One step further into the blackbox: a pilot study of how to build more confidence around an AI-based decision system of breast nodule assessment in 2D ultrasound.

OBJECTIVES: To investigate how a DL model makes decisions in lesion classification with a newly defined region of evidence (ROE) by incorporating "explainable AI" (xAI) techniques. METHODS: A data set of 785 2D breast ultrasound images acquired from 367 females. The DenseNet-121 was used to classify whether the lesion is benign or malignant. For performance assessment, classification results are evaluated by calculating accuracy, sensitivity, specificity, and receiver operating characteristic for experiments of both coarse and fine regions of interest (ROIs). The area under the curve (AUC) was evaluated, and the true-positive, false-positive, true-negative, and false-negative results with breakdown in high, medium, and low resemblance on test sets were also reported. RESULTS: The two models with coarse and fine ROIs of ultrasound images as input achieve an AUC of 0.899 and 0.869, respectively. The accuracy, sensitivity, and specificity of the model with coarse ROIs are 88.4%, 87.9%, and 89.2%, and with fine ROIs are 86.1%, 87.9%, and 83.8%, respectively. The DL model captures ROE with high resemblance of physicians' consideration as they assess the image. CONCLUSIONS: We have demonstrated the effectiveness of using DenseNet to classify breast lesions with limited quantity of 2D grayscale ultrasound image data. We have also proposed a new ROE-based metric system that can help physicians and patients better understand how AI makes decisions in reading images, which can potentially be integrated as a part of evidence in early screening or triaging of patients undergoing breast ultrasound examinations. KEY POINTS: • The two models with coarse and fine ROIs of ultrasound images as input achieve an AUC of 0.899 and 0.869, respectively. The accuracy, sensitivity, and specificity of the model with coarse ROIs are 88.4%, 87.9%, and 89.2%, and with fine ROIs are 86.1%, 87.9%, and 83.8%, respectively. • The first model with coarse ROIs is slightly better than the second model with fine ROIs according to these evaluation metrics. • The results from coarse ROI and fine ROI are consistent and the peripheral tissue is also an impact factor in breast lesion classification.

The decreased lncRNA ZEB2-AS1 in pre-eclampsia controls the trophoblastic cell line HTR-8/SVneo's invasive and migratory abilities via the miR-149/PGF axis.

Pre-eclampsia (PE) is a pregnancy disease that causes maternal death and threatens the health of newborns. Accumulating evidence has revealed the essential roles of long noncoding RNAs (lncRNAs) in the progression of PE. The present investigation determined lncRNA ZEB2 antisense RNA 1 (ZEB2-AS1) expression in PE and looked into the potential role of ZEB2-AS1 in modulating trophoblastic cell functions. Quantitative real-time polymerase chain reaction evaluated gene expression. Western blot analyzed the placental growth factor (PGF) protein level. Cell counting kit-8 and Transwell invasion assays assessed the proliferative and invasive abilities of placental trophoblast cells, respectively. Wound healing assay determined cell migratory potentials. Dual-luciferase reporter assay assessed the targeting relationship among ZEB2-AS1, miR-149, and PGF. Downregulation of lncRNA ZEB2-AS1 was detected in placentas from patients with PE when compared with those from normal pregnancies. Moreover, ZEB2-AS1 upregulation markedly promoted proliferative, migratory, and invasive potentials in HTR-8/SVneo cells, while knockdown of ZEB2-AS1 had the opposite effects. The effects on HTR-8/SVneo cells mediated by ZEB2-AS1 was correlated with the miR-149/PGF axis. These findings indicate that ZEB2-AS1 contributes to PE progression by affecting cell proliferative and invasive capacities via the miR-149/PGF axis in HTR-8/SVneo cells. In sum, we identified that ZEB2-AS1 was a novel aberrantly expressed lncRNA in the placentas of PE patients and lncRNA ZEB2-AS1 modulated trophoblastic cell line HTR-8/SVneo's proliferative and invasive potentials via targeting the miR-149/PGF axis.

How much can AI see in early pregnancy: A multi-center study of fetus head characterization in week 10-14 in ultrasound using deep learning.

PURPOSE: To investigate if artificial intelligence can identify fetus intracranial structures in pregnancy week 11-14; to provide an automated method of standard and non-standard sagittal view classification in obstetric ultrasound examination METHOD AND MATERIALS: We proposed a newly designed scheme based on deep learning (DL) - Fetus Framework to identify nine fetus intracranial structures: thalami, midbrain, palate, 4th ventricle, cisterna magna, nuchal translucency (NT), nasal tip, nasal skin, and nasal bone. Fetus Framework was trained and tested on a dataset of 1528 2D sagittal-view ultrasound images from 1519 females collected from Shenzhen People's Hospital. Results from Fetus Framework were further used for standard/non-standard (S-NS) plane classification, a key step for NT measurement and Down Syndrome assessment. S-NS classification was also tested with 156 images from the Longhua branch of Shenzhen People's Hospital. Sensitivity, specificity, and area under the curve (AUC) were evaluated for comparison among Fetus Framework, three classic DL models, and human experts with 1-, 3- and 5-year ultrasound training. Furthermore, 4 physicians with more than 5 years of experience conducted a reader study of diagnosing fetal malformation on a dataset of 316 standard images confirmed by the Fetus framework and another dataset of 316 standard images selected by physicians. Accuracy, sensitivity, specificity, precision, and F1-Score of physicians' diagnosis on both sets are compared. RESULTS: Nine intracranial structures identified by Fetus Framework in validation are all consistent with that of senior radiologists. For S-NS sagittal view identification, Fetus Framework achieved an AUC of 0.996 (95%CI: 0.987, 1.000) in internal test, at par with classic DL models. In external test, FF reaches an AUC of 0.974 (95%CI: 0.952, 0.995), while ResNet-50 arrives at AUC∼0.883, 95% CI 0.828-0.939, Xception AUC∼0.890, 95% CI 0.834-0.946, and DenseNet-121 AUC∼0.894, 95% CI 0.839-0.949. For the internal test set, the sensitivity and specificity of the proposed framework are (0.905, 1), while the first-, third-, and fifth-year clinicians are (0.619, 0.986), (0.690, 0.958), and (0.798, 0.986), respectively. For the external test set, the sensitivity and specificity of FF is (0.989, 0.797), and first-, third-, and fifth-year clinicians are (0.533, 0.875), (0.609, 0.844), and (0.663, 0.781), respectively.On the fetal malformation classification task, all physicians achieved higher accuracy and F1-Score on Fetus selected standard images with statistical significance (p < 0.01). CONCLUSION: We proposed a new deep learning-based Fetus Framework for identifying key fetus intracranial structures. The framework was tested on data from two different medical centers. The results show consistency and improvement from classic models and human experts in standard and non-standard sagittal view classification during pregnancy week 11-13+6. CLINICAL RELEVANCE/APPLICATION: With further refinement in larger population, the proposed model can improve the efficiency and accuracy of early pregnancy test using ultrasound examination.

Up-regulation of miR-299 suppressed the invasion and migration of HTR-8/SVneo trophoblast cells partly via targeting HDAC2 in pre-eclampsia.

Pre-eclampsia (PE), a pregnancy-associated disorder, is a major contributor to maternal mortality and morbidity worldwide. Recently, microRNAs (miRNAs) were found to be associated with the pathogenesis of PE. The present study investigated the function of miR-299 in HTR-8/SVneo trophoblast cells and explored its underlying mechanism in the pathogenesis of PE. The miR-299 and histone deacetylase 2 (HDAC2) mRNA expression levels were determined by quantitative real-time PCR. Transwell invasion and wound healing assays were used to measure cell invasive and migratory ability. Luciferase reporter assay was performed to confirm the downstream targets of miR-299. Western blot was performed to measure the protein expression of HDAC2. The expression level of miR-299 in placental tissues from pregnant women with severe PE was significantly higher than that from normal pregnant women. MiR-299 overexpression suppressed the invasion and migration of trophoblast cells; and knock-down of miR-299 promoted the invasion and migration of trophoblast cells. Bioinformatics prediction and luciferase reporter assay confirmed that miR-299 directly targeted the 3' untranslated region of HDAC2. The mRNA expression level of HDAC2 in placental tissues from pregnant women with severe PE was significantly lower than that from normal pregnant women, and was negatively correlated the expression level of miR-299 in placental tissues from pregnant women with severe PE. HDAC2 siRNA transfection suppressed the mRNA and protein expression levels of HDAC in trophoblast cells compared with control group, and HDAC2 siRNA transfection also suppressed the trophoblast cell invasion and migration compared with control group. Enforced expression of HDAC2 in trophoblast cells attenuated the inhibitory effects of miR-299 overexpression on cell invasion and migration. In summary, the results showed the up-regulation of miR-299 in the placental tissues from women with severe PE and miR-299 suppressed the invasion and migration of trophoblast cells partly via targeting HDAC2.

Gestational diabetes mellitus is associated with decreased adipose and placenta peroxisome proliferator-activator receptor γ expression in a Chinese population.

Peroxisome proliferator-activated receptors γ (PPARγ) is a member of nuclear receptor superfamily, and studies have demonstrated that dysregulation of PPARγ was associated with gestational diabetes mellitus (GDM), which is one of the most common metabolic abnormalities occurring during pregnancy. However, the results regarding the associations between PPARγ and GDM were conflicting among different studies. The present study aimed to determine the expression of PPARγ in adipose and placenta from GDM women in a Chinese population and to further explore the role of PPARγ in GDM women. The adipose and placenta tissues were isolated from GDM women and healthy pregnant women at term. The mRNA and protein expressions of PPARγ in adipose and placenta tissues were determined by qRT-PCR and western blot, respectively. Univariate correlation analysis was used to analyze the relationship between PPARγ expression and clinical characteristics of patients. The levels of tryglycerides and HbA1c were significantly higher, while the levels of low density lipoprotein (LDL) cholesterol, adiponectin and insulin were significantly lower in the GDM women than that in the healthy pregnant women. The mRNA and protein expression of PPARγ in both adipose and placenta from GDM women were significantly lower than that from healthy pregnant women. PPARγ mRNA expression in both adipose and placenta positively correlated with LDL cholesterol and adiponectin levels, and negatively correlated with tryglycerides and glucose levels at 0 h, 1 h and 2 h of 75 g oral glucose tolerance test. In summary, our results suggest that PPARγ may be a key modulator in the development of GDM, due to the roles of PPARγ in glucose homeostasis and adipose tissue development and function.