AIM: The measurement of the length of the cervix is just a method of routine, but the analysis of the grey-scale texture of the cervix represents a method poorly evaluated to date which may have the scope of improving the prediction of premature birth. Before it can be used for clinical purposes, the method must be standardised. METHODS: 66 patients were examined endosonographically. The analysis was done both visually and automatically, using a commercially available software. Inside a ROI (region of interest), we calculated the mean of the histogram, and homogeneity as well as contrast of the co-occurrence matrix of the grey-scale values. The influence of ultrasound wave propagation as well as location, extension and sound angle of the ROI were examined, and the results of the analysis performed by man and machine were compared. Additionally, we compared the textures of the inner and outer cervical os as well as the anterior and posterior lip of the cervix. RESULTS: As homogeneity and contrast were significantly influenced by the propagation of ultrasound waves, we developed a new angle-independent parameter combining the results of the 45 degrees and the 90 degrees calculation. Small changes in location, extension and sound angle of the ROI did not influence our results significantly. The texture analysis performed by the software was significantly better than the one performed by the examiner. The maximum of brightness was detected in the region of the internal os/anterior lip. Contrast as well as homogeneity showed significant differences between all parts of the cervix. CONCLUSION: The analysis of the grey-scale texture of the cervix can deliver clinically relevant results in the prediction of premature birth. The external influence on sound wave propagation as well as different locations inside the cervix must be taken into consideration.
AIM: The measurement of the length of the cervix is just a method of routine, but the analysis of the grey-scale texture of the cervix represents a method poorly evaluated to date which may have the scope of improving the prediction of premature birth. Before it can be used for clinical purposes, the method must be standardised. METHODS: 66 patients were examined endosonographically. The analysis was done both visually and automatically, using a commercially available software. Inside a ROI (region of interest), we calculated the mean of the histogram, and homogeneity as well as contrast of the co-occurrence matrix of the grey-scale values. The influence of ultrasound wave propagation as well as location, extension and sound angle of the ROI were examined, and the results of the analysis performed by man and machine were compared. Additionally, we compared the textures of the inner and outer cervical os as well as the anterior and posterior lip of the cervix. RESULTS: As homogeneity and contrast were significantly influenced by the propagation of ultrasound waves, we developed a new angle-independent parameter combining the results of the 45 degrees and the 90 degrees calculation. Small changes in location, extension and sound angle of the ROI did not influence our results significantly. The texture analysis performed by the software was significantly better than the one performed by the examiner. The maximum of brightness was detected in the region of the internal os/anterior lip. Contrast as well as homogeneity showed significant differences between all parts of the cervix. CONCLUSION: The analysis of the grey-scale texture of the cervix can deliver clinically relevant results in the prediction of premature birth. The external influence on sound wave propagation as well as different locations inside the cervix must be taken into consideration.