Objective:

To explore the diagnostic value of logistic regression model based on the characteristics of high-frequency ultrasound images in the diagnosis of pathological subtypes of phyllodes tumor of the breast (PTB).

Methods:

The clinical and high-frequency ultrasonographic data of 84 patients with PTB confir

med by operation and pathology were retrospectively analyzed. According to the pathological subtypes

patients were divided into benign

borderline and malignant groups. The quantitative indicators of ultrasound images included the maximum diameter

the ratio of longitudinal and longitudinal diameter

the resistance index and the peak systolic flow rate

while the qualitative indicators included morphology

margin

parenchymal echo

cystic changes

calcification focus and blood flow signal. Nonparametric rank-sum test was used to compare the differences in quantitative and qualitative indicators of ultrasound images among the three groups with different pathological grades

and then multi- classification ordered logistic regression model was used to screen out risk factors affecting pathological severity in the groups with statistical significance.

Results:

Among 84 cases of PTB

43 cases were diagnosed pathologically as benign

28 cases were borderline

and 13 cases were malignant. The comparison of the tumor aspect ratio

morphology

margin

and calcification in the benign

borderline and malignant PTB ultrasound images showed no significant difference (

P

＞0.05); while the comparison of the largest diameter of the lesion

parenchymal echo

cystic changes

blood flow signal

resistance index and peak systolic flow velocity were statistically significant (

P

＜0.05). The results of multi-class ordinal logistic regression showed that cystic degeneration and peaksystolic flow rate were high risk factors affecting the grade of PTB pathological subtypes

and their odds atios (95% CI) were 0.258 (0.027-2.469) and 1.194 (1.058-1.349)

respectively.

Conclusion:

The logistic regression model in high-frequency ultrasound images can provide better help for the pathological grading of PTB

and the focus should be on whether the tumor has cystic degeneration and the peak systolic flow rate.