[home] [Personal Program] [Help]

---

Session: Poster session II
Session starts: Thursday 24 January, 16:00

---

Ali Ataei (Orthopaedic research lab, Radboud university medical center)
Florieke Eggermont (Orthopaedic research lab, Radboud university medical center)
Milan Baars ()
Yvette van der Linden (Leiden University Medical Center)
Nico Verdonschot (Orthopaedic research lab, Radboud university medical center)
Esther Tanck (Orthopaedic research lab, Radboud university medical center)

Abstract:
Bone is the third most commonly affected tissue by metastases. Metastases in the femur are often painful and increase the risk of pathological fracture. Accurate delineation of bone metastases is amongst others important for radiotherapy planning and determining exact radiation fields. Hence, the aim of this study was to investigate the inter- and intra-operator reliability in manual segmentation of femoral metastatic lesions. CT-scans of 54 patients with femoral metastases (19 osteolytic, 17 osteoblastic and 18 mixed) were included in this study. Two inexperienced operators were trained and then asked to segment metastatic lesions in all patients twice with a four-week time interval. Dice coefficients (DC) were calculated to quantify the inter- and intra-operator overlap of the segmentations. A DC˃0.8 indicates good reliability and a DC between 0.6 and 0.8 is acceptable. Mean DCs for inter-operator reliability were 0.54 (±0.28) and 0.50 (±0.32) for the first and second time segmentation, respectively. Mean DCs for intra-operator reliability were 0.56 (±0.28) for operator I and 0.71 (±0.23) for operator II. For large lesions (over 20000 voxels), the mean DC was 0.82 (±0.07), whereas for smaller lesions (less than 20000 voxels), the mean DC was 0.60 (±0.28). Osteoblastic lesions had the highest mean inter- and intra-operator DC with 0.57 in the first time, 0.60 in the second time, 0.60 for operator I and 0.78 for operator II compared to osteolytic (0.54, 0.45, 0.56, 0.63) and mixed lesions (0.51, 0.46, 0.52, 0.74). The mean DC of neither inter- nor intra-operator reliability test were satisfying. Only one of the operators showed an acceptable intra-operator reliability. Additionally, segmentation of larger lesions resulted in good inter- and intra-operator reliability. Higher reliability for osteoblastic lesions can be justified by their more clear edges on CT-scan images. This study reveals that the current manual segmentation approach might results in unacceptable reliability rates. Hence, development and use of an automated segmentation method might be beneficial.