European Journal of Radiology, 10 (1990) 98-104 Elsevier
98
EURRAD
00024
Magnetic resonance imaging of splenic iron overload Lionel ArrivC*, Siegfried Thurnher **, Hedvig Hricak and David C. Price Department of Radiology, Universityof Cal$omia, School of Medicine, San Francisco, CA U.S.A. (Received 30 July 1989; revised version received 25 October
Key words: Magnetic resonance,
comparative
1989; accepted
study; Magnetic resonance, spleen; Magnetic resonance; iron, magnetic resonance
28 October
resonance,
1989)
iron overload;
Spleen, magnetic
Abstract The value of magnetic resonance (MR) imaging in assessing iron overload in the spleen was retrospectively investigated in 40 consecutive patients. MR appearance, measure of signal intensity and Tl- and TZrelaxation times were correlated with the histologically determined level of iron in the spleen in each patient. Histologic examination revealed no iron overload in 19 patients, mild iron overload in seven, moderate iron overload in six, and severe iron overload in eight. All 19 patients with no splenic iron overload and 11 of the other 21 patients with splenic iron overload were correctly identified by MR imaging (sensitivity 52%) specificity loo%, accuracy 75 %). Splenic iron overload was diagnosed when a decrease of signal intensity of the spleen compared with those of adipose tissue and renal cortex was demonstrated. MR images demonstrated all eight cases of severe, three of the six cases of moderate, and none of the seven cases of mild iron overload. Only spleens with severe iron overload had a significant mean decrease in signal intensity and Tl- and TZ-relaxation times. Although specific, MR imaging is poorly sensitive to splenic iron overload.
Introduction The term ‘iron overload’ refers to an increase in body iron stores. Iron can be deposited either in parenchymal cells, particularly of the liver, or in reticuloendothelial cells, particularly of the spleen, bone marrow and liver, though there is often deposition in both [ 11. Iron overload can result from an abnormal increase in the amount of iron absorbed, parenteral administration of iron or blood transfusions [2]. Detection of mild iron overload is necessary for early diagnosis and quantiflcation of the iron level in severe iron overload is necessary to monitor therapy [ 31. Blood determination of ferritin levels is considered [4] the most reliable noninvasive method of assessing iron stores. Histologic examination of biopsy specimens, however, remains the procedure of choice [4]. *
Present address: Department of Radiology; Hopital Beaujon, 92118 Clichy Cedex, France. ** Present address: Departement Diagnostische Radiologie, CH-8091, Ztirich, Switzerland. Address for reprints: Lionel Arrive, M.D., Service de Radiologie, HBpital Beaujon, 92118 Clichy Cedex, France. 0720-048X/90/$03.50
0 1990 Elsevier Science Publishers
Several reports [ 5-81 have described the decreased signal intensity of the liver on magnetic resonance (MR) images of animal and human livers with iron overload. The present, retrospective, in vivo study was designed to determine whether MI? imaging can be used to quantitatively assess iron overload in the spleen. The spleen’s MR appearance, signal intensity, and Tl- and TZrelaxation times were correlated with histologically determined iron levels. Patients and Methods Patients Forty consecutive patients, 23 female and 17 male (aged 13-78 years, mean age 43 years), were retrospectively studied. Twenty-six of the 40 patients had been referred for MR imaging because of clinically suspected iron overload (primary hemochromatosis in six, dietary hemosiderosis in three, transfusional hemosiderosis in 17). The other 14 patients had undergone MR imaging for Hodgkin’s disease (seven patients), non-Hodgkin’s lymphoma (three patients), splenic abscess (three patients), and splenic infarction (one patient).
B.V. (Biomedical Division)
99
Histologic correlation of iron deposition levels was available in all 40 patients. Histologic specimens had been obtained of the spleen in 30 patients (22 by means of surgery, six by autopsy, and two by biopsy). Histologic specimens had been obtained by biopsy of the liver (five patients) or bone (five patients) in the other ten patients. Analysis of histologicfindings The histologic slides from the 40 patients were reviewed in conference by two of us to determine the degree of iron deposition. The ten patients with only liver or bone marrow tissue available for histologic analysis were included in our study because the amount of iron deposition in the reticuloendothelial system (RES) of the liver and bone marrow parallels that in the RES of splenic tissue [9]. Iron deposition, visible on hematoxylin and eosin stained tissue, was scored and categorized into four groups. A score of 0 indicated no iron deposition, ‘1 + ’ indicated mild iron deposition (perivascular only or less than two RES cells containing hemosiderin granules as seen per high-power-field (HPF) magnification of 400), ‘2 + ’ indicated moderate iron overload (two to ten RES cells/HPF magnification of 400), and ‘3’ indicated severe iron overload (more than ten RES cells/HPF magnification of 400). A4R imaging MR imaging had been performed in all 40 patients with a Diasonics MT/S system operated at 0.35 T. MR imaging had been additionally performed in live patients with a General Electric Signa system operated at 1.5 T. At 0.35 T, multislice spin-echo (SE) images were obtained with repetition times (TRs) of 500-2000 ms and echo times (TEs) of 30 ms for the first echo and 60 ms for the second echo. Both transverse Tl-predominant (SE 500/30) and transverse T2-predominant (SE 2000/60) images were obtained in all patients. In addition, Tl-predominant coronal images were obtained in three of the patients and Tl-predominant sag&al images in nine. In the five patients imaged at 1.5 T, both transverse Tl-predominant (SE 600/25) and transverse T2-predominant (SE 2000/60) images were obtained. Tl-predominant coronal images were obtained in two of these live. Analysis of iUR images The MR images were evaluated in conference by two of us who did not evaluate the histologic slides. These observers had no knowledge of clinical information or histologic results. Analyzed were: (a) the signal intensity of the spleen compared with those of the liver, renal cortex, striated muscle, and adjacent subcutaneous
adipose tissue on Tl- and T2-predominant images. (b) Percent contrast of the spleen compared to liver, renal cortex and adjacent, subcutaneous adipose tissue on Tl- and T2-predominant images at 0.35 T. (c) Tland T2-relaxation times of the spleen at 0.35 T. (d) Homogeneity of the spleen’s signal intensity on Tland T2-predominant images. (e) Longest dimension of the spleen-splenomegaly was diagnosed when splenic length was greater than 15 cm [ lo]. Signal intensities of the spleen, liver, renal cortex and subcutaneous adipose tissue were measured with use of the cursor to define the regions of interest (ROIs). The percentage of contrast between the spleen and the liver, renal cortex, or adipose tissue was determined as follows : % contrast
=
(signal intensity of spleen - signal intensity of the tissue) x loo signal intensity of spleen + signal intensity of the tissue
After all MR measurements were made of all patients, the MR results were compared to the histologic results. Sensitivity, specificity, and accuracy, were calculated. All quantitative data were expressed as group mean k 1 standard deviation (S.D.). The comparison of the measurements between the different groups was performed using the Student’s t-test for unpaired data. Results Comparison of histologicand MR findings Histologic determination of iron deposition in all 40 specimens revealed no iron overload in 19, mild iron overload in seven, moderate iron overload in six, and severe iron overload in eight. From MR imaging findings, all 19 patients with no splenic iron overload were identified as such. Of the 21 patients with iron overload, only 11 patients were correctly identified. Splenic iron overload was diagnosed when a decrease of signal intensity of the spleen compared with those of adipose tissue and renal cortex was demonstrated. Overall sensitivity, specificity and accuracy, therefore, were 52%) 100% and 75 %, respectively. However, results differed with respect to the degree of iron overload: MR image analysis identified all of the eight severe, three of the six moderate, and none of the seven mild cases of iron overload. In addition to the histologic findings of iron overload described above, 26 of the 30 splenic tissue specimens obtained at surgery revealed no abnormality, three showed splenic abscess, and one showed splenic infarc-
100 TABLE I Percentage
of contrast on Tl-predominant
images of splenic tissue with various levels of iron
Degree of iron deposition (severity)
Number of patients
None (0) Mild (1+) Moderate (2 + ) Severe (3 + )
19 7 6 8
Percentage
contrast
(mean + S.D.)
spleen/liver
spleen/renal
-l* 8 -8k23 +6*27 +21 k 89
+7*13 +9& 19 +2* 17 -23 f 12*
cortex
spleen/fat -85 f 14 -8Ok 8 - 89 f 23 - 106 f 31**
* = p < 0.001. ** = p < 0.05.
TABLE II Percentage
of contrast
on T2-predominant
images of splenic tissue with various levels of iron
Degree of iron deposition (severity)
Number of patients
None (0) Mild (1+) Moderate (2 + ) Severe (3 + )
19 7 6 8
Percentage
contrast
(mean f SD.)
spleen/liver
spleen/renal
+60 + 16 +54 + 11 +50*23 +29 k 69
-7* 13 -6+ 12 -23 f 35 - 78 f 26+
cortex
spleen/fat -26 + 22 -13+ 7 -43 f 29 - 103 f 37*
* = p < 0.001.
tion. These last four specimens had also demonstrated iron overload. Two of the three splenic abscesses and the splenic infarction were depicted on MR images. MR appearance of normal spleen (no iron overload)
On Tl-predominant images of the 19 spleens with no iron overload, MR signal intensity of the spleen was similar to or slightly lower than that of the liver, similar to that of the renal cortex, and markedly lower than that
TABLE III Tl- and TZ-relaxation times of splenic tissue with various levels of iron Degree of iron deposition (severity)
Number of patients
None (0) Mild (1+) Moderate (2 + ) Severe (3 + )
19 7 6 8
* = p < 0.005. ** = p < 0.001.
Relaxation time (mean + S.D.) Tl 802 859 890 597
T2 + f k *
157 128 249 97*
75* 7 81k 9 67 + 14 39 * 19**
of adipose tissue. On T2-predominant images, MR signal intensity of the spleen was markedly higher than that of the liver, similar to that of the renal cortex, and approaching that of adipose tissue. These findings on Tl- and TZpredominant images were regardless of field strength. The percent contrast of spleen to the other tissues studied and the relaxation times of the spleens at 0.35 T are given in Tables I-III. Nine patients demonstrated inhomogeneity of the normal splenic tissue on both Tl- and T2-predominant images (Fig. la and b). The inhomogeneity in signal intensity was between the anterior and posterior aspects of the spleen and was due to radiofrequency-field inhomogeneity. Normal spleen varied in size, but in none did the length exceed 15 cm. MR appearance of splenic iron overload
In the seven patients with mild iron overload, no change in splenic MR signal intensity was observed on either Tl- or T2-predominant images. Percent contrast in relation to other tissue and relaxation times of the spleen were not significantly different from those of normal spleen (Tables I-III). Splenomegaly was demonstrated in one spleen that had an associated splenic abscess. The abscess was visible as a high-intensity area
Fig. 1. Normal spleen showing image degradation due to radiofrequency-field inhomogeneity (0.35 T field strength). (a) Tl-predominant (SE 500/30) and (b), T2-predominant (SE 2000/60) images show inhomogeneous signal intensity within normal tissues. Low intensity areas are seen in the lateral segment of the liver (L), anterior aspect of the spleen (S), and adjacent subcutaneous fat (arrows).
of TZpredominant images. Another abscess in another spleen in this group was not depicted by MR. In the six patients with moderate iron overload, no change in splenic MR signal intensity was visible on Tl-predominant images (Fig. 2a). A decrease in splenic MR signal intensity was observed on T2-predominant images in three of the six patients (Fig. 2b). This decrease was moderate and depicted only by careful comparison with fat or renal cortex signal intensity, i.e., splenic MR signal intensity was lower than those of fat and renal cortex, but remained markedly higher than that of striated muscle. Percent contrast of spleen to other tissues and relaxation times of the spleen were not significantly different from those of normal spleen
(Tables I-III). Splenomegaly was demonstrated in two patients. In one, an associated splenic infarction was also depicted (Fig. 2a and b). In four of the eight patients with severe iron overload, a decrease in splenic MR signal intensity was observed on Tl-predominant images. A decrease in signal intensity was observed on TZpredominant images in all eight patients. This latter decrease was either marked (signal intensity of spleen lower than that of muscle) (Fig. 3) or moderate (signal intensity of spleen lower than those of fat and renal cortex but higher than that of muscle) (Fig. 4). Comparison with liver signal intensity was not reliable because the liver was also of abnormal signal intensity (Fig. 5a and b).
Fig. 2. Splenic infarction associated with moderate splenic iron overload (0.35 T field strength). (a) Tl-predominant (SE 500/30) image shows no definite abnormality in signal intensity. Splenic infarction (I) in the anterior aspect of the spleen cannot be differentiated. (b) TZ-predominant (SE 2000/60) image demonstrates clearly the high signal intensity infarction (I). The rest of the spleen shows a moderately decreased signal intensity best evaluated by comparison with adjacent fat.
Fig. 3. Severe splenic iron overload (0.35 T field strength). TZ-predominant (SE 2000/60) image shows the spleen (S) is enlarged and that both spleen and liver (L) are of markedly low signal intensity.
On Tl-predominant images, contrasts between spleen and renal cortex (p < 0.001) and between spleen and fat (p < 0.05) were significantly increased, reflecting a decrease of splenic signal intensity (Table I). On T2predominant images, contrasts between spleen and renal cortex and spleen and fat were also significantly increased (p < 0.001) (Table II). These marked increases in contrast also reflected a marked decrease of splenic signal intensity. On both Tl- and T2-predominant images, contrast between spleen and liver was markedly variable because the liver was also involved by iron overload, as indicated by the large standard deviations (Tables I and II). Both Tl- (p < 0.005) and T2-relaxation times (p < 0.001) were significantly lower than in patients with no iron overload (Table III).
Fig. 4. Splenic abscess associated with severe splenic iron overload (0.35 T field strength). TZ-predominant (SE 2000/60) image shows the decrease in splenic (S) signal intensity to be moderate, but allows the differentiation of the higher signal intensity abscess (arrows).
Splenomegaly was observed in 3 patients. In one, the associated abscess was depicted (Fig. 4). As with normal spleen, inhomogeneity of splenic tissue due to radiofrequency-field inhomogeneity was frequently observed regardless of the degree of iron overload. Discussion
In iron overload, reticuloendothelial iron, as in the spleen, is mostly in the form of hemosiderin, whereas parenchymal iron, as in the liver, is mostly in the form of ferritin [ 111. The complexed iron within hemosiderin
Fig. 5. Severe splenic iron overload (1.5 T field strength). (a) Coronal Tl-predominant (SE 600/25) image shows the signal intensity of the spleen (S) to be similar to that of the liver (L) and markedly lower than that of fat, renal cortex, or the muscle. (b) Transverse TZ-predominant (SE 2000/60) image. Both spleen (S) and liver (L) remain of very low signal intensity.
103
or ferritin is primarily in the oxidized ferric (Fe+ ‘) state. This ferric iron exhibits paramagnetic properties. Several animal and human studies [5-81 have shown alterations in MR signal intensity due to iron overload both in vitro and in vivo. In vitro, the presence of ferric iron shortens both Tl- and T2-relaxation times [ 51; the shortening in Tl is greater at low concentrations of ferric iron and the shortening in T2 is greater at high concentrations. Therefore, in vitro low-concentration ferric solutions are reflected as high signal intensity and high-concentration ferric solutions are reflected as low signal intensity. In in-vivo animal and human studies [ 6-81, only low signal intensity was reported in hepatic iron overload. In rats, Stark et al. [6] found that liver iron content correlated with decrease in MR signal intensity; they also were able to detect low levels of iron in the liver. In humans, low MR signal intensity of the liver relative to an extreme decrease in T2 and a moderate decrease in Tl of the liver were found [ 71. Hernandez et al. [ 81 in a study of 15 patients with transfusional hemosiderosis were only able to distinguish patients with very high concentrations of iron in the liver:. Adler et al. [ 121 reported generalized diminished splenic MR signal intensity in 13 patients with splenic iron overload due to sickle-cell disease. However, correlation between MR and pathological findings was not made. In sickle-cell disease, in addition to iron deposition, microscopic perivascular and parenchymal calcifications and tissue fibrosis are found in the splenic tissue [ 131. Therefore calcifications and fibrosis may also contribute to the diminished splenic MR signal intensity
[=I* In all our patients with mild iron overload, no decrease in splenic MR signal intensity was demonstrated. In our six patients with moderate iron overload, the decrease of splenic MR signal intensity was observed in only three, and in those three it was only present on T2-predominant images. At 0.35 T, MR imaging is not sensitive to mild or moderate splenic iron overload. As spin dephasing by paramagnetic substances is field dependent, detection of splenic iron overload should be easier with a high field magnet. In our study, MR at 1.5 T was performed in only five patients. In all our patients with severe iron overload, the decrease of splenic MR signal intensity was shown on T2-predominant images, but in only half these patients was the decrease visible on Tl-predominant images. No increases of splenic MR signal intensity occurred, even in mild iron overload. Therefore, iron overload in RES cells of the spleen appears to have the same result on MR images as iron overload in parenchymal cells of the liver.
For visual evaluation of splenic signal intensity, comparison with other tissues particularly the renal cortex and adipose tissue is a helpful internal standard. Slight inhomogeneity of splenic tissue should not be considered a finding in determining iron overload, because such inhomogeneity due to radiofrequency-field inhomogeneity is frequently observed in normal spleens. As previously reported [6] and observed in our study, MR signal intensity values and calculation of relaxation times have limited precision; standard deviations of these measurements were large in both normal and abnormal spleens. Nevertheless, we did find a significant decrease in splenic MR signal intensity and in Tl- and TZrelaxation,times in the spleens of patients with severe iron overload. These results agree with studies of iron overload in the liver [5-71. In a recent in-vitro study of thalassemic spleens with various degrees of iron overload, Gomori et al. [ 141 found a decrease in Tl- and TZrelaxation times, but only the latter decrease was significant. In our patients with severe iron overload, decrease of splenic T2-relaxation time was more marked than the decrease in Tl, but both decreases were significant. A significant decrease of Tl-relaxation time has also been reported [ 71 for livers with severe iron overload. This marked decrease in splenic TZrelaxation time was reflected as a decrease in splenic MR signal intensity on T2-predominant images. We also saw a decrease in splenic MR signal intensity on our Tl-predominant images, because the markedly decreased T2-relaxation time contributed to the T 1-predominant image. In severe iron overload, the accurate estimation of splenic size, easily obtained with MR imaging, is useful when discussing the indication of splenectomy. However, accurate estimation of splenic size is also easily obtained with ultrasonography or computed tomography. In a preliminary study [ 151, MR imaging was found to be a nonsensitive technique for detecting focal lesions of the spleen. However, such detection might be easier in spleens with iron overload [ 12, 151. In our study, three of the four focal lesions found at surgery were depicted on MR images. The abscess not depicted by MR imaging was in a spleen with only mild iron overload. In some other of our cases, contrast between the focal lesion and spleens that showed diminished MR signal intensity was increased, iron served as a natural contrast agent. In summary, iron overload in the reticuloendothelial cells of the spleen appears to have the same result on MR images as iron overload in the parenchymal cells of the liver. Though sensitive for severe splenic iron overload, MR imaging is not sensitive to mild or moderate
104
splenic iron overload. In addition, precise quantification of iron levels is not feasible. Therefore, we believe that MR imaging should not be used to diagnose splenic iron overload. MR imaging detection of focal splenic lesions may be facilitated if splenic iron overload is also present.
References Jacobs A, Worwood M. Iron metabolism, iron deficiency and overload. In: Hardisty RM, Weatherall DJ, eds. Blood and its disorders, 2nd Edn. Oxford: Blackwell Scientific, 1982; 149-197. Powell LW, Halliday JW. Idiopathic haemochromatosis. In: Jacobs A, Worwood M, eds. Iron in biochemistry and medicine, II. London; Academic Press, 1980; 461-498. Rao KV, Anderson WR. Hemosiderosis and hemochromatosis in renal transplant recipients. Am J Nephrol 1985; 5: 419-430. Bomford A, Corrigall A, Walker RJ, Williams R. Iron metabolism in haemochromatosis with reference to a chelatable iron pool and changes in iron absorption. In: Kief H, ed. Iron metabolism and its disorders. Proceedings of the third workshop conference Hoechst, Schloss Reisensburg, 6-9 April, 1975. New York: Elsevier, 1975; 21 l-220. Brasch RC, Wesbey GE, Gooding CA, Koerper MA. Magnetic resonance imaging of transfusional hemosiderosis complicating thalassemia major. Radiology 1984; 150: 767-771.
6 Stark DD, Bass NM, Moss AA, et al. Nuclear magnetic resonance imaging of experimentally induced liver disease. Radiology 1983; 148: 743-751. 7 Stark DD, Moseley ME, Bacon BR, et al. Magnetic resonance imaging and spectroscopy of hepatic iron overload. Radiology 1985; 154: 137-142. 8 Hernandez RJ, Samaik SA, Lande I, et al. MR evaluation of liver iron overload. J Comput Assist Tomogr 1988; 12: 91-94. 9 Roth VO, Jasinski B, Van Bidder H. Das Gewebeeisen beim Menschen bei normalen und pathologischen Zustiinden. Helv Med Acta 1951; 18: 159-174. 10 Greenberg M. Spleen. In: Greenberg M, ed. Essentials of body computed tomography. Philadelphia: W.B. Saunders, 1983; 178-199. 11 Matsuno T, Mori M, Awai M. Distribution of ferritin and hemosiderin in the liver, spleen and bone marrow of normal, phlebotomized and iron overloaded rats. Acta Med Okayama 1985; 39: 347-360. 12 Adler DD, Glazer GM, Aisen AM. MRI of the spleen: normal appearance and findings in sickle-cell anemia. AJR 1986; 147: 843-845. 13 Diggs LW. Siderotibrosis of the spleen in sickle cell anemia. JAMA 1935; 104: 538-541. 14 Gomori JM, Grossman RI, Drott HR. MR relaxation times and iron content of thalassemic spleens: an in vitro study. AJR 1988; 150: 567-569. 15 Hahn PF, Weissleder R, Stark DD, Saini S, Elizondo G, Ferrucci JT. MR imaging of focal splenic tumors. AJR 1988; 150: 823-827.
98
EURRAD
00024
Magnetic resonance imaging of splenic iron overload Lionel ArrivC*, Siegfried Thurnher **, Hedvig Hricak and David C. Price Department of Radiology, Universityof Cal$omia, School of Medicine, San Francisco, CA U.S.A. (Received 30 July 1989; revised version received 25 October
Key words: Magnetic resonance,
comparative
1989; accepted
study; Magnetic resonance, spleen; Magnetic resonance; iron, magnetic resonance
28 October
resonance,
1989)
iron overload;
Spleen, magnetic
Abstract The value of magnetic resonance (MR) imaging in assessing iron overload in the spleen was retrospectively investigated in 40 consecutive patients. MR appearance, measure of signal intensity and Tl- and TZrelaxation times were correlated with the histologically determined level of iron in the spleen in each patient. Histologic examination revealed no iron overload in 19 patients, mild iron overload in seven, moderate iron overload in six, and severe iron overload in eight. All 19 patients with no splenic iron overload and 11 of the other 21 patients with splenic iron overload were correctly identified by MR imaging (sensitivity 52%) specificity loo%, accuracy 75 %). Splenic iron overload was diagnosed when a decrease of signal intensity of the spleen compared with those of adipose tissue and renal cortex was demonstrated. MR images demonstrated all eight cases of severe, three of the six cases of moderate, and none of the seven cases of mild iron overload. Only spleens with severe iron overload had a significant mean decrease in signal intensity and Tl- and TZ-relaxation times. Although specific, MR imaging is poorly sensitive to splenic iron overload.
Introduction The term ‘iron overload’ refers to an increase in body iron stores. Iron can be deposited either in parenchymal cells, particularly of the liver, or in reticuloendothelial cells, particularly of the spleen, bone marrow and liver, though there is often deposition in both [ 11. Iron overload can result from an abnormal increase in the amount of iron absorbed, parenteral administration of iron or blood transfusions [2]. Detection of mild iron overload is necessary for early diagnosis and quantiflcation of the iron level in severe iron overload is necessary to monitor therapy [ 31. Blood determination of ferritin levels is considered [4] the most reliable noninvasive method of assessing iron stores. Histologic examination of biopsy specimens, however, remains the procedure of choice [4]. *
Present address: Department of Radiology; Hopital Beaujon, 92118 Clichy Cedex, France. ** Present address: Departement Diagnostische Radiologie, CH-8091, Ztirich, Switzerland. Address for reprints: Lionel Arrive, M.D., Service de Radiologie, HBpital Beaujon, 92118 Clichy Cedex, France. 0720-048X/90/$03.50
0 1990 Elsevier Science Publishers
Several reports [ 5-81 have described the decreased signal intensity of the liver on magnetic resonance (MR) images of animal and human livers with iron overload. The present, retrospective, in vivo study was designed to determine whether MI? imaging can be used to quantitatively assess iron overload in the spleen. The spleen’s MR appearance, signal intensity, and Tl- and TZrelaxation times were correlated with histologically determined iron levels. Patients and Methods Patients Forty consecutive patients, 23 female and 17 male (aged 13-78 years, mean age 43 years), were retrospectively studied. Twenty-six of the 40 patients had been referred for MR imaging because of clinically suspected iron overload (primary hemochromatosis in six, dietary hemosiderosis in three, transfusional hemosiderosis in 17). The other 14 patients had undergone MR imaging for Hodgkin’s disease (seven patients), non-Hodgkin’s lymphoma (three patients), splenic abscess (three patients), and splenic infarction (one patient).
B.V. (Biomedical Division)
99
Histologic correlation of iron deposition levels was available in all 40 patients. Histologic specimens had been obtained of the spleen in 30 patients (22 by means of surgery, six by autopsy, and two by biopsy). Histologic specimens had been obtained by biopsy of the liver (five patients) or bone (five patients) in the other ten patients. Analysis of histologicfindings The histologic slides from the 40 patients were reviewed in conference by two of us to determine the degree of iron deposition. The ten patients with only liver or bone marrow tissue available for histologic analysis were included in our study because the amount of iron deposition in the reticuloendothelial system (RES) of the liver and bone marrow parallels that in the RES of splenic tissue [9]. Iron deposition, visible on hematoxylin and eosin stained tissue, was scored and categorized into four groups. A score of 0 indicated no iron deposition, ‘1 + ’ indicated mild iron deposition (perivascular only or less than two RES cells containing hemosiderin granules as seen per high-power-field (HPF) magnification of 400), ‘2 + ’ indicated moderate iron overload (two to ten RES cells/HPF magnification of 400), and ‘3’ indicated severe iron overload (more than ten RES cells/HPF magnification of 400). A4R imaging MR imaging had been performed in all 40 patients with a Diasonics MT/S system operated at 0.35 T. MR imaging had been additionally performed in live patients with a General Electric Signa system operated at 1.5 T. At 0.35 T, multislice spin-echo (SE) images were obtained with repetition times (TRs) of 500-2000 ms and echo times (TEs) of 30 ms for the first echo and 60 ms for the second echo. Both transverse Tl-predominant (SE 500/30) and transverse T2-predominant (SE 2000/60) images were obtained in all patients. In addition, Tl-predominant coronal images were obtained in three of the patients and Tl-predominant sag&al images in nine. In the five patients imaged at 1.5 T, both transverse Tl-predominant (SE 600/25) and transverse T2-predominant (SE 2000/60) images were obtained. Tl-predominant coronal images were obtained in two of these live. Analysis of iUR images The MR images were evaluated in conference by two of us who did not evaluate the histologic slides. These observers had no knowledge of clinical information or histologic results. Analyzed were: (a) the signal intensity of the spleen compared with those of the liver, renal cortex, striated muscle, and adjacent subcutaneous
adipose tissue on Tl- and T2-predominant images. (b) Percent contrast of the spleen compared to liver, renal cortex and adjacent, subcutaneous adipose tissue on Tl- and T2-predominant images at 0.35 T. (c) Tland T2-relaxation times of the spleen at 0.35 T. (d) Homogeneity of the spleen’s signal intensity on Tland T2-predominant images. (e) Longest dimension of the spleen-splenomegaly was diagnosed when splenic length was greater than 15 cm [ lo]. Signal intensities of the spleen, liver, renal cortex and subcutaneous adipose tissue were measured with use of the cursor to define the regions of interest (ROIs). The percentage of contrast between the spleen and the liver, renal cortex, or adipose tissue was determined as follows : % contrast
=
(signal intensity of spleen - signal intensity of the tissue) x loo signal intensity of spleen + signal intensity of the tissue
After all MR measurements were made of all patients, the MR results were compared to the histologic results. Sensitivity, specificity, and accuracy, were calculated. All quantitative data were expressed as group mean k 1 standard deviation (S.D.). The comparison of the measurements between the different groups was performed using the Student’s t-test for unpaired data. Results Comparison of histologicand MR findings Histologic determination of iron deposition in all 40 specimens revealed no iron overload in 19, mild iron overload in seven, moderate iron overload in six, and severe iron overload in eight. From MR imaging findings, all 19 patients with no splenic iron overload were identified as such. Of the 21 patients with iron overload, only 11 patients were correctly identified. Splenic iron overload was diagnosed when a decrease of signal intensity of the spleen compared with those of adipose tissue and renal cortex was demonstrated. Overall sensitivity, specificity and accuracy, therefore, were 52%) 100% and 75 %, respectively. However, results differed with respect to the degree of iron overload: MR image analysis identified all of the eight severe, three of the six moderate, and none of the seven mild cases of iron overload. In addition to the histologic findings of iron overload described above, 26 of the 30 splenic tissue specimens obtained at surgery revealed no abnormality, three showed splenic abscess, and one showed splenic infarc-
100 TABLE I Percentage
of contrast on Tl-predominant
images of splenic tissue with various levels of iron
Degree of iron deposition (severity)
Number of patients
None (0) Mild (1+) Moderate (2 + ) Severe (3 + )
19 7 6 8
Percentage
contrast
(mean + S.D.)
spleen/liver
spleen/renal
-l* 8 -8k23 +6*27 +21 k 89
+7*13 +9& 19 +2* 17 -23 f 12*
cortex
spleen/fat -85 f 14 -8Ok 8 - 89 f 23 - 106 f 31**
* = p < 0.001. ** = p < 0.05.
TABLE II Percentage
of contrast
on T2-predominant
images of splenic tissue with various levels of iron
Degree of iron deposition (severity)
Number of patients
None (0) Mild (1+) Moderate (2 + ) Severe (3 + )
19 7 6 8
Percentage
contrast
(mean f SD.)
spleen/liver
spleen/renal
+60 + 16 +54 + 11 +50*23 +29 k 69
-7* 13 -6+ 12 -23 f 35 - 78 f 26+
cortex
spleen/fat -26 + 22 -13+ 7 -43 f 29 - 103 f 37*
* = p < 0.001.
tion. These last four specimens had also demonstrated iron overload. Two of the three splenic abscesses and the splenic infarction were depicted on MR images. MR appearance of normal spleen (no iron overload)
On Tl-predominant images of the 19 spleens with no iron overload, MR signal intensity of the spleen was similar to or slightly lower than that of the liver, similar to that of the renal cortex, and markedly lower than that
TABLE III Tl- and TZ-relaxation times of splenic tissue with various levels of iron Degree of iron deposition (severity)
Number of patients
None (0) Mild (1+) Moderate (2 + ) Severe (3 + )
19 7 6 8
* = p < 0.005. ** = p < 0.001.
Relaxation time (mean + S.D.) Tl 802 859 890 597
T2 + f k *
157 128 249 97*
75* 7 81k 9 67 + 14 39 * 19**
of adipose tissue. On T2-predominant images, MR signal intensity of the spleen was markedly higher than that of the liver, similar to that of the renal cortex, and approaching that of adipose tissue. These findings on Tl- and TZpredominant images were regardless of field strength. The percent contrast of spleen to the other tissues studied and the relaxation times of the spleens at 0.35 T are given in Tables I-III. Nine patients demonstrated inhomogeneity of the normal splenic tissue on both Tl- and T2-predominant images (Fig. la and b). The inhomogeneity in signal intensity was between the anterior and posterior aspects of the spleen and was due to radiofrequency-field inhomogeneity. Normal spleen varied in size, but in none did the length exceed 15 cm. MR appearance of splenic iron overload
In the seven patients with mild iron overload, no change in splenic MR signal intensity was observed on either Tl- or T2-predominant images. Percent contrast in relation to other tissue and relaxation times of the spleen were not significantly different from those of normal spleen (Tables I-III). Splenomegaly was demonstrated in one spleen that had an associated splenic abscess. The abscess was visible as a high-intensity area
Fig. 1. Normal spleen showing image degradation due to radiofrequency-field inhomogeneity (0.35 T field strength). (a) Tl-predominant (SE 500/30) and (b), T2-predominant (SE 2000/60) images show inhomogeneous signal intensity within normal tissues. Low intensity areas are seen in the lateral segment of the liver (L), anterior aspect of the spleen (S), and adjacent subcutaneous fat (arrows).
of TZpredominant images. Another abscess in another spleen in this group was not depicted by MR. In the six patients with moderate iron overload, no change in splenic MR signal intensity was visible on Tl-predominant images (Fig. 2a). A decrease in splenic MR signal intensity was observed on T2-predominant images in three of the six patients (Fig. 2b). This decrease was moderate and depicted only by careful comparison with fat or renal cortex signal intensity, i.e., splenic MR signal intensity was lower than those of fat and renal cortex, but remained markedly higher than that of striated muscle. Percent contrast of spleen to other tissues and relaxation times of the spleen were not significantly different from those of normal spleen
(Tables I-III). Splenomegaly was demonstrated in two patients. In one, an associated splenic infarction was also depicted (Fig. 2a and b). In four of the eight patients with severe iron overload, a decrease in splenic MR signal intensity was observed on Tl-predominant images. A decrease in signal intensity was observed on TZpredominant images in all eight patients. This latter decrease was either marked (signal intensity of spleen lower than that of muscle) (Fig. 3) or moderate (signal intensity of spleen lower than those of fat and renal cortex but higher than that of muscle) (Fig. 4). Comparison with liver signal intensity was not reliable because the liver was also of abnormal signal intensity (Fig. 5a and b).
Fig. 2. Splenic infarction associated with moderate splenic iron overload (0.35 T field strength). (a) Tl-predominant (SE 500/30) image shows no definite abnormality in signal intensity. Splenic infarction (I) in the anterior aspect of the spleen cannot be differentiated. (b) TZ-predominant (SE 2000/60) image demonstrates clearly the high signal intensity infarction (I). The rest of the spleen shows a moderately decreased signal intensity best evaluated by comparison with adjacent fat.
Fig. 3. Severe splenic iron overload (0.35 T field strength). TZ-predominant (SE 2000/60) image shows the spleen (S) is enlarged and that both spleen and liver (L) are of markedly low signal intensity.
On Tl-predominant images, contrasts between spleen and renal cortex (p < 0.001) and between spleen and fat (p < 0.05) were significantly increased, reflecting a decrease of splenic signal intensity (Table I). On T2predominant images, contrasts between spleen and renal cortex and spleen and fat were also significantly increased (p < 0.001) (Table II). These marked increases in contrast also reflected a marked decrease of splenic signal intensity. On both Tl- and T2-predominant images, contrast between spleen and liver was markedly variable because the liver was also involved by iron overload, as indicated by the large standard deviations (Tables I and II). Both Tl- (p < 0.005) and T2-relaxation times (p < 0.001) were significantly lower than in patients with no iron overload (Table III).
Fig. 4. Splenic abscess associated with severe splenic iron overload (0.35 T field strength). TZ-predominant (SE 2000/60) image shows the decrease in splenic (S) signal intensity to be moderate, but allows the differentiation of the higher signal intensity abscess (arrows).
Splenomegaly was observed in 3 patients. In one, the associated abscess was depicted (Fig. 4). As with normal spleen, inhomogeneity of splenic tissue due to radiofrequency-field inhomogeneity was frequently observed regardless of the degree of iron overload. Discussion
In iron overload, reticuloendothelial iron, as in the spleen, is mostly in the form of hemosiderin, whereas parenchymal iron, as in the liver, is mostly in the form of ferritin [ 111. The complexed iron within hemosiderin
Fig. 5. Severe splenic iron overload (1.5 T field strength). (a) Coronal Tl-predominant (SE 600/25) image shows the signal intensity of the spleen (S) to be similar to that of the liver (L) and markedly lower than that of fat, renal cortex, or the muscle. (b) Transverse TZ-predominant (SE 2000/60) image. Both spleen (S) and liver (L) remain of very low signal intensity.
103
or ferritin is primarily in the oxidized ferric (Fe+ ‘) state. This ferric iron exhibits paramagnetic properties. Several animal and human studies [5-81 have shown alterations in MR signal intensity due to iron overload both in vitro and in vivo. In vitro, the presence of ferric iron shortens both Tl- and T2-relaxation times [ 51; the shortening in Tl is greater at low concentrations of ferric iron and the shortening in T2 is greater at high concentrations. Therefore, in vitro low-concentration ferric solutions are reflected as high signal intensity and high-concentration ferric solutions are reflected as low signal intensity. In in-vivo animal and human studies [ 6-81, only low signal intensity was reported in hepatic iron overload. In rats, Stark et al. [6] found that liver iron content correlated with decrease in MR signal intensity; they also were able to detect low levels of iron in the liver. In humans, low MR signal intensity of the liver relative to an extreme decrease in T2 and a moderate decrease in Tl of the liver were found [ 71. Hernandez et al. [ 81 in a study of 15 patients with transfusional hemosiderosis were only able to distinguish patients with very high concentrations of iron in the liver:. Adler et al. [ 121 reported generalized diminished splenic MR signal intensity in 13 patients with splenic iron overload due to sickle-cell disease. However, correlation between MR and pathological findings was not made. In sickle-cell disease, in addition to iron deposition, microscopic perivascular and parenchymal calcifications and tissue fibrosis are found in the splenic tissue [ 131. Therefore calcifications and fibrosis may also contribute to the diminished splenic MR signal intensity
[=I* In all our patients with mild iron overload, no decrease in splenic MR signal intensity was demonstrated. In our six patients with moderate iron overload, the decrease of splenic MR signal intensity was observed in only three, and in those three it was only present on T2-predominant images. At 0.35 T, MR imaging is not sensitive to mild or moderate splenic iron overload. As spin dephasing by paramagnetic substances is field dependent, detection of splenic iron overload should be easier with a high field magnet. In our study, MR at 1.5 T was performed in only five patients. In all our patients with severe iron overload, the decrease of splenic MR signal intensity was shown on T2-predominant images, but in only half these patients was the decrease visible on Tl-predominant images. No increases of splenic MR signal intensity occurred, even in mild iron overload. Therefore, iron overload in RES cells of the spleen appears to have the same result on MR images as iron overload in parenchymal cells of the liver.
For visual evaluation of splenic signal intensity, comparison with other tissues particularly the renal cortex and adipose tissue is a helpful internal standard. Slight inhomogeneity of splenic tissue should not be considered a finding in determining iron overload, because such inhomogeneity due to radiofrequency-field inhomogeneity is frequently observed in normal spleens. As previously reported [6] and observed in our study, MR signal intensity values and calculation of relaxation times have limited precision; standard deviations of these measurements were large in both normal and abnormal spleens. Nevertheless, we did find a significant decrease in splenic MR signal intensity and in Tl- and TZrelaxation,times in the spleens of patients with severe iron overload. These results agree with studies of iron overload in the liver [5-71. In a recent in-vitro study of thalassemic spleens with various degrees of iron overload, Gomori et al. [ 141 found a decrease in Tl- and TZrelaxation times, but only the latter decrease was significant. In our patients with severe iron overload, decrease of splenic T2-relaxation time was more marked than the decrease in Tl, but both decreases were significant. A significant decrease of Tl-relaxation time has also been reported [ 71 for livers with severe iron overload. This marked decrease in splenic TZrelaxation time was reflected as a decrease in splenic MR signal intensity on T2-predominant images. We also saw a decrease in splenic MR signal intensity on our Tl-predominant images, because the markedly decreased T2-relaxation time contributed to the T 1-predominant image. In severe iron overload, the accurate estimation of splenic size, easily obtained with MR imaging, is useful when discussing the indication of splenectomy. However, accurate estimation of splenic size is also easily obtained with ultrasonography or computed tomography. In a preliminary study [ 151, MR imaging was found to be a nonsensitive technique for detecting focal lesions of the spleen. However, such detection might be easier in spleens with iron overload [ 12, 151. In our study, three of the four focal lesions found at surgery were depicted on MR images. The abscess not depicted by MR imaging was in a spleen with only mild iron overload. In some other of our cases, contrast between the focal lesion and spleens that showed diminished MR signal intensity was increased, iron served as a natural contrast agent. In summary, iron overload in the reticuloendothelial cells of the spleen appears to have the same result on MR images as iron overload in the parenchymal cells of the liver. Though sensitive for severe splenic iron overload, MR imaging is not sensitive to mild or moderate
104
splenic iron overload. In addition, precise quantification of iron levels is not feasible. Therefore, we believe that MR imaging should not be used to diagnose splenic iron overload. MR imaging detection of focal splenic lesions may be facilitated if splenic iron overload is also present.
References Jacobs A, Worwood M. Iron metabolism, iron deficiency and overload. In: Hardisty RM, Weatherall DJ, eds. Blood and its disorders, 2nd Edn. Oxford: Blackwell Scientific, 1982; 149-197. Powell LW, Halliday JW. Idiopathic haemochromatosis. In: Jacobs A, Worwood M, eds. Iron in biochemistry and medicine, II. London; Academic Press, 1980; 461-498. Rao KV, Anderson WR. Hemosiderosis and hemochromatosis in renal transplant recipients. Am J Nephrol 1985; 5: 419-430. Bomford A, Corrigall A, Walker RJ, Williams R. Iron metabolism in haemochromatosis with reference to a chelatable iron pool and changes in iron absorption. In: Kief H, ed. Iron metabolism and its disorders. Proceedings of the third workshop conference Hoechst, Schloss Reisensburg, 6-9 April, 1975. New York: Elsevier, 1975; 21 l-220. Brasch RC, Wesbey GE, Gooding CA, Koerper MA. Magnetic resonance imaging of transfusional hemosiderosis complicating thalassemia major. Radiology 1984; 150: 767-771.
6 Stark DD, Bass NM, Moss AA, et al. Nuclear magnetic resonance imaging of experimentally induced liver disease. Radiology 1983; 148: 743-751. 7 Stark DD, Moseley ME, Bacon BR, et al. Magnetic resonance imaging and spectroscopy of hepatic iron overload. Radiology 1985; 154: 137-142. 8 Hernandez RJ, Samaik SA, Lande I, et al. MR evaluation of liver iron overload. J Comput Assist Tomogr 1988; 12: 91-94. 9 Roth VO, Jasinski B, Van Bidder H. Das Gewebeeisen beim Menschen bei normalen und pathologischen Zustiinden. Helv Med Acta 1951; 18: 159-174. 10 Greenberg M. Spleen. In: Greenberg M, ed. Essentials of body computed tomography. Philadelphia: W.B. Saunders, 1983; 178-199. 11 Matsuno T, Mori M, Awai M. Distribution of ferritin and hemosiderin in the liver, spleen and bone marrow of normal, phlebotomized and iron overloaded rats. Acta Med Okayama 1985; 39: 347-360. 12 Adler DD, Glazer GM, Aisen AM. MRI of the spleen: normal appearance and findings in sickle-cell anemia. AJR 1986; 147: 843-845. 13 Diggs LW. Siderotibrosis of the spleen in sickle cell anemia. JAMA 1935; 104: 538-541. 14 Gomori JM, Grossman RI, Drott HR. MR relaxation times and iron content of thalassemic spleens: an in vitro study. AJR 1988; 150: 567-569. 15 Hahn PF, Weissleder R, Stark DD, Saini S, Elizondo G, Ferrucci JT. MR imaging of focal splenic tumors. AJR 1988; 150: 823-827.
