Introduction	Choose Top of pageABSTRACTIntroduction <<Materials and MethodsResultsDiscussionReferencesCITING ARTICLES

Neuroblastoma is the most common extracranial solid malignancy in pediatric patients and is responsible for a large proportion of cancer deaths in children younger than 5 years [1, 2]. It is an embryonic tumor arising from the neural crest cells, which give rise to the adrenal medulla and the sympathetic nervous system [3]. The tumor is most frequently situated in the adrenal gland or anywhere else along the sympathetic nervous system chain [1]. It shows an infiltrative growth pattern with early metastatic disease, with metastases being revealed in one half to two thirds of the patients at presentation [4]. Imaging of neuroblastoma consists of sonography, CT, MRI, and radionuclide examinations such as scintigraphic bone scanning and metaiodobenzylguanidine (MIBG) scintigraphy [4, 5]. The therapeutic spectrum is supportive care with no treatment (stage IVS), definitive excision if possible, or chemotherapy before and after surgery partially combined with total-body irradiation, and followed by autologous bone marrow transplantation [6].

MIBG labeled with iodine-131 (¹³¹I) was developed by Wieland et al. [7–9] and Sisson et al. [10] at the University of Michigan as an imaging agent for tissues and tumors of the sympathetic nervous system, originally for the localization and treatment of pheochromocytomas [11]. Scintigraphy with ¹²³I- or ¹³¹I-labeled MIBG has become a well-established method in the diagnosis and staging of neuroblastoma [12] because of its high specificity, which is reported in the literature to be between 90% and 100% [6, 11, 13].

MRI has several advantages in the diagnosis of neuroblastoma: high sensitivity in detecting bone marrow abnormalities [14], lack of ionizing radiation, high intrinsic soft-tissue contrast resolution [5], depiction of internal structure [15], and exact definition of intraspinal tumor extension or diaphragmatic involvement of thoracic tumors [16]. All these factors are decisive, especially for operative planning.

In the literature, comparative studies about the diagnostic value of MIBG scintigraphy and MRI in staging neuroblastoma in pediatric patients have been published [4, 14]. The basis of any staging procedure is the detection of single tumor lesions. Therefore, our approach was to assess the diagnostic value of an integrated analysis with both modalities with the following aims: to assess the sensitivity and specificity of MIBG scintigraphy and MRI separately in the diagnosis and follow-up of single neuroblastoma lesions and to evaluate the benefit of a combined analysis with MRI and MIBG scintigraphy.

Materials and Methods	Choose Top of pageABSTRACTIntroductionMaterials and Methods <<ResultsDiscussionReferencesCITING ARTICLES

Fifty ¹²³I MIBG scintigraphy and 50 MRI examinations were performed over 5 years 9 months in 28 pediatric patients (18 boys and 10 girls; mean age, 3.2 years; age range, 1 week–11 years) and were analyzed retrospectively. All patients had suspected or proven neuroblastoma. Of the 50 examinations, 22 combined MRI and MIBG scintigraphy studies were performed for the primary diagnosis before the start of chemotherapy. Twenty-eight follow-up examinations were performed in patients with proven neuroblastoma.

MIBG scans were obtained 24 hr after slow IV injection of ¹²³I-labeled MIBG (3.7 MBq/kg of body weight). Supersaturated potassium iodide was administered 1 day before the examination and was continued for 3 days to block thyroid uptake. Anterior and posterior images of the whole body were recorded with a double-headed gamma camera (Prism 2000 XP [medium-energy collimator, 256 × 256 matrix], Philips Medical Systems, Best, The Netherlands). In addition, single-photon emission computed tomography images were acquired 24 hr after injection.

MRI was performed with a 1.5-T scanner (Magnetom Vision, Siemens, Erlangen, Germany). Head or body array coils were used according to the size of the patient. Axial, coronal, or sagittal spin-echo or fast spin-echo T1- and T2-weighted images and short tau inversion recovery (STIR) images were obtained. The slice thickness ranged from 3 to 6 mm. In addition, gadopentetate dimeglumine (Magnevist, Schering, Berlin, Germany) was IV injected (0.2 mL/kg of body weight) to assess contrast enhancement of suspected lesions on T1-weighted sequences. Uncooperative patients were sedated or anesthetized by an anesthesiologist with continuous cardiac and respiratory monitoring.

The inclusion criterion for our study was a maximum time frame of 30 days between MIBG scintigraphy and MRI. For patients being treated with chemotherapy, the maximum time interval between the two examinations was 10 days. The mean time interval between the two modalities was 6.4 days (range, 0–29 days). Suspect tumor lesions were included only if they were in the field of view on images from both modalities.

Individual analyses of each of the two modalities were performed by two experienced observers. Analysis was performed with knowledge of clinical data, but without knowledge of the findings on the other imaging modality. To determine whether regression or disappearance of lesions was evident on follow-up examinations, observers needed to know the findings from the previous examination on the same modality. Combined assessments of both modalities were performed by the same observers after the separate analyses. MRIs were reviewed on a PC using eFilm Workstation 1.5.3 software (eFilm Medical, Toronto, ON, Canada), and MIBG scans were reviewed on a Hermes workstation (Nuclear Diagnostics, Haegersten, Sweden).

Image analyses were performed on a lesion-related basis (in contrast to a case-related evaluation). A total of 115 lesions were evaluated. Because 46 of the 115 lesions were analyzed on at least one (up to five) follow-up examination, the total number of lesion-related evaluations was 193.

On both modalities, each lesion was judged either positive or negative with regard to neuroblastoma involvement. For observers to reach a decision about lesions with discrepant results on both modalities, a diagnostic confidence score of three levels was established for each modality: 1, both observers were uncertain about a positive or negative finding; 2, one observer was uncertain and one observer was certain; and 3, both observers were certain. This diagnostic confidence score was assigned to each suspect lesion on MIBG scintigraphy and MRI separately. For lesions with discrepant findings on both modalities, the finding of the modality with the higher diagnostic confidence score was accepted. If results from both modalities were discrepant and had the same diagnostic confidence score value, the lesion was judged positive.

On MIBG scintigraphy, a lesion was judged positive for neuroblastoma when nonphysiologic focal uptake was seen. On MRI, a lesion was presumably vital neuroblastoma tissue when high signal intensity on STIR sequences and the presence of gadolinium enhancement on T1-weighted spin-echo sequences were detected.

The gold standard was established by histopathologic findings (n = 45) or follow-up control examinations (n = 148). Especially for patients with stage IV neuroblastoma, histologic verification of all metastases is impossible. Therefore, on follow-up control examinations, a minimum of 6 months was used for verification of lesions. In these cases, a lesion was classified as a false-positive finding if it disappeared without tumor therapy during the observation period. A lesion was classified as a true-positive finding if it persisted or progressed during follow-up or if it showed clear regression under specific therapy.

Discussion	Choose Top of pageABSTRACTIntroductionMaterials and MethodsResultsDiscussion <<ReferencesCITING ARTICLES

In studies assessing neuroblastoma lesions in pediatric patients, the gold standard can hardly be represented exclusively by histology [17]. Especially in patients with stage IV disease, biopsy of all suspect lesions cannot be performed. Therefore, follow-up examinations must be used to verify the diagnosis of suspect lesions. Nevertheless, this limitation represents a weakness of the study design because this gold standard can be responsible for classification errors.

A lesion-based assessment such as our study has only limited value for clinical grading according to the International Neuroblastoma Staging System [18]. However, for an evaluation of the potential of different imaging modalities, clinical grading is inadequate when assessing the accuracy of detection of a single lesion that would be important to detect for accurate staging.

On MIBG scintigraphy, the sensitivity (69%) and specificity (85%) for the detection of neuroblastoma lesions in our study were markedly lower compared with published data. In several earlier studies, MIBG scintigraphy was found to be 100% specific and approximately 90% sensitive for the detection of neuroblastoma [6, 11, 13]. One possible reason for this disagreement is that case-related assessments were performed in those studies in contrast to the lesion-related analyses in our study. In cases of coexistent true-positive and false-negative lesions, the whole examination is graded true-positive in a case-related study, thereby not taking into account all false-negative lesions. Another possible reason is that the sensitivity of MIBG scintigraphy is limited in the detection of single bone and bone marrow metastases, as has been reported in the literature [13, 19]. Gordon et al. [19] reported negative findings on MIBG scintigraphy in 66 (29%) of 227 skeletal lesions, whereas bone scintigraphy findings were positive. Troncone et al. [13] showed the sensitivity of MIBG scintigraphy (61%) in depicting individual bone marrow lesions is limited. Bone and bone marrow metastases were the primary source (32/39 cases) of false-negative MIBG scintigraphy findings in our study (Fig. 1A, 1B, 1C). Possible reasons for the limited detection rate of bone metastases are the coexistence of hot and cold MIBG lesions in the same patient [12] and metabolic changes of the tumor and its foci under treatment [20].

Additional false-negative MIBG scintigraphy findings resulted in skin, lymph node, and liver metastases and in one primary adrenal neuroblastoma (Fig. 2A, 2B, 2C). These findings are concordant with published studies showing the difficulties in differentiating liver metastases from physiologic uptake in the liver [20, 21] and a general false-negative rate of 8–10% on MIBG scintigraphy [12, 22].

On the other hand, false-positive findings were seen in liver lesions such as hemangioma, focal nodular hyperplasia, and a normal liver due to an inhomogeneous uptake pattern in the liver. However, the diagnostic confidence score was low in these cases (Table 2). Misinterpretation of a normal renal pelvis as a neuroblastoma lesion has been reported [23] and occurred in two of our patients. Bonnin et al. [20] described the spectrum of false-positive MIBG findings due to a physiologic uptake in a normal adrenal gland, in the bowel, and in the musculature, as confirmed in three of our patients.

The increase in the sensitivity and specificity on follow-up examinations in our study was caused by the observers' knowledge of the findings from the previous examination, which was necessary so the observers could determine whether regression or disappearance of lesions was evident at follow-up. The observers' knowledge of the findings on previous examinations represents a bias for the results of follow-up examinations.

MRI has advantages in evaluating the extent of a primary tumor and a possible vessel encasement and in determining operability [3, 4, 24–26]. In our study, the sensitivity of MRI (86%) was markedly higher than its specificity (77%). These results are consistent with studies reporting MRI to be highly sensitive, but hardly specific [27, 28]. However, on follow-up examinations, we found a decrease in sensitivity (68%) compared with the primary diagnosis (93%). Our hypothesis for this decrease is that MRI shows a rapid signal intensity change for lesions under chemotherapy, as seen in bony metastases and residual vital tumors (Fig. 5A, 5B, 5C, 5D), that leads to false-negative findings. However, additional follow-up examinations (MIBG scintigraphy, biopsy, or both) showed a tumor persistence or progression in these lesions. Nine of these 12 false-negative MRI findings on follow-up occurred in one patient, thereby leading to this high decrease of sensitivity. Two other neuroblastoma lesions in our study were misinterpreted as a rhabdomyosarcoma and a Wilms' tumor (Fig. 4A, 4B, 4C, 4D), because the origin of the mass could not be determined on MRI. In that point, our findings agree with the study of Ng and Kingston [27] but disagree with the findings of Dietrich et al. [28] who describe a clear and easy differentiation between Wilms' tumor and neuroblastoma on MRI.

False-positive MRI results were seen in inactive posttherapeutic neuroblastoma metastases, reactive changes after surgery (Fig. 3A, 3B, 3C) or after irradiation, and enlarged lymph nodes without neuroblastoma involvement. A possible reason for these false-positive results is the inability to distinguish metabolically active from inactive tumor tissue by means of MRI. Two ganglioneuromas and one pheochromocytoma were incorrectly classified as neuroblastoma because the internal structure and growth pattern of these tumors are identical on MRI [29].

A comparison of MIBG scintigraphy and MRI in the diagnosis of neuroblastoma has been reported in some studies in which researchers primarily assessed bone marrow metastases. A high rate of false-negative bone marrow metastases, which can be detected on MRI, on MIBG scintigraphy and poor specificity of MRI was reported by these authors [14, 30, 31]. These findings are consistent with our study findings—a relatively low sensitivity on MIBG scintigraphy and low specificity on MRI in detecting not only bone marrow metastases but also all neuroblastoma lesions. Using integrated imaging for neuroblastoma diagnosis, we found a marked increase in both sensitivity (99%) and specificity (95%).

Of special interest are the diagnoses for the 76 cases with discrepant findings on both modalities (Table 2). False-negative MIBG scintigraphy findings (mean diagnostic confidence score, 2.5) occurred primarily in bone and bone marrow metastases. In these cases, MRI findings were clearly positive, with a mean diagnostic confidence score of 2.9. A lower mean diagnostic confidence score was seen in all other false-positive and false-negative MRI and MIBG findings compared with true-positive and true-negative results. The lowest diagnostic confidence scores were for false-positive findings, which is not surprising because of a tendency to misinterpret doubtful findings as positive.

Rapid signal intensity changes in lesions under chemotherapy on MRI were responsible for false-negative results, and thus for the drop of sensitivity at follow-up compared with the primary diagnosis. In contrast, the sensitivity of MIBG scintigraphy showed an increase from the primary diagnosis to follow-up (Table 1). Posttherapeutic changes were difficult to distinguish from residual tumors and led to false-positive results on MRI that could be clarified with MIBG scintigraphy (Fig. 3A, 3B, 3C). In cases in which MRI could not definitively detect the origin of the tumor, additional MIBG scintigraphy was decisive in determining neuroblastoma (Fig. 4A, 4B, 4C, 4D).

With combined imaging, one false-negative (proven bone marrow metastasis) and three false-positive diagnoses (two ganglioneuromas and one pheochromocytoma) remained (Fig. 6A, 6B). In all three cases with false-positive diagnoses on integrated imaging, the therapeutic strategy was not affected. A tumor was detected in these cases, and operative biopsy was the next diagnostic step, which is not different from neuroblastoma.

Because of its high soft-tissue contrast [5], MRI was decisive in finding the correct differential diagnosis in our cases of thymic hyperplasia, lymphangioma, and adrenal hemorrhage. All these findings were true-negative on both modalities. In two discrepant MRI (true-negative) and MIBG (false-positive) findings, MRI (higher diagnostic confidence score) led to the correct diagnosis of focal nodular hyperplasia and a hemangioma of the liver on combined imaging.