Search

Objective: Hepatocellular carcinoma (HCC) has become a pressing health problem facing the world today due to its high morbidity, high mortality, and late discovery. As a diagnostic criteria of HCC, the exact threshold of Alpha-fetoprotein (AFP) is controversial. Therefore, this study was aimed to systematically estimate the performance of AFP in diagnosing HCC and to clarify its optimal threshold. Methods: Medline and Embase databases were searched for articles indexed up to November 2019. English language studies were included if both the sensitivity and specificity of AFP in the diagnosis of HCC were provided. The basic information and accuracy data included in the studies were extracted. Combined estimates for sensitivity and specificity were statistically analyzed by random-effects model using MetaDisc 1.4 and Stata 15.0 software at the prespecified threshold of 400 ng/mL, 200 ng/mL, and the range of 20–100 ng/mL. The optimal threshold was evaluated by the area under curve (AUC) of the summary receiver operating characteristic (SROC). Results: We retrieved 29,828 articles and included 59 studies and 1 review with a total of 11,731 HCC cases confirmed by histomorphology and 21,972 control cases without HCC. The included studies showed an overall judgment of at risk of bias. Four studies with AFP threshold of 400 ng/mL showed the summary sensitivity and specificity of 0.32 (95%CI 0.31–0.34) and 0.99 (95%CI 0.98–0.99), respectively. Four studies with AFP threshold of 200 ng/mL showed the summary sensitivity and specificity of 0.49 (95%CI 0.47–0.50) and 0.98 (95%CI 0.97–0.99), respectively. Forty-six studies with AFP threshold of 20–100 ng/mL showed the summary sensitivity and specificity of 0.61 (95%CI 0.60–0.62) and 0.86 (95%CI 0.86–0.87), respectively. The AUC of SROC and Q index of 400 ng/mL threshold were 0.9368 and 0.8734, respectively, which were significantly higher than those in 200 ng/mL threshold (0.9311 and 0.8664, respectively) and higher than those in 20–100 ng/mL threshold (0.8330 and 0.7654, respectively). Furthermore, similar result that favored 400 ng/mL were shown in the threshold in terms of AFP combined with ultrasound. Conclusion: AFP levels in serum showed good accuracy in HCC diagnosis, and the threshold of AFP with 400 ng/mL was better than that of 200 ng/mL in terms of sensitivity and specificity no matter AFP is used alone or combined with ultrasound. [ABSTRACT FROM AUTHOR]

Introduction: Ovarian tumors are the most common diagnostic challenge for gynecologists and ultrasound examination has become the main technique for assessment of ovarian pathology and for preoperative distinction between malignant and benign ovarian tumors. However, ultrasonography is highly examiner-dependent and there may be an important variability between two different specialists when examining the same case. The objective of this work is the evaluation of different well-known Machine Learning (ML) systems to perform the automatic categorization of ovarian tumors from ultrasound images. Methods: We have used a real patient database whose input features have been extracted from 348 images, from the IOTA tumor images database, holding together with the class labels of the images. For each patient case and ultrasound image, its input features have been previously extracted using Fourier descriptors computed on the Region Of Interest (ROI). Then, four ML techniques are considered for performing the classification stage: K-Nearest Neighbors (KNN), Linear Discriminant (LD), Support Vector Machine (SVM) and Extreme Learning Machine (ELM). Results: According to our obtained results, the KNN classifier provides inaccurate predictions (less than 60% of accuracy) independently of the size of the local approximation, whereas the classifiers based on LD, SVM and ELM are robust in this biomedical classification (more than 85% of accuracy). Conclusions: ML methods can be efficiently used for developing the classification stage in computer-aided diagnosis systems of ovarian tumor from ultrasound images. These approaches are able to provide automatic classification with a high rate of accuracy. Future work should aim at enhancing the classifier design using ensemble techniques. Another ongoing work is to exploit different kind of features extracted from ultrasound images. [ABSTRACT FROM AUTHOR]