The current retrospective study was performed with data deriving from a private radiologic centre, performing – apart from numerous other imaging modalities – MRI scans of all body sites. The local medical ethics committee has approved the study (33–630 ex 20/21) that involved patients having received an MRI of the knee for any indication between 01.01.2007 and 01.03.2020. During that time, 44,762 patients had undergone at least one knee MRI scan. Of these, 5182 had undergone MRI scans of both knee joints, resulting in a total of 24,125 and 25,819 MRI scans of the left and right knee, respectively.

All MRI reports of the knee were searched for at least one of the following terms by one of the co-authors (J.W.) who had performed an extensive study of literature and had received intensive training on imaging features of cartilage lesions before data acquisition: “Enchondrom” (enchondroma), “kartilaginäre Läsion” (cartilaginous lesion), “cartilaginäre Läsion” (cartilaginous lesion), “kartilaginärer Tumor” (cartilaginous tumour), “cartilaginärer Tumor” (cartilaginous tumour), “chondrogene Läsion” (chondrogenic lesion), “chondrogener Tumor” (chondrogenic tumour), “Chondrosarkom” (chondrosarcoma), “ACT - atypischer chondromatöser Tumor“ (atypical chondromatous tumour), “atypische chondromatöse Läsion“ (atypical chondromatous lesion).

Altogether, 697 patients’ MRI reports were positive for at least one search term and were subsequently analysed in further detail (Fig. 1). Of these, 21 patients had to be excluded due to the following reasons: In 11 patients, reason for referral had been a suspected cartilaginous tumour but MRI could not confirm this tentative diagnosis; 3 patients had MRI reports containing not only findings of MRIs of the knee but also of other body regions, in whom they had been diagnosed with a cartilage lesion; 2 patients were initially suspected to have an enchondroma, yet follow-up MRIs led to a change of primary diagnosis; 5 patients had undergone surgical removal of the cartilaginous lesion prior to index imaging.

Flow chart representing the filtering of patients with a cartilaginous tumour

Reports of the remaining 676 patients were re-examined together with related MRI scans, and a definitive radiologic diagnosis of cartilaginous tumours was confirmed in 635 patients. However, 41 patients had inconclusive reports and images. Therefore, the advice of a senior radiologist was sought for these cases. Thereafter, 20 more patients were excluded, as they did not show typical features of EC/ACT, resulting in 656 patients ultimately eligible.

MRI-based differentiation into EC (Fig. 2) and ACT (Fig. 3) was made based on tumour characteristics suspicious of aggressive behavior as proposed by Mulligan et al. [5], Murphey et al. [4], van de Sande et al. [11] and Douis et al. [12], i.e. tumour size > 4.9 cm, periosteal reaction, perilesional edema or deep endosteal scalloping involving ≥ 2/3 of cortical thickness. With either one of these features being positive, chondrogenic lesions were classified as ACT. All tumours exhibiting at least one of these features were thoroughly examined by an experienced orthopedic oncologist, who excluded 5 more patients due to diagnosis of bone infarction (n = 3), postoperative changes (n = 1) and intraosseous ganglion (n = 1).

EC of the right tibia: (A) proton density, fat suppression, coronal, turbo spin echo; (B) proton density, fat suppression, transversal, turbo spin echo; (C) proton density, fat suppression, sagittal, turbo spin echo; (D) t1, coronal, turbo spin echo (year of MRI: 2016)

ACT of the left femur: (A) proton density, fat suppression, coronal, turbo spin echo; (B) proton density, fat suppression, transversal, turbo spin echo; (C) proton density, fat suppression, sagittal, turbo spin echo; (D) t1, coronal, turbo spin echo (year of MRI: 2016)

Overall, 651 patients had a cartilaginous tumour, with 21 patients presenting with 2 cartilaginous lesions at the same time, resulting in 672 cases of EC/ACT.

MRI was considered positive for a cartilaginous tumour on identification of a smooth or lobulated lesion that presented itself as a focal geographic area within the bone marrow, showing low signal intensity on proton-density-weighted and on T1-weighted images and high signal intensity on proton-density fat-suppressed images. Lesions positive for these characteristics but strictly subchondral in location had to be excluded, as they most likely represent different entities such as subchondral cysts, intraosseous ganglia, subchondral edema or contusion.

Apart from patient gender and age, the following tumour-specific features were ascertained: lesion size (maximal tumour diameter in cm), lesion site (femur, tibia, fibula, patella), tumour location (eccentric or central; epiphyseal, epimetaphyseal, metaphyseal, metadiaphyseal, diaphyseal), endosteal scalloping, perilesional edema and periosteal reaction. Furthermore, indication for MRI (tumour associated symptoms or follow-up examinations; no tumour related indication; no documented indication) and whether the patients had undergone dynamic contrast MRI was ascertained.

MRI examinations were performed on two different 3T MRI systems (Siemens Magnetom Skyra/Siemens Magnetom Vida; both Siemens Healthcare Diagnostics GmbH, Austria) with a 15/18-channel knee coil. Sequences acquired were (1) coronal proton density (PD) with fat suppression (FS) (field of view (VF) 160/140 mm; Matrix (M) 307 × 384/307 × 384; repetition time (TR) 3000/3200 ms; echo time (TE) 34/25 ms; slice thickness (ST) 3/3 mm; interslice gap (IG) 0.6/0.6 mm); (2) transversal PD with FS (VF 160/150 mm; M 307 × 384/307 × 384; TR 5460/4110 ms; TE 37/35 ms; ST 2.5-3/3 mm; IG 0.6/0.6 mm); (3) sagittal PD with FS (VF 160/140 mm; M 307 × 384/307 × 384; TR 2920/2920 ms; TE 34/34 ms; ST 3/3 mm; IG 0.6/0.6 mm); (4) coronal T1 weighted turbo spin echo (TSE) (VF 160/140 mm; M 346 × 384/290 × 484; TR 690/690 ms; TE 11/19 ms; ST 3/3 mm; IG 0.6/0.6 mm; flip angle (FA) 180°/150°).

Gd-DTPA (dose 2 ml/kg body weight) via venous access was administered in 77 of 651 patients (81 of 672 tumours), followed by an MRI examination with the sequence “(4) coronal T1 weighted turbo spin echo” that was subtracted from the native T1 sequence. Additionally, T1 sequences with fat suppression in different dimensions were recorded.

Statistical analyses were carried out with Stata Version 16.1 for Mac (StataCorp, College Station, Texas, US). Patient demographics were summarized for based on the total number of patients with cartilaginous lesions. Tumour demographics were summarized from the total number of lesions found in patients. Prevalence was calculated based on the diagnosis of a benign cartilaginous lesion per patient, and not based on the number of cartilage lesions found in total. In line with this, patients without ECs/ACTs undergoing MRI scans of both knee joints during the defined period were counted once only for calculation of prevalence. Normally and non-normally distributed variables were given as means and corresponding standard deviations, as well as medians and corresponding interquartile ranges (IQR), respectively. Fisher’s exact test and t-test were used to assess differences in binary (or ordinary) and continuous variables, respectively, depending on radiological diagnosis. A p-value of < 0.05 was considered statistically significant.