Introduction

Chondrosarcomas are rare malignant neoplasms of cartilage tissues that can develop in the skeletal system. Skull base chondrosarcomas (SBCs) are a rare subtype of chondrosarcoma, accounting for less than 10% of all chondrosarcomas and 0.15% of all intracranial tumors [1]. Accurate radiologic diagnosis is essential for elaborating tailored surgical strategies and subsequent treatment plans, balancing invasiveness and functional preservation. However, owing to the rarity of SBC and the presence of other tumors that can mimic SBC, physicians often encounter a diagnostic conundrum.

Several articles have reported detailed radiologic features of SBC; SBCs typically appear isointense or hypointense on T1-weighted images (T1WI) and hyperintense on T2-weighted images (T2WI) with avid contrast enhancement [2,3,4,5]. However, accurate preoperative diagnosis of SBC may be challenging for the following reasons. First, SBCs are mostly located in ventrolateral skull base, in which several other skull base tumors can arise. Second, surgeons sometimes encounter atypical-looking SBCs, such as centrally located SBCs and SBCs with minimal enhancement. Third, there remains a paucity of imaging features for other supportive imaging modalities, such as positron-emission tomography. In particular, chordomas are often indistinguishable from SBCs since they have several common radiologic features [4, 6, 7]. In the recent literature, radiomics analysis is reported to provide better diagnostic accuracy [8], but it remains too complicated to be widely used in clinical practice. Alternatively, diffusion-weighted imaging may be a promising tool for the differential diagnosis [9, 10].

Recent studies highlight significant biological differences between SBCs and chordomas; in general, better prognosis is expected and thus a less aggressive, functional preservation-oriented surgical strategy is recommended in SBCs, while more intensive, marginal resection is advocated for chordomas [11,12,13,14,15]. To create a tailored surgical strategy according to tumor histology and to explain the associated risk and prognosis to patients, accurate presurgical diagnosis is invaluable. The goal of the present study was to describe the detailed radiologic features of SBCs and to make comparisons with chordomas, creating a useful method in distinguishing SBCs from chordomas.

Methods

Patient selection

The study was approved by the institutional review board (IRB#2231). Written informed consent was obtained from all the study participants. This study adhered to the STROBE guideline [16]. Clinical and radiologic data of patients with histologically diagnosed SBC between 2002 and 2021 were retrospectively collected from hospital medical records and image databases. For comparison purposes, consecutive patients who had a new histological diagnosis of skull base chordoma before 2021 were retrospectively reviewed and collected until the same number of patients with SBC were obtained. To focus on pre-surgical diagnosis, only those in whom digital radiologic data before initial intervention were available for review were included since surgeries/radiotherapies and subsequent recurrences could alter imaging features, including location, signal intensities (SIs), and enhancement patterns.

Data acquisition

The patients’ baseline characteristics were collected from the medical records. MRIs were performed using 1.5- or 3.0-T machines (1.5-T field strength, SIGNA Excite, GE Healthcare, IL, US; MAGNETOM Vision and MAGNETOM Avanto, Siemens Healthineers, Erlangen, Germany; 3.0-T field strength, SIGNA HDxt and SIGNA Premier, GE Healthcare; Skyra, Siemens Healthineers; Ingenia CX, Philips, Amsterdam, Netherlands) with the following acquisition parameters: T1WI, repetition time (TR)/echo time (TE) 400–520/2.26–15 ms at 5-mm thickness for Siemens machines, 659/8 ms at 5-mm thickness for Philips machine, and TR/TE 400–640/8.4–11 ms at 3–5-mm thickness for GE machines; T2WI, TR/TE 3200–4260/84–105 ms at 3–5-mm thickness for Siemens machines, TR/TE 4160–4163/90 ms at 3-mm thickness for Philips machine, and TR/TE 4000–5420/78.8–98.2 ms at 3–5-mm thickness for GE machines; DWI, TR/TE 4400–4750/81–86 ms at 5-mm thickness, b = 1000 s/mm2 for Siemens machines, TR/TE 4841/66 ms at 5-mm thickness, b = 1000 s/mm2 for Philips machine, and TR/TE 4500–7800/59.3–74 ms at 5-mm thickness, b = 1000 s/mm2 for GE machines; contrasted T1WI, TR/TE 18–24/2.3–3.4 ms at 0.4–0.8-mm thickness for Siemens machines, TR/TE 20/3.8–3.9 ms at 0.8-mm thickness for Philips machine, and TR/TE 20–24/2.2–3.1 ms at 0.8–2-mm thickness for GE machines. Apparent diffusion coefficient (ADC) maps were generated by built-in software provided by the MRI scanner vendors. SIs on T1WI and T2WI were measured by two co-authors who were blinded to the patient details using an ellipsoid-shaped region of interest, which was placed around the center of the tumor and adjusted as large as possible but not exceeding the tumor matrix, excluding regions of intratumoral calcification and hemorrhage as well as regions of entrapment of mucus (mucocele). Since SIs on T1- and T2-weighted images are relative values influenced by acquisition conditions and other factors, the SI ratio ([SI of the tumor]/[SI of the white matter]) was calculated for standardization. ADC values, which are absolute values in contrast to T1WI- and T2WI-SIs, were also measured using a similar method. Using gadolinium-enhanced T1WI, the degree and patterns of enhancement were evaluated separately; the degree was graded as “avid” if the enhancement was similar to the nasal mucosa, “fair” if it was less than the nasal mucosa but similar to muscle tissues, or “little” if it was less than the muscle tissues (Fig. 1A). The patterns of enhanced internal structures were classified as “homogenous” or “heterogenous”; in particular, “arabesque” was recorded if a characteristic “septal or ring-and-arc enhancement” [17, 18] was seen, forming an arabesque-like pattern (Fig. 1B). The tumor volume was estimated using the simplified formula for volume of ellipsoid: abc/2 [19]. Origins of SBCs were estimated based on the “centerpoint” of the tumor; a synchondrosis that was the closest to the tumor’s center point was considered as the origin (Fig. 1C) [4]. When 18F-fluorodeoxyglucose PET data were available, lesion to normal tissue ratios (L/N ratios), which were calculated using the contralateral cerebral cortex or cerebellum and generally preferred to standardized uptake values for brain lesions [20], were recorded. The presence of intratumoral calcification and associated bone erosion, which are often indistinguishable, was evaluated using CT images.

Fig. 1
figure 1

Examples of degrees (A) and patterns (B) of contrast enhancement and skull base synchondroses (C). Yellow asterisk, tumor; yellow arrow, nasal mucosa; yellow arrowhead, muscle tissues. Pet-Occ petro-occipital synchondrosis, Sph-Sph pterygo-sphenoidal synchondrosis, Sph-Eth spheno-ethmoidal synchondrosis, Sph-Occ spheno-occipital synchondrosis, Sph-Pet spheno-petrosal synchondrosis

Full size image

Statistical analysis

Differences of the baseline characteristics and radiologic features between SBC and chordoma were tested using Fisher’s exact test (categorical variables) and Mann–Whitney U test (continuous variables). Logistic regression analysis was used to examine the clinical and radiologic factors associated with the diagnosis of SBC. Multivariate analysis was performed using factors that were significant in the bivariate analysis. Finally, using the factors that had a significant association with the diagnosis of SBC, a prediction method for SBC diagnosis was developed. Receiver operating characteristic (ROC) analysis was performed to determine the cutoff value that maximized the diagnostic potential based on the Youden index [21]. A P value of < 0.05 was considered significant in all statistical analyses. All statistical analyses were performed using JMP Pro 16.0 (SAS Institute, NC, USA).

Results

Overview of the cohorts

Among 55 patients with a histological diagnosis of SBC, 41 were included in the study after excluding 16 patients in whom imaging data before initial intervention were not available. There were five (12.2%) grade I, 31 (75.6%) grade II, and two (4.9%) grade III conventional SBCs, one (2.4%) myxoid subtype, one (2.4%) mesenchymal subtype, and one (2.4%) unspecified subtype. The representative cases are illustrated in Figs. 2 and 3. For comparison, 41 consecutive patients with chordomas (30 conventional, 10 chondroid, and one dedifferentiated) were included in the study. The clinical and radiologic features of the patients are summarized in Table 1; compared to chordoma, no factors were significantly associated with SBC, although there were the following tendencies: slightly more female predominance (63.4% vs. 51.2%, P = 0.27), slightly younger age at presentation (41.0 years vs. 48.0 years, P = 0.24), more frequent occurrence of headache/neck pain (14.6% vs. 4.9%, P = 0.26).

Fig. 2
figure 2

A 49-year-old woman with right petro-occpital chondrosarcoma (grade II conventional). The tumor appears as a T1-hypointensity (A), T2-hyperintensity (B) mass with avid homogenous enhancement (C). The apparent diffusion coefficient value is high at 1865 × 10–6 mm2/s (D). The tumor has no calcifications (E). Positron emission tomography demonstrates that the tumor is mildly colder than the normal brain tissues (F), with the lesion-to-normal tissue ratio being 0.65. Right carotid angiogram is performed to examine the patency of the right carotid artery, which demonstrates its complete occlusion, with its distal part (blue arrow) patent through a retrograde collateral from the right ophthalmic artery (green arrow). Notably the cavernous portion of the internal carotid artery (purple arrow) is also patent, and there is faint tumor stain (red arrow) via the inferolateral trunk (G, lateral view). Red asterisk and arrowhead indicate the tumor and a remnant of the occluded right internal carotid artery, respectively

Full size image
Fig. 3
figure 3

A 56-year-old man with spheno-ethmoidal chondrosarcoma (grade I conventional). The tumor appears as a T1-hypointensity (A), T2-mixed intensity (B) mass with “arabesque”-like avid enhancement (C). The apparent diffusion coefficient value is high at 2326 × 10–6 mm2/s (D). There are tumoral calcifications on CT (E). Positron emission tomography demonstrates that the tumor is significantly colder than the normal brain tissues (F), with the lesion-to-normal tissue ratio being 0.2. Left external carotid angiogram demonstrates visible tumor stain (red arrow) from the left sphenopalatine artery (pink arrow) branching from the left internal maxillary artery (yellow arrow) (G, A-P view). Red asterisk indicates the tumor

Full size image
Table 1 Summary of the clinical and radiologic characteristics of the patients
Full size table

Radiologic features

Most SBCs exhibited hypointensity (27, 65.9%) or isointensity (12, 29.3%) on T1WI, and hyperintensity (35, 85.4%) or hyperintensity mixed with smaller hypointense areas (6, 14.6%) on T2WI. Regarding the patterns of enhancement on MRI, the arabesque pattern (18, 43.9%) was the most common, followed by the heterogeneous (14, 34.1%) and homogenous (9, 22.0%) patterns. Avid enhancement was the most common (23, 56.1%), followed by fair (14, 34.2%) and little (4, 9.8%) enhancement.

Data on T1SI- and T2SI-ratios was available in all patients, whereas ADC value was available in 32 SBC and 39 chordoma patients. Compared to chordomas, SBCs were significantly associated with a higher T2SI-ratio (P < 0.01) and ADC value (P < 0.01) (Table 1). While only a few chordomas exhibited an arabesque enhancement pattern (2, 4.9%), the majority of them had either heterogeneous (21, 51.2%) or homogenous (15, 36.6%) enhancement. Massive necrosis was seen in two patients (2.4%), while small areas of necrosis in background homogenous enhancement were seen in 2 (4.9%). Regarding the intensity of enhancement, little (16, 39.0%) enhancement was relatively more common in chordoma, while avid (16, 39.0%) or fair (9, 22.0%) enhancement was also possible.

Data on PET was available in 15 SBC and 21 chordoma patients. In SBCs, the PET L/N ratio was generally low-to-intermediate, with the median value being 0.36 (interquartile range, 0.25–0.50). No significant differences in the presence of tumoral calcifications or the PET L/N ratio were observed between SBCs and chordomas (Table 1).

Tumor location

A midline mass without significant laterality was a predominant feature of chordoma (37, 90.2%), while it was also observed in some SBCs (6, 14.6%; P < 0.01). The origins of the SBCs were presumed to be the petro-occipital (27, 65.9%),, spheno-petrosal (5, 12.2%), spheno-ethmoidal (3, 7.3%), pterygo-sphenoidal (2, 4.9%), and spheno-occipital (2, 4.9%) synchondroses. There were two patients in whom SBCs arose from an atypical lateral region (one intraorbital and one lateral petrous). A summary of the tumor locations according to presumed tumor origin is illustrated in Fig. 4.

Fig. 4
figure 4

Representative figures of chondrosarcoma arising from each synchondrosis are shown: A spheno-ethmoid; B spheno-occipital; C Spheno-petrosal; D pterygo-sphenoidal; E petro-occipital (rostral side, extending in the cavernous sinus); F petro-occipital (middle level, centered in the petrous apex); G petro-occipital (caudal side, extending in the jugular foramen); H intraorbital; I lateral petrous

Full size image

Factors predictive of a diagnosis of SBC

Bivariate logistic regression analysis demonstrated that the T1SI-ratio (odds ratio [OR] 0.12; 95% confidence interval [CI] 1.39 × 10–2–0.96; P = 0.04), T2SI-ratio (OR 3.93; 95% CI 1.65–9.40; P < 0.01), ADC value (OR 1.01; 95% CI 1.01–1.02; P < 0.01), little enhancement (OR 0.17, 95% CI 5.04 × 10−2–0.57, P < 0.01), and non-midline location (OR 54.0, 5% CI 14.0–2.07 × 102, P < 0.01) were significantly associated with an SBC diagnosis. Among them, only ADC value (OR 1.01; 95% CI 1.00–1.02; P < 0.01) remained significant in the multivariate analysis (Table 2).

Table 2 Bi- and Multi-variate logistic regression analysis for factors associated with a diagnosis of SBC
Full size table

ROC analysis was performed to further evaluate the association between SBC diagnosis and ADC value. The area under the curve was 0.99, and 1750 × 10–6 mm2/s was found to maximize the Youden index, which was considered the best cutoff value (Fig. 5). An ADC value of ≥ 1750 × 10–6 mm2/s in fact had a strong association with an SBC diagnosis (OR 5.89 × 102; 95% CI 51.0–6.80 × 103; P < 0.01), yielding a sensitivity of 93.9%, specificity of 97.4%, positive predictive value of 96.9%, and negative predictive value of 95.0%.

Fig. 5
figure 5

The receiver-operating curve for the association between the diagnosis of chondrosarcoma and apparent diffusion coefficient value. AUC area under the curve

Full size image

Radiologic parameters according to the histological subtypes

The radiologic parameters according to the histological subtypes of SBCs are summarized in Table 3. Between grade I and II SBCs, no significant differences were observed in tumor volume (P = 0.42), midline location (P = 0.55), T1SI-ratio (P = 0.77), T2SI-ratio (P = 0.06), ADC value (P = 0.27), and presence of tumoral calcification (P = 1.00); whereas PET L/N ratio was significantly higher in grade II than in grade I SBCs (P = 0.04).

Table 3 Summary of the radiologic characteristics according to the histological subtypes
Full size table

As for chordomas, there were no statistical differences between conventional and chondroid chordomas in the radiologic parameters except that tumor volume was significantly larger in chondroid chordomas (P = 0.04). The T2SI-ratio of dedifferentiated chordoma appeared to be lower than that of conventional and chondroid chordomas, though statistical analysis could not be performed because there was only one patient with dedifferentiated chordoma.

When comparing grade I/II SBCs to conventional chordomas, T2SI-ratio (median 2.51 vs. 2.06; P < 0.01) and ADC value (median 2053 × 10–6 mm2/s vs. 1329 × 10–6 mm2/s; P < 0.01) were significantly higher in grade I/II SBCs. When comparing grade I/II SBCs to chondroid chordomas, tumor volume (median, 10.8 mL vs. 29.8 mL; P < 0.01) was significantly smaller, T2SI-ratio (median 2.51 vs. 1.84; P < 0.01) and ADC value (median 2053 × 10–6 mm2/s vs. 1349 × 10–6 mm2/s; P < 0.01) were significantly higher, and tumoral calcification (72.2% vs. 33.3%; P = 0.05) was marginally more common in grade I/II SBCs.

Discussion

The present study depicted the in-depth radiologic features of SBC, in contrast to chordoma. In particular, we found that ADC values were very useful for distinguishing SBCs from chordomas, with the difference in the median values reaching approximately 1.5-fold (2065 × 10–6 mm2/s vs. 1329 × 10–6 mm2/s). The multivariate analysis demonstrated that the unit odds ratio for ADC value was 1.01 (1.011596); when using more practical examples, this means that the estimated odds of a tumor being SBC increased by 1.78 or 3.17 for each 50- or 100-point increase in ADC value, respectively. Based on the subtype analyses, the difference in ADC values between SBCs and chordomas was apparently significant regardless of the tumor subtypes. Notably, similar levels of difference in ADC values between SBCs and chordomas were observed in previous studies (2051 × 10–6 mm2/s vs. 1474 × 10–6 mm2/s and 2017 × 10–6 mm2/s vs. 1264 × 10–6 mm2/s, respectively), suggesting the robustness of this finding [9, 10]. Since ADC value is a measure of the magnitude of water molecule diffusion within tissues, it theoretically reflects the extent of tissue cellularity [10]. Accordingly, ADC value is known to provide a good insight into histological characteristics of the lesions, and can be a good biomarker when distinguishing benign from malignant tumors [22]. Indeed, there are a few previous studies reporting the diagnostic value of diffusion-weighted imaging, which is a secondary product calculated from ADC map [9, 10]. As previously suggested, the difference in ADC values between SBCs and chordomas may possibly result from the difference in their cellularities, structures of extracellular matrices, and/or intracellular components (nuclear-to-cytoplasmic ratio, amount of intracellular proteins, etc.) [10, 23].

The key finding of our study is that an ADC value ≥ 1750 × 10–6 mm2/s is highly predictive of SBC diagnosis. Moreover, the diagnostic potential will be further improved if other supportive imaging features of SBC, as detailed below, are considered together. The strength of this prediction method is not only its accuracy, but also its excellent feasibility. Li et al. recently reported a radiomics-based method to distinguish chordoma and SBC with great accuracy [8]. Given the large number of patients and involvement of multiparametric MRI, their method may be more robustly applicable, though this type of method significantly lacks feasibility. However, our method can be readily used in clinical practice, given that ADC values are easy to measure as long as ADC maps are available.

Some MRI findings can support the diagnosis of SBC. Our study confirmed that SBCs exhibit hypointensity to isointensity on T1WI, which is in line with the previous literature [4]. On T2WI, they typically exhibited hyperintensity, which was occasionally mixed with less dominant hypointense areas, likely due to the presence of calcifications. While T2 SI appears to be higher in SBCs than in chordomas, the difference may not be sufficient to be used alone for distinguishing SBC. The majority had fair or avid enhancement. Minimal enhancement was possible, but significantly more likely in chordomas as shown in the univariate analysis. Although arabesque-like enhancement patterns can be seen in some chordomas, they seem to be a hallmark of SBC. This feature, which was previously introduced as “septal or ring-and-arc enhancement” [17, 18], is thought to correspond to fibrovascular bundles surrounding the cartilaginous nodules and is indicative of low-grade SBCs [2]. However, it should be noted that this feature is somewhat subjective and susceptible to an inter-observer error. Although each of the above-mentioned features alone may not be sufficient to support an SBC diagnosis, they will be good markers when combined with the above-mentioned ADC value and tumor volume-based prediction methods.

Regarding the clinical characteristics, including age, sex, and mode of presentation, no significant differences between SBCs and chordomas were observed, although there was a tendency for female predominance in SBCs. Given that mild-to-moderate female predominance has been repeatedly observed in many studies, it appears to be very robust [9, 10, 24,25,26]. Nevertheless, this predominance alone is not useful to support an SBC diagnosis, since chordoma does not have an apparent sex difference.

Limitations

The present study has several limitations. The retrospective nature and relatively small sample size might have affected the results owing to possible selection bias and underpowered analyses. In addition, our measurements of SIs did not involve the entire volume of the tumor, and thus, the values might be susceptible to intratumoral variability. Especially considering cases of heterogeneous tumors, the measurement of the entire volume might better reflect the characteristics of the tumor. Next, since intratumoral calcification leads to lower ADC values, the present method may not be appropriate for tumors with massive calcification. Moreover, given ADC values may somewhat vary between different MRI machines, the external validity of the present ADC value-based method needs to be further evaluated, and some modifications may be required in the future. Lastly, since the numbers of high grade tumors were limited in our cohort, our result should be cautiously applied when high grade tumors are expected based on their clinical courses. Despite the fact that, in our cohort, the grade III chondrosarcoma exhibited as high an ADC value as the less aggressive subtypes while the dedifferentiated chordoma exhibited as low an ADC value as the less aggressive subtypes, further case accumulation is needed to reach a more robust conclusion. Despite these limitations, our study included one of the largest numbers of patients with “primary” SBC who did not lose the clinical/radiologic features at initial manifestation and thus were the most suitable for examining presurgical diagnosis methods. In addition, the entire volume measurement of SIs and ADC values would be unrealistic in daily clinical practice, and our method has certain advantages in feasibility.

Conclusion

The ADC value appears to be a promising tool to distinguish SBC from chordoma, and an ADC value of ≥ 1750 × 10–6 mm2/s may be an optimal cutoff. The prediction accuracy increases when other features of SBC, such as non-midline location and arabesque-like enhancement, are considered together.