Enrico Martin

127 Diagnostic accuracy of non-invasive tests Discussion MRI characteristics could varyingly detect MPNST, but absence of a target sign was highly sensitive. Ill-defined margins and perilesional edema could adequately distinguish MPNSTs from BPNST. FDG-PET has the highest diagnostic accuracy for detecting MPNST in NF1 patients, with equal efficacy when using SUVmax or T/L ratio. Ideal threshold value of SUVmax seems to be ≥3.5. Functional MRI and liquid biopsies may be useful tools as well, but do require more research. MRI in nerve sheath tumors Both MPNSTs and BPNSTs can exhibit rather different characteristics on MRI, highlighted by findings in this study. Presence of a target sign was the only MRI characteristic that could rule out MPNST, because of its nLR of less than 0.1. 28 Based on this characteristic, biopsies could be obviated for tumors with target signs. However, two studies reported 6/94 MPNSTs in this meta-analyses with a target sign. 16,32 One may argue that in order to omit a biopsy, in addition to the presence of a target sign, perilesional edema and ill-defined margins should be absent as well. Moreover, many BPNSTs do not show a target sign; 59.9% (range: 43.3-94.3%) in this meta-analysis. Nevertheless, in the remaining 40.1%, a biopsy may possibly be omitted. Presence of perilesional edema and ill-defined margins can adequately detect MPNST as the pLRs are more than 10, but biopsies may still be needed because these features can be present in a minority of BPNST as well. Unfortunately, perilesional edema and ill-defined margins are only present in 29-92% and 25-68% of MPNSTs respectively. Other characteristics that only have a moderate ability to differentiate MPNST and BPNST should therefore also be considered, including cystic changes, heterogeneity on T1, intratumoral lobulation, and large tumor size. An ideal combination of moderately specific characteristics adjacent to ill-defined margins and perilesional edema is still lacking, but may further reduce the need for biopsies. This is partially reflected by the diagnostic accuracy of qualitative assessment of MRI’s which could not outperform either sensitivity or specificity of single characteristics. 32,33,41 Likewise, studies that reported diagnostic algorithms combining features decreased in sensitivity, albeit a rise in specificity. 22,40 Hence, conventional MRI’s are imperfect and further diagnostics including FDG-PET in NF1 and biopsies may still be necessary in many cases. Luckily, interobserver agreement of MRI characteristics are very good to excellent, making them reproducible for use. 29,31 Functional MRI sequences may provide additional value in MPNST as DWI and ADC mapping yielded higher accuracy of detecting malignancy than conventional MRI characteristics. 21,38 MPNSTs show increased cellularity which makes ADCmin values relevant. Its use has however only been tested in two distinct populations and warrants further investigation. FDG-PET in NF1 patients FDG-PET scans are increasingly being applied to detect malignancy in NF1 patients with varying frequency of use across centers. Many efforts have been made to find ideal 6