METHODS: Patients who underwent CBCT at Wonkwang University Daejeon Dental Hospital between January 2019 and December 2022 and were diagnosed with COD and PC by clinical, radiologic, and, if necessary, histopathologic examination were included. Twenty-five patients each were retrospectively enrolled in the COD and PC group. All lesions observed on axial CBCT images were manually segmented using the open-access software MaZda version 4.6 to establish the regions of interest, which were then subjected to texture analysis. Among the 279 texture features obtained, 10 texture features with the highest Fisher coefficients were selected. Statistical analysis was performed using the Mann-Whitney U-test, Welch\'s t-test, or Student\'s t-test. Texture features that showed significant differences were subjected to receiver operating characteristics (ROC) curve analysis to evaluate the differential diagnostic ability of COD and PC.
RESULTS: The COD group consisted of 22 men and 3 women, while the PC group consisted of 14 men and 11 women, showing a significant difference between the two groups in terms of sex (p=0.003). The 10 selected texture features belonged to the gray level co-occurrence matrix and included the sum of average, sum of entropy, entropy, and difference of entropy. All 10 selected texture features showed statistically significant differences (p<0.05) when comparing patients with COD (n=25) versus those with PC (n=25), osteolytic-stage COD (n=11) versus PC (n=25), and osteolytic-stage COD (n=11) versus cementoblastic-stage COD (n=14). ROC curve analysis to determine the ability to differentiate between COD and PC showed a high area under the curve ranging from 0.96 to 0.98.
CONCLUSIONS: Texture analysis of CBCT images has shown good diagnostic value in the differential diagnosis of COD and PC, which can help prevent unnecessary endodontic treatment, invasive biopsy, or surgical intervention associated with increased risk of infection.
方法:2019年1月至2022年12月在旺光大学大田牙科医院接受CBCT检查并通过临床诊断为COD和PC的患者,放射学,and,如有必要,包括组织病理学检查。25例患者均被纳入COD和PC组。在轴向CBCT图像上观察到的所有病变均使用开放获取软件MaZda4.6版手动分割,以建立感兴趣的区域,然后进行纹理分析。在获得的279个纹理特征中,选择具有最高Fisher系数的10个纹理特征。使用Mann-WhitneyU检验进行统计分析,韦尔奇的t检验,或学生的t检验。对显示显着差异的纹理特征进行受试者工作特征(ROC)曲线分析,以评估COD和PC的鉴别诊断能力。
结果:COD组由22名男性和3名女性组成,PC组由14名男性和11名女性组成,显示两组之间在性别方面的显着差异(p=0.003)。选取的10个纹理特征属于灰度共生矩阵,包括平均值之和,熵的总和,熵,和熵的差异。当比较患有COD(n=25)的患者与患有PC(n=25)的患者时,所有10个选定的质地特征均显示出统计学上的显着差异(p<0.05)。溶骨阶段COD(n=11)与PC(n=25),溶骨阶段COD(n=11)与成水泥阶段COD(n=14)。确定区分COD和PC的能力的ROC曲线分析显示出0.96至0.98范围内的高曲线下面积。
结论:CBCT图像的纹理分析在COD和PC的鉴别诊断中显示出良好的诊断价值。这有助于防止不必要的牙髓治疗,侵入性活检,或手术干预与感染风险增加相关。