151751-Najiba-Chargi

40 CHAPTER 2 We do believe that at the current time, the SMA at C3 can provide a good estimation of SMM of HNC patients without the need for additional diagnostics and at minimal effort, with con - siderable accuracy. To facilitate implementation of SMM measurement in clinical practice, we believe the long-term focus should shift towards using artificial intelligence such as deep learning andmachine learning to develop an automatic, whole muscle volume analysis based on routinely available CT imaging or MRI. Research into thesemethods are ongoing, and the ex - pectation is that whole- or portion-of-bodymeasurement of SMMwill provide amuchmore ac- curate representation of a patients overall body composition and skeletal muscle status than the SMA on a single CT slide or a single muscle, with no or very little manual work involved. 22–24 Indeed, the use of the SMA at the level of L3 as an estimation of whole body skeletal muscle volume is based on studies using whole-body MRI for manual segmentation and calculation of whole-body skeletal muscle volume; in these studies, whole body MRI is referenced as the gold standard. 6,25 Manual segmentation of whole body MRI is time-consuming and therefore clinically not feasible. However, when software is available to perform automatic skeletal muscle volume analysis, a whole-body analysis approach seems preferred. In the short term, future studies may be aimed at developing gender-specific references values for SMA at the level of C3, to allow for the use of SMA at the level of C3 as a direct measure of SMM and to overcome the problemof several different cut-offs for low SMM that are currently available. 26,27 There are limitations to our study that need to be addressed. Most patients in our study pre - sentedwith advanced stage disease; in our center, the indication for FDG-PET/CT is a suspected advanced stage disease. Inherently to the use of FDG-PET/CT, patients with limited disease are underrepresented in this study. We excluded patients who had received prior treatment for HNC for this validation study, because the effect of prior local treatment (e.g., radiother - apy or surgery) on the accuracy of delineation of SMA at C3 is not known and may cloud its relationship with SMA at L3. It is well-known that patients with tobacco-related cancers of the upper aero-digestive tract have a substantial risk of developing a second primary malig- nancy in the same region. In another study by our group, also imaging of patients who had undergone prior treatment was also used, and found that low SMM as identified at the level of C3 was associated with adverse outcomes in patients with and without prior treatment. 9 Some patients with HNC will undergo MRI instead of CT imaging. In this study we only used CT imaging, according to the protocol described by Swartz et al. 7 Two recent studies also showed excellent correspondence between SMA on CT imaging and MRI, and concluded that CT and MRI can be used interchangeably. 28,29 The effect of different posture and different angles (e.g. in laryngeal cancer, CT scans are often angulated to better visualize the vocal cords) was not evaluated in this study, but may influence SMA. 30 Future research should clarify this, but we expect that this problem will be overcome by using whole-body or portion-of-body skeletal muscle volumes using artificial intelligence. Our current study confirms the previously found strong correlation between SMA at the level of C3 and SMA at the level of L3. This method allows for research into the predictive and prognostic effect of low SMM in HNC patients, using routinely performed imaging of the head