321 Value of Lipid- and Water- Suppression MR Images in Distinguishing Between Blood and Lipid Within Ovarian Masses Ruben Kier1 The distinction between blood and lipid in ovarian masses on MR imaging is important Robert C. Smith in the differential diagnosis of these lesions. However, this Is often difficult on routine Shirley M. McCarthy MR images because both blood and lipid within tumors can have the same signal intensfty as subcutaneous fat. Accordingly, we studied the value of lipid- and watersuppression MR images in making this distinction in 16 patients (21 lesions). As proved by surgery (six patients) or laparoscopy (10 patients), there were 16 endometriomas, one hemorrhagic leiomyosarcoma, and four lipid-containing mature cystic teratomas. The signal intensity in all 17 hemorrhagic lesions was greater than that of subcutaneous fat on lipid-suppression images and less than that of fat on water-suppression Images. This compared with the signal intensity of the four lesions that contained lipid, in which the signal intensity was similar to that of subcutaneous fat on both the lipid- and watersuppression images. Thus, the lipid- and water-suppression MR images allowed an accurate distinction between the two. Our experience suggests that the appearance of blood and lipid in ovarian tumors is sufficiently different on lipid- and water-suppression MR images to allow an accurate distinction between the two. The two techniques should be useful in the differential diagnosis of such lesions by MR imaging. AJR 158:321-325, February 1992 Ovarian masses have been difficult to characterize with MR imaging, since most appear hypointense relative to subcutaneous fat on Ti -weighted images and hyperintense relative to fat on T2-weighted images [1]. Exceptions to this signal behavior include both hemorrhagic (e.g., endometnomas) and lipid-containing (e.g., mature cystic teratomas) lesions. In these cases, the MR signal may be entirely or partly isointense relative to subcutaneous fat on both Ti - and T2-weighted images [2-5]. Consequently, these two abnormalities may be difficult to distinguish with conventional MR techniques. Accordingly, we studied the value of lipid- and watersuppression MR imaging to make this distinction. Subjects and Methods Over a 1 2-month period, we identified 24 women with a total of 30 adnexal masses that had signal intensities that were isointense relative to subcutaneous fat on Ti -weighted MR images. Sixteen of these patients (21 lesions), those for whom surgical or laparoscopic findings were available, were included in the study. Surgery confirmed the presence of sebum Received July 29, 1 991 ; accepted after revision in four mature cystic teratomas (four patients) and the presence of blood in one cystic September 17, 1991. leiomyosarcoma (one patient) and two endometriomas (one patient). Laparoscopy performed I Allauthors: Department of Radiology, Yale Uni- after MR imaging confirmed the presence of blood in two endometriornas (two patients). versity School of Medicine, 333 Cedar St., New Twelve lesions in eight patients with laparoscopic diagnoses of endometnosas that preceded Haven, CT 06510. Address reptint requests to A. MR imaging were presumed to represent endometriomas containing blood. Kier. Imaging was performed on a 1.5-T scanner (Signa, General Electric, Milwaukee, WI) with 0361-803X/92/1582-0321 a spin-echo (SE) technique, 28-cm field of view, two excitations, 1 28 x 256 matrix, and 5- C American Aoentgen Ray Society mm slice thickness with a 2.5-mm gap. All 16 patients were studied with Ti-weighted SE
322 KIER ET AL. AJA:158, February 1992 I It. a s,. 4s a. S m J, images, 400-800/i 1-20 (TRITE), in either the coronal plane only (seven patients), axial plane only (five), or both planes (four). Axial T2-weighted images were obtained either as part of a conventional SE 1900-2000/20,80 sequence (1 3 of 16 patients) or as a fast SE sequence (2900/1 26) (three of 1 6 patients). Supplementary Ti - weighted SE images (400-700/i i -20) were obtained first (1 5 of 16 patients) with presaturation offat protons (lipid suppression) and then repeated (1 3 of i 6 patients) with presaturation of water protons (water suppression). To selectively suppress the signal from lipid, a non-slice-selective Fig. 1.-Endometrioma of left ovary in 35-year-old woman. A and B, 3-cm ovarian mass (arrows) appears isointense compared with subcutaneous fat on both coronal Tiweighted (650/17) (A) and axial T2- weighted (2000/80) (B) MR images. C and D, Mass (arrows) becomes markedly hyperintense relative to fat on axial Ti-weighted (500/il) MR image with lipid suppression (C) and hypointense relative to fat on axial Tiweighted (500/il) MR image with water suppression (D). 900 pulse with a narrow bandwidth centered on the precessional frequency of lipid was applied just prior to the slice-selective 90#{176} pulse of the standard SE sequence. A spoiler gradient was then applied immediately after the lipid-selective excitation pulse. For suppression of water, the narrow bandwidth pulse was centered on the precessional frequency of water [6]. This technique yields two additional sets of Ti -weighted images: one set with the lipid signal suppressed and the second set with the water signal suppressed. These two sequences each require 2 mm of imaging time. Allowing for prescan time, both sets of images can be obtained in less than 10 mm. Images were interpreted independently by two experienced radiologists. Disagreements were settled by joint review and consensus. A hemorrhagic lesion was diagnosed if the signal intensity of the mass was hyperintense relative to subcutaneous fat on lipid-suppressed images and hypomntense compared with subcutaneous fat on water-suppressed images. Mature cystic teratoma was diagnosed if the mass remained isointense relative to fat on both water-suppressed and lipid-suppressed images. The standard Ti - and T2- weighted SE images of the adnexal lesions were assessed separately and independently for the presence of a hypointense rim (suggestive of hemosiderin or fibrous tissue around an endometrioma) [2], atypical chemical-shift artifact within or at the interface of the lesion (suggestive of cystic teratoma) [3], and relative signal intensity on T2- weighted Results images. Lipid- and water-suppression MR imaging enabled correct categorization of all 21 lesions in the 1 6 patients. The signal intensity of all 1 7 hemorrhagic lesions (1 6 endometriomas and one hemorrhagic cystic leiomyosarcoma) was hyperintense relative to fat on lipid-suppressed images and hypoin-
AJR:158, February 1992 MR OF BLOOD VS LIPID IN OVARIAN MASSES 323 Fig. 2.-Endometrioma of right ovary In 33-year-old woman. A and B, 5-cm ovarian mass (arrows) appears Isointense compared with subcutaneous fat on both Ti-weighted (500/20) (A) and T2-weighted (2900/ 126)(B) axial MR Images. C and D, Lesion (arrows) becomes markedly hyperintense relative to subcutaneous fat on axial Ti-weighted (600/20) MR Image with lipid suppression (C) and hypointense relative to fat on axial Ti-weighted (600/20) MR image with water suppression (0). tense relative to fat on water-suppressed images (Figs. 1 and 2). Only the hemorrhagic Ieiomyosarcoma and one endometrioma displayed a hypointense rim suggestive of hemosiderin. On routine T2-weighted images, the hemorrhagic lesions were isointense compared with fat in seven patients (eight endometriomas), slightly hypointense relative to fat in two patients (three endometriomas), and markedly hyperintense relative to fat in four patients (five endometriomas and one Ieiomyosarcoma). These numbers include one patient with bilateral endometriomas, of which one was isointense with fat and the other was hyperintense relative to fat on T2-weighted images. All four cases of mature cystic teratoma were correctly characterized as lipid-containing lesions on the basis of the presence of areas that remained isointense with subcutaneous fat on water-suppressed images (Figs. 3 and 4). All teratomas demonstrated suppression of lipid signal on lipidsuppressed images, remaining isointense with subcutaneous fat in three cases (Fig. 3) and becoming slightly hypointense relative to subcutaneous fat in one case (Fig. 4C). Two of the I -. r_n ---n h teratomas demonstrated chemical-shift artifact at the interface between tumor fat and adjacent tissues (Fig. 4B). Compared with subcutaneous fat, the signal intensity of teratomas on T2-weighted images was isointense in three cases and slightly hypointense in one case. Discussion Lipid-containing lesions have high signal intensity on Ti - weighted images because lipid has a short Ti relaxation time. Hemorrhagic lesions can display signal characteristics similar to those of lipid-containing lesions on both Ti - and T2- weighted images if the lesion contains methemoglobin, which shortens both Ti and T2 relaxation times [7]. Recent series have demonstrated signal intensity isointense with fat on Tiweighted images in 52 of 88 endometriomas [2] and 22 of 23 cystic teratomas [3]. In some cases, the diagnosis of endometrioma can be suggested if the lesion has a hypointense rim on both Ti -
324 KIER ET AL. AJR:158, February 1992 and T2-weighted MR images, although this feature is absent in the majority of lesions [2]. Only one endometrioma in this series and the hemorrhagic leiomyosarcoma had a border with low signal, suggesting hemosiderin deposition. Ovarian cystic teratomas can be suggested if there is atypical chemical-shift artifact due to lipid within the tumor, although this feature may be difficult to detect at lower field strengths [3], at standard bandwidths [8], or at curved interfaces [9]. Only two of four cystic teratomas in this series demonstrated chemical-shift artifact at their margins. Slight differences in signal intensity between lesion and subcutaneous fat on T2-weighted images did not prove helpful for lesion characterization. Although hemorrhagic lesions in two patients appeared slightly hypointense relative to subcutaneous fat on T2-weighted images, one cystic teratoma also appeared slightly hypointense relative to subcutaneous fat. Since the lipid and water composition of subcutaneous fat differs from that of the sebum in cystic teratoma, slight differences in signal intensity may be encountered [3] (Fig. 4C). Fig. 3.-Mature cystic teratoma of right ovary in 38-year-old woman. A-D, 4-cm ovarian mass (arrows) remains Isointense compared with subcutaneous fat on axial Ti-weighted (800/20) MR Image (A), T2-weighted (2000/80) MR image (B), Ti-weighted (600/29) MR image with lipid suppression (C), and Ti-weighted (700/il) MR image with water suppression (0). For ovarian masses that appear isointense with fat on Ti - weighted images, supplementary lipid- and water-suppression sequences are not always necessary to differentiate hemorrhagic from lipid-containing lesions. In this series, seven of i 7 hemorrhagic lesions could be characterized by either marked hypenntensity compared with fat on routine T2- weighted images or a hypointense rim. Two offour teratomas could be characterized by atypical chemical-shift artifacts. Thus, lipid and water suppression improved characterization of 57% of lesions (i 2/2i) in this study population. Lipid- and water-suppression MR imaging can differentiate between hemorrhagic lesions and cystic teratomas, because the source of high signal on Ti -weighted images is the water proton for hemorrhage vs the lipid proton for cystic teratoma. All cases in this series were differentiated successfully with these two techniques. Although these techniques can exclude lipid as a source of high signal on Ti -weighted images, persistence of high signal on lipid-suppressed Ti -weighted images will not be unique to endometrioma. Other hemorrhagic lesions, especially hemorrhagic cysts, may be indistin-
AJR:158, February 1992 MR OF BLOOD VS LIPID IN OVARIAN MASSES 325 Fig. 4.-Mature cystic teratoma in 28-year-old woman. A, 4-cm left ednexal mass (arrow) Is Isolntense compared with subcutaneous faton coronal Ti-weighted (500/ 20) MR image. B, On axial 12-weighted (1900/80) MR Image, left ovarian mass (straight arrow) appears isointense compared with subcutaneous fat, although slightly heterogeneous. Chemical-shift artifact may be present along superomedial and superolateral margins of mass (curved arrows), suggesting cystic teratoma. C, Anterior component (arrow) of ovarian lesion becomes slightly hypointense relative to subcutanaous fat on axial Ti-weighted (567/il) MR Image with lipid suppression, Indicating that lipid In this teratoma differs from lipid in subcutaneous fat. 0, Mass (arrow) remains Isointense compared with subcutaneousfaton axlal Ti-weighted (587/11) MR Image with water suppression. guishable from endometriomas [i]. Similarly, any lipid-contaming lesion may have signal behavior identical to that of cystic teratoma, although lipid is rare in other ovarian tumors. In this study, we routinely used both lipid-suppression and water-suppression pulse sequences. Although either Sequence alone may distinguish blood from lipid, use of both sequences subjectively improved diagnostic confidence. Because suppression of signal often was incomplete within lesions (Figs. i D, 20, and 3C), the lipid- and water-suppression images were complementary. The lipid-suppressed images accentuated the conspicuity of hemorrhagic lesions, whereas the water-suppressed images enhanced the conspi- Cuity of lipid-containing lesions. We conclude that the appearance of blood and lipid in ovarian tumors is sufficiently different on lipid- and watersuppression MR images to allow an accurate distinction between the two. The two techniques should be useful in the differential diagnosis of lesions that appear isointense to subcutaneous fat on routine MR images. REFERENCES 1. Mitchell DG, Mintz MC, Spritzer CE, et al. Adnexal masses: MA imaging observations at 1.5T, with US and CT correlation. Radiology 1987; 162:319-324 2. Zawin M, McCarthy 5, Scoutt L, Comite F. Endometriosis: appearance and detection at MA imaging. Radiology 1989;171 :693-696 3. Togashi K, Nishimura K, Itoh K, et al. Ovarian cystic teratomas: MA imaging. Radiology i987;162:669-673 4. Dooms GC, Hricak H, Tscholakoft D. Adnexal structures: MR imaging. Radiology i986;158:639-646 5. Nyberg DA, Porter BA, Olds MO, Olson DO, Andersen A, Wesby GE. MA imaging of hemorrhagic adnexal masses. J ComM Assist Tomogr i987;1 1:664-669 6. Keller PJ, Hunter WW, Schmalbrock P. Multisection fat-water imaging with chemical shift selective presaturation. Radiology 1987;164:539-541 7. Gomori JM, Grossman Al. Mechanisms responsiblefor the MR appearance and evolution of intracranial hemorrhage. RadioGraphics i988;8:427-439 8. Smith AS, Weinstein MA, Hurst GC, Defremer OR, Cole RA, Duchesneau PM. Intracranial chemical-shift arttfacts on MA images of the brain: observations and relation to sampling bandwidth. AiR i990;154: 1275-1283 9. Smith AC, Lange AC, McCarthy SM. Chemical-shift artdact: dependence on shape and orientation of the lipid-water interface. Radiology i99i; 181:225-229