A Deep Learning Approach for Classifying Benign, Malignant, and Borderline Ovarian Tumors Using Convolutional Neural Networks and Generative Adversarial Networks.

A systematic review and meta-analysis of the use of ultrasound to diagnose borderline ovarian tumours

Ovarian tumors of low malignant potential, often termed "borderline tumors," have been defined as those that have some but not all of the morphologic features of malignancy (i.e., they are not invasive). With the use of data obtained by the western Washington population-based Cancer Surveillance System for 1975-83, the incidence of serous and mucinous borderline epithelial ovarian tumors was analyzed, as well as the survival of women who developed them. The incidence of borderline tumors increased with increasing age, although at a pace somewhat slower than that of malignant ovarian tumors. There was an upward trend in the incidence of borderline tumors starting in the late 1970's, a trend not present for malignant tumors. Only 12% of borderline tumors were not confined to the ovary, as opposed to 40% of malignant Grade I and 73% of other malignant ovarian neoplasms. At 5 years following diagnosis, the survival of women with borderline tumors was 93% that of the general female population. This percentage varied little by stage or histologic type. Given the reduced survival of women with these ovarian tumors and the lack of a sharp histologic distinction between borderline and Grade I malignant lesions, it is recommended that borderline ovarian tumors be routinely ascertained by population-based cancer registries.

5564 Background: The objective of this study was to evaluate accuracy of frozen section in borderline ovarian tumors and to determine the tumor characteristics that lead to higher likelihood of inaccurate intraoperative diagnosis (IAIOD). IAIOD is a clinical problem that restricts the diagnostic accuracy of frozen section in borderline ovarian tumors. Methods: This was a retrospective chart review of 622 consecutive cases that were diagnosed with pelvic mass and underwent surgery at busy gynecology services of two institutions, between 2006-2011. Of these cases, 52 were diagnosed as borderline ovarian tumors by frozen section. Experienced pathologists performed frozen section with second opinion from specialized gynecologic pathologists as needed. Terms such as “at least borderline” were also evaluated to help stratify patients. Frozen section and final permanent histology reports were compared. Patient and tumor characteristics that may cause IAIOD such as age of patient, histological subtype, size of tumor, bilaterality, CA-125 levels were studied. Staging was performed when borderline or malignant ovarian tumors were identified by frozen section. Results: Agreement of the frozen section results with final pathology was observed in 37 out of 52 patients with a diagnostic accuracy of 71.15 %. Under diagnosis occurred in 12 out of 52 patients and over diagnosis occurred in 3 out of 52 patients. Age >40 years, size of tumor >5 cm, bilaterality of tumors and CA-125 were not found significant in causing IAIOD when chi-square analysis was performed. Characterization by “at-least” borderline terminology at the time of frozen section did not help identify patients with higher likelihood of IAIOD. Conclusions: In our study, the rate of IAIOD was high, at 23% despite experienced pathologists and using “at-least borderline” terminology. Traditionally described features leading to inaccuracy with frozen sections, such as large tumors with mucinous histology did not increase the risk of IAIOD in this study. Patients with unilateral, small tumors and non-mucinous histology had greatest risk for IAIOD. Given this information, full staging of all borderline ovarian tumors identified at time of frozen section should remain the standard of practice.

Oxidative stress-induced antioxidant enzyme expression is an early phenomenon in ovarian carcinogenesis

paper, we focus on exploring the relationship between weave patterns and their mechanical properties in woven fiber composites through Machine Learning. Specifically, we explore the interactions between woven architectures and in-plane stiffness properties through Deep Convolutional Neural Network (DCNN) and Generative Adversarial Network (GAN). Our research is important for exploring how woven composite’s pattern is related to its mechanical properties and accelerating woven composite design as well as optimization. We focus on two tasks: (1) Stiffness prediction: Predicting in-plane stiffness properties for given weave patterns. Our DCNN extracts high-level features through several convolutional and fully connected layers to determine the final predictions. (2) Weave pattern prediction: Predicting weave patterns for target stiffness properties, which can be treated as the reverse task of the first one. Due to many-to-one mapping between weave patterns and the composite properties, we utilize a Decoder Neural Network as our baseline model and compare its performance with GAN and Genetic Algorithm. We represent the weave patterns as 2D checkerboard models and use finite element analysis (FEA) to determine in-plane stiffness properties, which serve as input data for our ML framework. We show that: (1) for stiffness prediction, DCNN can predict stiffness values for a given weave pattern with relatively high accuracy (above 93%); (2) for weave pattern prediction, the GAN model gives the best prediction accuracy (above 92%) while Decoder Neural Network has the best time efficiency. HAOTIAN FENG

Borderline ovarian tumors: Guidelines from the French national college of obstetricians and gynecologists (CNGOF)

Deep learning (DL) algorithms could improve the classification of ovarian tumours assessed with ultrasound (US) images and clinical information. This study aimed to develop a DL model for diagnosing ovarian tumours based on US images and clinical information, and to compare the performance of the DL model with radiologist assessment. This study retrospectively collected data from four hospitals involving women who underwent US examinations for ovarian tumours. Additionally, data were prospectively randomly collected from two other hospitals. Histopathological analysis served as the reference standard. The retrospective dataset was divided into training, test and internal validation sets. The prospective dataset was used as an external validation set. A DL model was developed. The performance of the DL model was compared with the eleven radiologists’ assessments in the internal and external validation sets. The performance of radiologists' assessments with or without DL model were also compared. A total of 1518 women (6432 images) with ovarian tumours were collected, including 364 women in the internal validation dataset and 340 women in the external validation dataset. In the external validation dataset, the accuracy for the DL model's classification of benign, borderline, and malignant ovarian tumours were 85.5%, which were comparable to the expert assessments (85.3%, p > 0.05), but significantly higher than the average level of eleven different-experience radiologists (72.5%, p < 0.05). The diagnostic performance of junior and mid-level experienced radiologists were significantly improved after applying the DL model. The DL model developed using US images and clinical information can effectively classify benign, malignant, and borderline ovarian tumours with diagnostic performance comparable to the expert assessment and improving junior and mid-level experienced radiologists.

The aim of the study was to determine values of a quantitative morphometry analysis of nuclear characteristics and argyrophilic nucleolar organizer regions (AgNORs) in differential cytodiagnosis of benign, atypically proliferating (borderline) and malignant serous ovarian tumours. Cytological imprints of benign (n = 20), borderline (n = 19) and malignant (n = 20) ovarian serous tumours were analysed. A computerized, digital analysis was used to determine morphometric nuclear features, the number and characteristics of single AgNORs, cluster AgNORs, total AgNOR and AgNOR area/nucleus (relative area) ratio. According to their size AgNORs were classified in three categories. A one-way variance analysis and post hoc test (Scheffé) were used for statistical analysis. The morphometric nuclear analysis showed that benign, borderline and malignant serous ovarian tumours are statistically different (P < 0.001) according to the area and outline, the values being highest in malignant tumours and lowest in the borderline group. Digital analysis of AgNORs in benign, borderline and malignant groups showed that the total AgNOR number increases with progression of the lesion (meaning tumour malignancy) significantly (P < 0.001) between benign and malignant as well as between borderline and malignant serous ovarian tumours (P < 0.001). The progression of the lesion malignancy was accompanied by a significant (P < 0.001) progressive increase of the total and relative AgNOR area per nucleus. The AgNOR size increases from benign to malignant tumours and a statistically significant difference (P < 0.001) was observed in all three groups regarding small and large AgNORs. Combining different markers of morphometric nuclear characteristics and AgNOR values could improve differential cytodiagnosis of benign, borderline and malignant serous ovarian tumours.

Borderline Ovarian tumours (BOTs) are a rare and heterogenous histological group which can carry significant malignant potential if not treated properly. Effective preoperative diagnosis can be challenging however IOTA ADNEX model is capable of differentiating between subgroups of malignancy. Pattern recognition by expert examiner is thought to be the best diagnostic method. The aim of this study was to check the diagnostic efficacy of the preoperative US algorithm based on the mentioned methods. In a tertiary university hospital's ultrasound specialised outpatient 611 non-selected pre- and postmenopausal patients were scanned by the same experienced, IOTA certified examiner. ADNEX score was calculated in all cases where it was possible (age restriction under 14 years) but the subjective opinion of the examiner was recorded in all cases. According to the subjective assessment 81 cases (13.26%) were selected as a candidate for being BOT. Of those who underwent operation BOT was diagnosed in 18 cases (22.22%): serous BOT in 13 cases (72.22%), mucinous BOT in 4 cases (22.22%), rare type BOT in one case (5.56%). Malignancy was detected in 11 (13.58%) further cases: serous adenocarcinoma in 5 cases (45.45%), mucinous adenocarcinoma in 3 cases (27.27%), endometrioid tumour in 2 cases (18.18%) and granulosa cell tumour in one case (9.09%). Benign lesions were reported in all other cases. There was only one case without any preoperative finding for BOT which turned to be borderline in the histopathological evaluation. The number of papillary projections, the presence of blood flow and the ADNEX score all support the BOT diagnosis. IOTA terms and definitions and IOTA models such ADNEX can be a reliable tool for the selection of high-risk population of patients carrying risk of malignancy. Experienced examiners' pattern recognition can elevate this diagnostic performance for a more sophisticated preoperative algorithm allowing a more powerful discrimination from malignant lesions.

Generative models have been applied in many fields and can be evaluated with many methods. In the evaluation of generative models, the proper evaluation metric varies with the application field. Therefore, the evaluation of generative adversarial networks is inherently challenging. In this study, conditional deep convolutional generative adversarial networks were applied in mechanical fault diagnosis and then evaluated. We proposed three evaluation metrics of conditional deep convolutional generative adversarial networks: Jensen-Shannon divergence, kernel maximum mean discrepancy, and the 1-nearest neighbor classifier which were used to distinguish generated samples from real samples, test mode collapsing and detect overfitting based on the dataset of Electronic Engineering Laboratory of Case Western Reserve University and the planetary gearbox dataset measured in the laboratory. The Jensen-Shannon divergence could not well distinguish generated samples from real samples. However, the two metrics (maximum mean discrepancy and 1-nearest neighbor classifier) well-distinguished generated samples from real samples, thus verifying the applicability of conditional deep convolutional generative adversarial networks in the field of mechanical diagnosis.

The present study aims to compare the clinical and reproductive outcomes of women who have undergone lymph node dissection with those who have not and to discuss the indications for retroperitoneal evaluation in the treatment of borderline ovarian tumors. The present study investigated 123 women who were finally diagnosed with borderline epithelial ovarian tumors at the study center between January 1999 and January 2009. A total of 68 patients (55.3%) were found to have a complete surgical staging procedure. The patients who underwent complete lymph node dissection were significantly younger than those who did not have lymphadenectomy. Thus, the ratio of postmenopausal women was significantly higher among those in whom retroperitoneal evaluation was performed. As expected, the intraoperative blood loss was significantly more and the hospital stay was significantly longer in patients who underwent complete lymph node dissection. Also, the non-serous borderline ovarian tumors were significantly more frequent in the latter patient group. However, overall and disease-free survival spans were found to be statistically similar for both study groups as well as the recurrence rate. Retroperitoneal evaluation can be spared in every woman with borderline ovarian tumor unless she is to undergo complete surgical staging procedure because of high recurrence risk, advanced stage disease or personal choice.

Breast cancer is the leading cause of death in women worldwide; prevention of possible death from breast cancer can be decreased by early identification ultrasound image analysis by classifying ultrasound images into three classes (Normal, Benign, and Malignant), where the dataset used has imbalanced data. Imbalanced data cause the classification system only to recognize the majority class, so it is necessary to handle imbalanced data. In this study, imbalanced data can be handled by implementing the Deep Convolution Generative Adversarial Network (DCGAN) method as the addition of synthetic images to the training data. The DCGAN method generates synthetic images with feature learning on a Convolutional Neural Network (CNN), making DCGAN more stable than the basic generative adversarial network method. Synthetic and original images were further classified using the CNN GoogleNet method, which performs well in image classification and with reasonable computation cost. Synthetic ultrasound images were generated using a tuning hyperparameter in the DCGAN method to adjust the input size on GoogleNet for imbalanced data handling. From the experiment result, the implementation of DCGAN-GoogleNet has a higher accuracy in handling imbalanced data than conventional augmentation and other previous research, with an accuracy value reaching 91.61%, which is 1% to 4% higher than the accuracy value in the previous method.

Due to the real working conditions, the collected mechanical fault datasets are actually limited and always highly imbalanced, which restricts the diagnosis accuracy and stability. To solve these problems, we present an imbalanced fault diagnosis method based on the generative model of conditional-deep convolutional generative adversarial network (C-DCGAN) and provide a study in detail. Deep convolutional generative adversarial network (DCGAN), based on traditional generative adversarial networks (GAN), introduces convolutional neural network into the training for unsupervised learning to improve the effect of generative networks. Conditional generative adversarial network (CGAN) is a conditional model obtained through introducing conditional extension into GAN. C-DCGAN is a combination of DCGAN and CGAN. In C-DCGAN, based on the feature extraction ability of convolutional networks, through the structural optimization, conditional auxiliary generative samples are used as augmented data and applied in machine fault diagnosis. Two datasets (Bearing dataset and Planetary gear box dataset) are carried out to validate. The simulation experiments showed that the improved performance is mainly due to the generated signals from C-DCGAN to balance the dataset. The proposed method can deal with imbalanced fault classification problem much more effectively. This model could improve the accuracy of fault diagnosis and the generalization ability of the classifier in the case of small samples and display better fault diagnosis performance.

We conducted an evaluation of incidence and survival of women with borderline ovarian tumors in Sweden. All women diagnosed with borderline ovarian tumor in the Swedish Cancer Register 1960-2007 (n = 6252) combined with follow up in the Swedish Death Registry to 1 July 2009 were included. Estimation of age-standardized relative survival rate according to time periods for diagnosis. The incidence of borderline ovarian tumors increased during the study period, with a steep increase during the 1980s. The age standardized 5-year relative survival including all borderline tumors diagnosed 2000-07 was 97% (95% CI 92-99%). In women aged ≤64 years, the 10-year relative survival related to age at diagnosis of borderline tumors ranged from 95 to 98% and was 89% in women aged 65-74 years. In a multivariable analysis including age and decade of diagnosis relative survival for every decade increased. The 10-year relative survival in women with mucinous and serous borderline tumors did not differ significantly (p = 0.121). Results of the present study are reassuring about long-term survival in women with borderline ovarian tumors. The age-standardized relative survival rate increased across time periods for diagnosis. There was no difference in long-term survival between mucinous and serous borderline ovarian tumors.

Thanks to the availability of large-scale data, deep Convolutional Neural Networks (CNNs) have witnessed success in various applications of computer vision. However, the performance of CNNs on Synthetic Aperture Radar (SAR) image classification is unsatisfactory due to the lack of well-labeled SAR data, as well as the differences in imaging mechanisms between SAR images and optical images. Therefore, this paper addresses the problem of SAR image classification by employing the Generative Adversarial Network (GAN) to produce more labeled SAR data. We propose special GANs for generating SAR images to be used in the training process. First, we incorporate the quadratic operation into the GAN, extending the convolution to make the discriminator better represent the SAR data; second, the statistical characteristics of SAR images are integrated into the GAN to make its value function more reasonable; finally, two types of parallel connected GANs are designed, one of which we call PWGAN, combining the Deep Convolutional GAN (DCGAN) and Wasserstein GAN with Gradient Penalty (WGAN-GP) together in the structure, and the other, which we call CNN-PGAN, applying a pre-trained CNN as a discriminator to the parallel GAN. Both PWGAN and CNN-PGAN consist of a number of discriminators and generators according to the number of target categories. Experimental results on the TerraSAR-X single polarization dataset demonstrate the effectiveness of the proposed method.