 1 A 65-year-old man presented with persistent discomfort and a feeling of fullness in the perineal region. He subsequently had difficulty in micturition and poor stream. Abdomino-perineal excision for rectal carcinoma had been done 9 years previously. Cystoscopy was normal. Computed tomography (CT) scan demonstrated a prolapsed bladder lying posteriorly against the Address correspondence to Dr Wilfred C. G. Peh, Department of Diagnostic Radiology, University of Hong Kong, Queen Mary Hospital, Hong Kong.
442
Figure 1. IVU—frontal view shows abnormal position of a misshapen bladder. The British Journal of Radiology, May 1992
Technical notes Calibration of a 0.08 Tesia magnetic resonance imager for in vivo r, and T2 measurement By S. F. Keevil, M A , MSc, M I P S M , G. Dolke, DCR(R), P. Armstrong, M B , BS, FRCR and * M . A. Smith, MSc, PhD, FIPSM Academic Department of Radiology, St Bartholomew's Hospital, West Smithfield, London EC1A 7BE and 'Department of Medical Physics, University of Leeds, Leeds General Infirmary, Great George Street, Leeds LS1 3EX, UK
(Received 28 November 1990 and in revised form 23 April 1991, accepted 6 August 1991) Keywords: Magnetic resonance imaging, Relaxation times, Calibration
The magnetic resonance unit at St Bartholomew's Hospital is equipped with an upgraded MD800 resistive magnetic resonance imager (Imatec International Limited) operating at a field strength of 0.08 Tesia. In vivo spin-lattice relaxation times (T{) can be measured using a saturation recovery/inversion recovery sequence (SRIR, T* = 1000 ms, TY = 200 ms) and spin-spin relaxation times (T2) using a four echo CarrPurcell-Meiboom-Gill sequence (CPMG, TK = 1000 ms, TE = 32, 64, 96, 128 ms). The SRIR sequence consists of selective 90° pulses every 1000 ms, alternate 90° pulses being preceded by 200 ms adiabatic fast passage (AFP) inversion pulses. It can be considered to be an interleaved saturation recovery and inversion recovery sequence with gradient echoes. In each case a relaxation time map is produced on a pixel-by-pixel basis without the need for elaborate off-line image processing. The aim of the work described was to investigate the accuracy and reproducibility of 7", and T2 values measured using the imager, and to determine suitable calibration factors before commencing a programme of clinical research.
Our T, and T2 measurements were investigated over a period of 3 months, but the accuracy of T{ measurements can only be assessed over a shorter period following correction of an error in the commercial Tx calculation software. T{ measurements showed good reproducibility but poorer accuracy before this correction, as has been demonstrated by Richards et al (1987). In the standard imaging protocol, the test object, together with a thermometer, was placed transversely in a head coil inside the bore of the magnet. After an interval of approximately 30 min to allow the temperature of the gels to stabilize, a single axial slice through the test object was imaged using the SRIR and CPMG sequences. Afieldof view of 256 mm was used, together with a slice thickness of 12 mm and a resolution of 128 x 128 pixels; two signal averages were collected in each case in order to improve signal to noise. These parameters are typical of those used for clinical imaging at our institution. The temperature in the magnet bore was checked before and after each sequence. Typical 71, and T2 maps of the test object are shown in Figs 1 and 2, respectively. The relaxation times of the gels were measured by positioning circular regions of interest as shown and were corrected to 20°C. Methods and materials In addition, the standard protocol was modified in The calibration was carried out using a relaxation time test object developed by the EEC Concerted order to investigate the effects of imaging in the sagittal Research Project Identification and Characterisation of and coronal planes, varying the slice thickness, varying Biological Tissues by NMR (EEC Concerted Research the number of signal averages, using a different imaging Project, 1988; Podo et al, 1988). This object consists of a coil and varying the field of view. Up to four slices were Perspex disc with a series of holes into which glass tubes imaged to guard against possible variation between are inserted. Each tube contains a quantity of agarose slices in multislice imaging. gel doped with gadolinium (III) chloride to produce predictable relaxation properties (Walker et al, 1988, Results 1989). Calibration data and temperature correction Temperature corrected Tx and T2 measurements factors for 0.08 T were available for the gels in the test (mean ±1 standard deviation (SD)) made during the object. period of this study are plotted against the true relaxation times of the gels in Figs 3 and 4, respectively. In the Correspondence should be addressed to Stephen Keevil, case of Tu only data obtained following the software Magnetic Resonance Unit, Academic Department of correction have been included. Linear regression was Radiology, St Bartholomew's Hospital, West Smithfield, used to produce the following fits and correlation London EC1A 7BE. coefficients. 438
The British Journal of Radiology, May 1992
Technical notes
Measured T,/ms 600-1
Gel T,/ms Figure 1. Tx map of the EEC relaxation time test object showing region of interest.
Figure 3. Measured T, values (mean + 1 SD) plotted against the true T, of the EEC gels.
Measured T2/ms 200
-i
1 00
Gel T,/ms Figure 4. Measured T2 values (mean + 1 SD) plotted against the true T2 of the EEC gels. Figure 2. T2 map of the EEC relaxation time test object showing region of interest.
The accuracy of our measurements is indicated by the differences between the true and measured relaxation times of the gels. For Tx, the mean difference across the Tx = 9 . 1 7 + 1.03 r i m e a s (r = 0.997) (1) range of gel relaxation times following the software correction was —6.1% (range —10.4 to —2.9%), and T2 = -3.04+1.08 r2meas (r = 0.996) (2) for T the mean difference was —3.8% (range —9.4 to 2 where Tx and T2 are the true relaxation times of the gels 2.0%). The coefficient of variation of measurements and r lmeas and T2meas the measured values. All relaxation made over the period of the study gives a measure of reproducibility. The mean coefficient of variation was times are in milliseconds. 1.9% (range 1.1 to 3.2%) for Tx and 4.2% (range 3.1 to The variation in the measured relaxation times of each of the gels over the course of the study is shown in 8.0%) for T2. There were no trends in accuracy or Figs 5 and 6. Fig. 5 includes Tx measurements obtained reproducibility across the range of relaxation times before the software correction referred to above, and covered by the gels. It is clear from Fig. 4 that the discontinuities can be seen at the point in time at which accuracy of the measured T2 of one gel appears to be this correction was introduced (Day 78). The discon- significantly worse than all the others. This is more tinuity is worst in the case of gels with relatively long Tx likely to be caused by an error in the stated "true" T2 since one of the consequences of the error was to than to a real discontinuity in the behaviour of the underestimate long Tx significantly. The earlier results imaging system. are valid as an indicator of reproducibility, but were No variation was noted in the measurement of either neither taken into account when calculating accuracy relaxation time when the imaging protocol was modified nor when performing linear regression. in the ways described above. There was no variation Vol. 65, No. 773
439
Technical notes
Measured T,/ms
Measured T,/ms
6OO-1 500-
3
D
•
D
3
5
J
o
400-
300-
3
°
o
O
200 -
1 00 0
Time / days
Time / days
(a)
(b)
Figure 5. Temporal variation in the measured Tx of each of the EEC gels over the course of the study. The discontinuities mark the point at which the calculation software was corrected. The true relaxation times of the gels were as follows: (a) closed circles 156 ms, open circles 240 ms, closed squares 238 ms, open squares 333 ms, closed triangles 419 ms. (b) Closed circles 239 ms, open circles 331 ms, closed squares 416 ms, open squares 519 ms.
Measured T2/ms
Measured T2/ms
1 50 -i
1 50 -i
A
o
10
zA
AA
o
20
30
40
50
60
70
80
90
100
Time / days (a)
10
20
30
40
50
60
70
80
90
100
Time / days (b)
Figure 6. Temporal variation in the measured T2 of each of the EEC gels over the course of the study. The true relaxation times of the gels were as follows: (a) closed circles 86 ms, open circles 106 ms, closed squares 126 ms, open squares 138 ms. (b) Open circles 49 ms, closed circles 62 ms, open squares 81 ms, open triangles 118 ms, closed squares 123 ms.
between the relaxation times measured from different slices when multislice imaging was used. This is perhaps largely because the order in which slices are imaged by our system is such that contiguous slices are never excited consecutively.
measurement errors, such as physiological motion, cannot be overcome by calibration, but it is clearly important that as many sources of error as possible are eliminated in this way. Unfortunately, calibration and quality assurance have not been addressed in most published work on relaxometry, including the extensive reviews by Bottomley et al (1984, 1987). Discussion The work of Richards et al (1987) demonstrated The widespread rejection of relaxometry as a clinical technique results from the belief that the wide range of excellent reproducibility of Tx measurements made with in vivo T{ and T2 values to be found in the literature for our imager using the SRIR sequence. There was slightly a given tissue or pathology (Bottomley et al, 1984; 1987) greater variability for Tx values above about 490 ms, reflects true biological variation. The alternative view and this was ascribed to the use of a simplified algo(Smith & Taylor, 1986) is that these variations may be rithm for calculation of 71,, which assumes that Tx is largely the result of inaccurate and imprecise measure- small compared with the repetition time (Redpath, ment. Some of the factors which may give rise to 1982). An upgrade in 1988 introduced a more accurate
440
The British Journal of Radiology, May 1992
Technical notes
algorithm which allows for partial saturation. The CPMG sequence for measurement of T2 was implemented at the same time. The work reported in this note demonstrates the improvement in Tx reproducibility due to the new algorithm and also the reproducibility of T2 measurements. Furthermore, the use of the European Community test object has enabled us to demonstrate the accuracy of our techniques relative to an internationally recognized standard. Our discovery and correction of an error in the r, calculation software demonstrates the importance of calibration and quality assurance of this kind to ensure accurate measurement of r, and T2 using commercial imagers. Since there is an approximately linear relationship between the measured and true T{ and T2 of the gels, simple calibration factors (eqns 1 & 2) can be used to bring in vivo measurement into agreement with the test object calibration. These calibration factors have been shown to remain valid when the standard imaging protocol is varied in the ways described above. The accuracy of Tx and T2 measurements made using our system is in marked contrast to the poor results obtained from some other systems using the test object (Lerski et al, 1988). Some of the possible sources of error in relaxation time evaluation have been discussed by Johnson et al (1987) and, for T2, by Crawley and Henkelman (1987). The SRIR sequence, the original "Aberdeen" imaging technique (Edelstein et al, 1980), is thought to be a particularly accurate way of measuring r, because the majority of 71, information is incorporated into the signal by means of non-selective adiabatic fast passage, providing efficient inversion throughout the sample despite inhomogeneities in the radiofrequency field (Abragam, 1961). The improved algorithm used to calculate Tx from SRIR data is shown in eqn (3), SI and 52 being the signals obtained following the saturation recovery and inversion recovery portions of the sequence, respectively. This normalized signal difference is compared with a "look-up" table in order to evaluate Tx. 51 - 5 2 = 2{exp(- r 1 / r , ) - e x p ( - TR/T{)} 51 l-exp(-rR/r1) (3) The expression takes account of partial saturation, but it is assumed that inversion is 100% efficient and that the slice selective excitation pulse nutate spins through precisely 90° throughout the slice of interest. Errors due to slice profile effects can arise unless TR » Tx. In fact, the slice profile generated by our system is satisfactory, since the radiofrequency pulse used for selective excitation is a three-cycle sync pulse apodized by a Hanning window function. It is therefore likely that incomplete inversion by the adiabatic fast passage (AFP) pulse is the dominant source of errors. The performance of the SRIR sequence for T, measurement, particularly with regard to the behaviour of the AFP pulse in the presence of Bo and Bx inhomogeneities, has been simulated and investigated experimentally by Hardy et al (1985). These authors demonstrate the dependence of the inversion Vol. 65, No. 773
produced by a linearly swept AFP on the sweep duration and derive a non-linear sweep function with superior performance. Our AFP pulses consist of a linear frequency sweep from —4 kHz to +4 kHz through resonance with a duration of 10 ms. In the case of T2 measurement, it is likely that the systematic error is largely due to imperfections in the excitation and refocusing pulses and approximations in the calculation algorithm. Acknowledgments This work was carried out within the context of the European Community Concerted Research Project Tissue Characterisation by MRS and MRI (COMAC-BMEII. 1.3). We are grateful to the project leader, Dr Franca Podo, and our other colleagues in this multinational collaboration for their support and encouragement.. Dr J. P. Ridgway, Dr R. A. Lerski and Dr A. Lunt gave valuable advice on the use of the European Community test object. S. F. Keevil is supported by the Bunzl Gift Fund. References ABRAGAM, A., 1961. The Principles of Nuclear Magnetism (Clarendon Press, Oxford). BOTTOMLEY, P. A . , FOSTER, T. H . , ARGERSINGER, R. E. &
PFEIFER, L. M., 1984. A review of normal hydrogen NMR relaxation times and relaxation mechanisms from 1-100 MHz: dependence on tissue type, NMR frequency, temperature, species, excision, and age. Medical Physics, 11, 425-448. BOTTOMLEY, P. A., HARDY, C. J., ARGERSINGER, R. E. &
ALLEN-MOORE, G., 1987. A review of 'H nuclear magnetic resonance relaxation in pathology: are 7", and T2 diagnostic? Medical Physics, 14, 1-37. CRAWLEY, A. P. & HENKELMAN, R. M., 1987. Errors in T2
estimation using multislice, multiple echo imaging. Magnetic Resonance in Medicine, 4, 34-47. EEC CONCERTED RESEARCH PROJECT, 1988. Identification and
characterisation of biological tissues by NMR. Concerted research project of the European Economic Community. IV. Protocols and test objects for the assessment of MRI equipment. Magnetic Resonance Imaging, 6, 195-199. EDELSTEIN, W. A., HUTCHISON, J. M. S., JOHNSON, G. &
REDPATH, T. W., 1980. Spin warp NMR imaging and applications to human whole body imaging. Physics in Medicine and Biology, 25, 751-756. HARDY, C. J., EDELSTEIN, W. A., VATIS, D., HARMS, R. &
ADAMS, W. J., 1985. Calculated Tx images derived from a partial saturation-inversion recovery pulse sequence with adiabatic fast passage. Magnetic Resonance Imaging, 3, 107-116. JOHNSON, G., ORMEROD, I. E. C , BARNES, D., TOFTS, P. S. &
MACMANUS, D., 1987. Accuracy and precision in the measurement of relaxation times from nuclear magnetic resonance images. British Journal of Radiology, 60, 143-153. LERSKI, R. A., MCROBBIE, D. W., STRAUGHAN, K., WALKER, P. M., DE CERTAINES, J. D. & BERNARD, A. M., 1988.
Identification and characterisation of biological tissues by NMR. Concerted research project of the European Economic Community. V. Multi-centre trial with protocols and prototype test objects for the assessment of MRI equipment. Magnetic Resonance Imaging, 6, 201-214. PODO, F., ORR, J. S., SCHMIDT, K. & BOVEE, W. M. M. J.,
1988. Identification and characterisation of biological tissues by NMR. Concerted research project of the European 441
1992, The British Journal of Radiology, 65, 442^44 Economic Community. I. Introduction, objectives and activities. Magnetic Resonance Imaging, 6, 175-178. REDPATH, T. W., 1982. Calibration of the Aberdeen NMR imager for proton spin-lattice relaxation time measurements in vivo. Physics in Medicine and Biology, 27, 1057-1065. RICHARDS, M. A., GREGORY, W. M., WEBB, J. A. W., JEWELL,
S. E. & REZNEK, R. H., 1987. Reproducibility of spin-lattice relaxation time (T,) measurement using an 0.08 tesla MD800 magnetic resonance imager. British Journal of Radiology, 60, 241-244. SMITH, M. A. & TAYLOR, D. G., 1986. The absence of tissue
specificity in MRI using in vivo Tx or T2 determination: true
Technical notes biological variation or technical artefact? British Journal of Radiology, 59, 82-83. WALKER, P. M., LERSKI, R. A., MATHUR-DE VRE, R., BINET, J.
& YANE, F., 1988. Identification and characterisation of biological tissues by NMR. Concerted research project of the European Economic Community. VI. Preparation of agarose gels as reference materials for NMR relaxation time measurement. Magnetic Resonance Imaging, 6, 215-222. WALKER, P. M., BALMER, C , ABLETT, S. & LERSKI, R. A., 1989.
A test material for tissue characterisation and system calibration in MRI. Physics in Medicine and Biology, 34,
5-22.
The disappearing bladder—modifying imaging techniques after rectal excision By W. C. G. Peh, FRCR, N. C. Chokshi, FRCR and *C. H. Young, FRCS Departments of Diagnostic Radiology and "Urology, Selly Oak Hospital, Birmingham B29 6JD, UK (Received 5 July 1991, accepted 3 October 1991) Keywords: Bladder diseases, Pelvic surgery, Rectal surgery, Urinary outflow obstruction
Urological complications after abdomino-perineal excision of the rectum are well recognized. One complication is chronic urinary retention, which occurs in 14-50% of patients (Watson & Williams, 1952; Buckwalter et al, 1955; Eickenberg et al, 1976; Neal et al, 1982; Watters et al, 1983). The three main theories explaining causes of bladder dysfunction are direct nerve injury, loss of bladder support and pericystitis (Buckwalter et al, 1955; Eickenberg et al, 1976). Vesical neuropathy is now generally accepted as the main cause of urinary retention and this has been reflected in the emphasis on urodynamic studies, conducted mostly by urologists (Rankin, 1969; Fowler, 1973; Fowler et al, 1978; Neal et al, 1982; Chang & Fan, 1983; Lupton, 1986). We present two male patients who developed posterior—inferior bladder prolapse subsequent to rectal excision and illustrate the modifications of standard imaging technique required to demonstrate the altered anatomy.
sacrum. An intravenous urogram (IVU) examination was done and on the frontal projection, the bladder was noted to have an abnormal shape. The bladder also appeared to be in a low position (Fig. 1). A lateral projection done with the patient
Case 1 A 65-year-old man presented with persistent discomfort and a feeling of fullness in the perineal region. He subsequently had difficulty in micturition and poor stream. Abdomino-perineal excision for rectal carcinoma had been done 9 years previously. Cystoscopy was normal. Computed tomography (CT) scan demonstrated a prolapsed bladder lying posteriorly against the Address correspondence to Dr Wilfred C. G. Peh, Department of Diagnostic Radiology, University of Hong Kong, Queen Mary Hospital, Hong Kong.
442
Figure 1. IVU—frontal view shows abnormal position of a misshapen bladder. The British Journal of Radiology, May 1992
