ESTRO 2024 - Abstract Book

S3811

Physics - Image acquisition and processing

ESTRO 2024

spinal metastases involvement evaluated on hydroxyapatite-enhanced images (HAP) produced from DECT imaging is acceptable compared with that on MRI images.

Material/Methods:

Of the patients treated with radiotherapy for osteolytic spinal metastases at our institution between 2019 and 2021, we included all 18 cases, including 46 vertebrae and 276 segments for which both DECT and MRI were performed. Patients included 13 men and 5 women, having a median age of 70 years (range of 45–94 years). The primary sites included 6 lung, 3 hepatobiliary, 3 renal, 2 breast, and 1 each of skin, colon, thyroid, and prostate, with 27 thoracic and 19 lumbar vertebrae examined. Each vertebra was divided into six segments according to the international consensus guidelines [1], and each segment was evaluated for the presence of metastatic involvement using MRI, HAP, and normal CT imaging (NL). Three radiation oncologists (RO1–3) and one diagnostic radiologist (DR) individually determined the presence or absence of invasion for each modality (e.g., MRI results were evaluated without reference to those of HAP and NL). The Kappa Coefficient (κ) was calculated for the presence or absence of invasion as evaluated by MRI for each evaluator and the agreement between the presence or absence of invasion as evaluated by HAP and NL. The index of strength of agreement reported by Landis and Koch was used to evaluate the calculated κ values [2]. The sensitivity and specificity of MRI, HAP, and NL evaluation of invasion by RO1–3 were calculated assuming the presence or absence of invasion evaluated by DR, using all three types of images as the gold standard.

Results:

Regarding the percent agreement between MRI and HAP or NL, median κ values were 0.47 (range, 0.38–0.54) and 0.32 (range, 0.25–0.5), respectively. According to the Landis and Koch index, the strength of agreement was “Moderate” for HAP and “Fair” for NL. For all evaluators, agreement with the MRI ratings was higher for HAP than for NL.

Sensitivity and specificity of RO1–3 with DR evaluation as the gold standard, was median 0.63 (range, 0.62–0.71) and 0.89 (range, 0.89–0.92) in MRI, median 0.56 (range, 0.52–0.71) and 0.92 (range, 0.83–0.98) in HAP, and median 0.45 (range, 0.42–0.47) and 0.85 (range, 0.85–0.93) in NL, respectively. Sensitivity of HAP was lower than that of MRI but better than that of NL. Regarding specificity, HAP was superior to MRI and NL.