Purpose: We aimed to make a fast and accurate distinction of malignant and benign lesions in cases with predominantly solitary or multifocal involvement using latest technology software and hardware systems in compute...Purpose: We aimed to make a fast and accurate distinction of malignant and benign lesions in cases with predominantly solitary or multifocal involvement using latest technology software and hardware systems in computed tomography. Materials and Methods: 53 cases were included in the study. Primary (n = 42, 31 benign, 11 malignant) or metastatic (n = 11) tumors were detected at various locations in the bone structure of the cervical to coccygeal vertebrae in all cases. 3D CT images taken using the same system and biopsy or post-operative histopathology findings were available for all cases. Thin section images taken retrospectively from the archives were converted to 3D images using the same program and parameters, which were then recorded in the same window settings by two radiologists. Only 3D images were then analyzed to investigate the presence or absence of the dirty interface sign. Results: Dirty interface sign was present in 17 malignant lesions and absent in the remaining 5 lesions. As for benign lesions, the sign was present in only two lesions and the remaining 29 were negative for the sign. There was a high level of consistency between the two radiologists. In conclusion, malignant and benign lesions affecting the bone spinal axis were distinguished based on the presence or absence of the dirty interface sign with 77.3% sensitivity, 93.5% specificity and 86.8% accuracy. Conclusion: When evaluated with standard bone window views, 3D views can be used successfully for the distinction of malignant and benign bone tumors. At least, 3D views generated using low dose regimes in highly developed systems can be used with similar purpose to that of diffusion weighted MRI sequences that give roughly outlined but fast and accurate information about the lesion.展开更多
The stage of a tumor is sometimes hard to predict, especially early in its development. The size and complexity of its observations are the major problems that lead to false diagnoses. Even experienced doctors can mak...The stage of a tumor is sometimes hard to predict, especially early in its development. The size and complexity of its observations are the major problems that lead to false diagnoses. Even experienced doctors can make a mistake in causing terrible consequences for the patient. We propose a mathematical tool for the diagnosis of breast cancer. The aim is to help specialists in making a decision on the likelihood of a patient’s condition knowing the series of observations available. This may increase the patient’s chances of recovery. With a multivariate observational hidden Markov model, we describe the evolution of the disease by taking the geometric properties of the tumor as observable variables. The latent variable corresponds to the type of tumor: malignant or benign. The analysis of the covariance matrix makes it possible to delineate the zones of occurrence for each group belonging to a type of tumors. It is therefore possible to summarize the properties that characterize each of the tumor categories using the parameters of the model. These parameters highlight the differences between the types of tumors.展开更多
目的:探讨基于患者临床信息的logistic回归模型在乳腺影像报告和数据系统(Breast Imaging Reporting and Data System,BI-RADS)4类中鉴别病灶良恶性的价值。方法:回顾并收集经过病理学检查证实的BI-RADS4类乳腺病灶患者221例(良性133例...目的:探讨基于患者临床信息的logistic回归模型在乳腺影像报告和数据系统(Breast Imaging Reporting and Data System,BI-RADS)4类中鉴别病灶良恶性的价值。方法:回顾并收集经过病理学检查证实的BI-RADS4类乳腺病灶患者221例(良性133例,恶性88例)的临床信息。采用logistic回归分析筛选能够鉴别病灶良恶性的临床信息特征,建立回归模型。比较BI-RADS联合模型与单独采用BI-RADS分类在鉴别乳腺良恶性病灶上的区别。结果:经logistic回归分析,发现9个临床信息特征与乳腺良恶性病灶相关,其中是否触及病灶(OR=7.196)、病灶是否固定(OR=10.150)、病灶最大径是否>2 cm(OR=4.208)等3个特征有较高的危险度(P<0.05)。单独采用BI-RADS分类,其诊断灵敏度为86.3%、特异度为69.9%、准确率为76.5%;将BI-RADS分类联合回归模型,其灵敏度为88.6%、特异度为73.7%、准确率为79.6%。结论:BI-RADS分类联合基于患者临床信息的logistic回归模型有助于提高鉴别乳腺病灶良恶性的诊断效能,减少不必要的良性活检。展开更多
文摘Purpose: We aimed to make a fast and accurate distinction of malignant and benign lesions in cases with predominantly solitary or multifocal involvement using latest technology software and hardware systems in computed tomography. Materials and Methods: 53 cases were included in the study. Primary (n = 42, 31 benign, 11 malignant) or metastatic (n = 11) tumors were detected at various locations in the bone structure of the cervical to coccygeal vertebrae in all cases. 3D CT images taken using the same system and biopsy or post-operative histopathology findings were available for all cases. Thin section images taken retrospectively from the archives were converted to 3D images using the same program and parameters, which were then recorded in the same window settings by two radiologists. Only 3D images were then analyzed to investigate the presence or absence of the dirty interface sign. Results: Dirty interface sign was present in 17 malignant lesions and absent in the remaining 5 lesions. As for benign lesions, the sign was present in only two lesions and the remaining 29 were negative for the sign. There was a high level of consistency between the two radiologists. In conclusion, malignant and benign lesions affecting the bone spinal axis were distinguished based on the presence or absence of the dirty interface sign with 77.3% sensitivity, 93.5% specificity and 86.8% accuracy. Conclusion: When evaluated with standard bone window views, 3D views can be used successfully for the distinction of malignant and benign bone tumors. At least, 3D views generated using low dose regimes in highly developed systems can be used with similar purpose to that of diffusion weighted MRI sequences that give roughly outlined but fast and accurate information about the lesion.
文摘The stage of a tumor is sometimes hard to predict, especially early in its development. The size and complexity of its observations are the major problems that lead to false diagnoses. Even experienced doctors can make a mistake in causing terrible consequences for the patient. We propose a mathematical tool for the diagnosis of breast cancer. The aim is to help specialists in making a decision on the likelihood of a patient’s condition knowing the series of observations available. This may increase the patient’s chances of recovery. With a multivariate observational hidden Markov model, we describe the evolution of the disease by taking the geometric properties of the tumor as observable variables. The latent variable corresponds to the type of tumor: malignant or benign. The analysis of the covariance matrix makes it possible to delineate the zones of occurrence for each group belonging to a type of tumors. It is therefore possible to summarize the properties that characterize each of the tumor categories using the parameters of the model. These parameters highlight the differences between the types of tumors.
文摘目的:探讨基于患者临床信息的logistic回归模型在乳腺影像报告和数据系统(Breast Imaging Reporting and Data System,BI-RADS)4类中鉴别病灶良恶性的价值。方法:回顾并收集经过病理学检查证实的BI-RADS4类乳腺病灶患者221例(良性133例,恶性88例)的临床信息。采用logistic回归分析筛选能够鉴别病灶良恶性的临床信息特征,建立回归模型。比较BI-RADS联合模型与单独采用BI-RADS分类在鉴别乳腺良恶性病灶上的区别。结果:经logistic回归分析,发现9个临床信息特征与乳腺良恶性病灶相关,其中是否触及病灶(OR=7.196)、病灶是否固定(OR=10.150)、病灶最大径是否>2 cm(OR=4.208)等3个特征有较高的危险度(P<0.05)。单独采用BI-RADS分类,其诊断灵敏度为86.3%、特异度为69.9%、准确率为76.5%;将BI-RADS分类联合回归模型,其灵敏度为88.6%、特异度为73.7%、准确率为79.6%。结论:BI-RADS分类联合基于患者临床信息的logistic回归模型有助于提高鉴别乳腺病灶良恶性的诊断效能,减少不必要的良性活检。
