Ultrasound semiautomatic versus manual estimation of carotid intima-media thickness: reproducibility and cardiovascular risk stratification

Caterina Beatrice Monti, Marco Alì, Davide Capra, Federico Wiedenmann, Giulia Lastella, Francesco Secchi, Francesco Sardanelli


Aims: Carotid intima-media thickness (CIMT) is used increasingly as an imaging biomarker of cardiovascular risk (CVR). Our aim was to compare semiautomatic CIMT (sCIMT) versus manual CIMT (mCIMT) for reproducibility and prediction of CVR.

Materials and methods: Two independent readers measured sCIMT and mCIMT on previously acquired images of the right common carotid artery of 200 consecutive patients. Measurements were performed twice, four weeks apart; sCIMT was reported along with an image quality index (IQI) provided by the software. CVR stratification was compared for thresholds established by mCIMT studies, adapted for sCIMT according to a regression model.

Results: sCIMT (median 0.67 mm, interquartile range [IQR] 0.57‒0.76 mm) was significantly lower (p<0.001) than mCIMT (median 0.76 mm, IQR 0.63‒0.84 mm; ρ=0.832, p<0.001, slope 0.714, intercept 0.124). Overall, intra-reader reproducibility was 76% for sCIMT and 83% for mCIMT (p=0.002), inter-reader reproducibility 75% and 76%, respectively (p=0.316). In 129 cases with IQI≥0.65, reproducibility was significantly higher (p≤0.004) for sCIMT than for mCIMT (intra-reader 85% versus 83%, inter-reader 80% versus 77%,). The agreement between sCIMT and mCIMT for CVR stratification was fair both overall (κ=0.270) and for IQI≥0.65 (κ=0.345), crude concordance being 79% and 88%, respectively.

Conclusions: Reproducibility of sCIMT was not higher than mCIMT overall but sCIMT was significantly more reproducible than mCIMT for high-IQI cases. sCIMT cannot be used for CVR stratification due to fair concordance with mCIMT, even for high IQI. More research is required to improve image quality and define sCIMT-based thresholds for stratification of CVR.


carotid intima-media thickness; Doppler; cardiovascular diseases; reproducibility of results; computer-assisted

Full Text:


DOI: http://dx.doi.org/10.11152/mu-2416