Multi-modality deep learning model reaches high prediction accuracy in the diagnosis of ovarian cancer.

We evaluated the diagnostic performance of a multimodal deep-learning (DL) model for ovarian mass differential diagnosis. This single-center retrospective study included 1,054 ultrasound (US)-detected ovarian tumors (699 benign and 355 malignant). Patients were randomly divided into training (n = 675), validation (n = 169), and testing (n = 210) sets. The model was developed using ResNet-50. Three DL-based models were proposed for benign-malignant classification of these lesions: single-modality model that only utilized US images; dual-modality model that used US images and menopausal status as inputs; and multi-modality model that integrated US images, menopausal status, and serum indicators. After 5-fold cross-validation, 210 lesions were tested. We evaluated the three models using the area under the curve (AUC), accuracy, sensitivity, and specificity. The multimodal model outperformed the single- and dual-modality models with 93.80% accuracy and 0.983 AUC. The Multimodal ResNet-50 DL model outperformed the single- and dual-modality models in identifying benign and malignant ovarian tumors.

Diagnosis of Deep Infiltrating Endometriosis Using Transvaginal Ultrasonography.

Objective: To evaluate the performance of transvaginal ultrasonography (TVS) in diagnosis of deep infiltrating endometriosis (DIE). Methods: We retrospectively analyzed 198 patients with pathological diagnosis of DIE in our hospital from January 2017 to December 2019 and assessed the performances of pre-operative TVS diagnosis of DIE with regarding to sensitivity (SE), specificity (SP), positive predictive value (PPV), and negative predictive value (NPV), using the pathological diagnosis as the ground truth. We also characterized the ultrasonographic features of the DIE lesions. Results: Among all the 198 cases, 170 cases were uterosacral ligament (USL) involvement, SE: 96.47% and SP: 85.71%; 79 cases were intestinal involvement, SE: 94.94% and SP: 94.96%; 57 cases were vaginal rectal septum (VRS) involvement, SE: 73.68% and SP: 94.33%; 20 cases were vaginal involvement, SE: 50% and SP: 97.21%; three cases were bladder involvement, SE: 66.7% and SP: 100%; nine cases were ureter involvement, SE: 55.56% and SP: 100%; and 10 cases were broad ligament involvement, SE: 10% and SP: 100%. Conclusion: TVS showed high accuracy in diagnosing DIE.

Evaluation of Uterosacral Ligament Involvement in Deep Endometriosis by Transvaginal Ultrasonography.

This study was designed to conclude the ultrasonic characteristics of uterosacral ligament (USL) lesions involved by endometriosis and evaluated the value of transvaginal sonography (TVS) in diagnosing USL involvement in deep infiltrating endometriosis (DIE). A total of one hundred and eighteen patients with DIE were included in the study and underwent surgery. All these patients were evaluated by transvaginal ultrasound examination by one trained examiner. The gold standard for diagnosis was surgery and histopathology. 85 patients with USL endometriosis were confirmed by surgical pathology. 84 patients were diagnosed USL endometriosis by TVS and 81 of which were confirmed by the gold standard. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy of TVS for diagnosing USL endometriosis were 95.3, 90.9, 96.4, 88.2, and 94.1%, respectively. According to the ultrasound characteristics of USL endometriosis, we summarized four types: Type I. thickened and stiff lesions, Type II. local nodules, Type III. irregular striped lesions, and Type IV. mixed lesions. The conclusion of the study was that TVS was a convenient, accurate and first-line diagnostic technique for USL endometriosis and the USL lesions could be summarized into four types according to the ultrasound morphological changes.

A fast kernel extreme learning machine based on conjugate gradient.

Kernel extreme learning machine (KELM) introduces kernel leaning into extreme learning machine (ELM) in order to improve the generalization ability and stability. But the Penalty parameter in KELM is randomly set and it has a strong impact on the performance of KELM. A fast KELM combining the conjugate gradient method (CG-KELM) is presented in this paper. The CG-KELM computes the output weights of the neural network by the conjugate gradient iteration method. There is no penalty parameter to be set in CG-KELM. Therefore, the CG-KELM has good generalization ability and fast learning speed. The simulations in image restoration show that CG-KELM outperforms KELM. The CG-KELM provides a balanced method between KELM and ELM.

Contrast-enhanced ultrasonography follow-up after radiofrequency ablation in normal rabbit kidney.

INTRODUCTION: The aim of this study was to evaluate the diagnostic performance of SonoVue-enhanced ultrasonography in the follow-up of rabbit kidney lesions induced by percutaneous radiofrequency ablation. MATERIAL AND METHODS: New Zealand rabbits (28) underwent percutaneous radiofrequency ablation (RFA) to generate renal lesions. Lesions were evaluated by conventional 2D ultrasound and contrast-enhanced ultrasonography (CEUS) at 1 day, 1 week, 1 month and 3 months after RFA, and the results were compared to gross pathology. RESULTS: One day and 1 week after RFA, renal lesions were wedge-shaped in the gross dissection cross-sectional profiles. Conventional ultrasound could not indicate the extent of the lesions; however, CEUS could exactly delineate the lesion shape and size. At 1 and 3 months, lesions were observed as hyperechogenic areas on conventional ultrasound, and as small perfusion defects on CEUS. The differences in the lesion measurements obtained by CEUS and in pathological specimens were not statistically significant (p > 0.05). CONCLUSIONS: The conclusion could be deduced from the study that SonoVue-enhanced ultrasonography was effective for the follow-up of normal rabbit kidney percutaneous radiofrequency ablation.

Detection of small inflammatory and metastatic lymph nodes in rabbits by sonography: evaluation with the resistive index ratio.

OBJECTIVE: The aim of this prospective study was to determine whether the resistive index (RI) ratio based on high-resolution spectral Doppler sonography would be useful in the differential diagnosis of small inflammatory and metastatic lymph nodes in rabbit models. METHODS: The infected and metastatic lymph node models we used were created by subcutaneously inoculating methicillin-susceptible Staphylococcus aureus and VX2 tumor cells respectively into the left hind limbs of 10 New Zealand White rabbits. High-resolution sonography was performed to investigate the popliteal fossa lymph nodes 2 weeks after the inoculation. The sizes, long-/short-axis ratios, and RI ratios (defined as the value of the peripheral RI relative to the central RI) of the nodes were evaluated with sonography and then compared with the histopathologic findings. RESULTS: No statistical differences were found between the volumes (mean +/- SD, 104 +/- 41 versus 87 +/- 24 mm(3); P > .05) and the long-/short-axis ratios (1.97 +/- 0.28 versus 2.03 +/- 0.26; P > .05) of 15 inflammatory and 14 metastatic lymph nodes. On spectral Doppler sonography, the RI ratio was higher in the metastatic lymph nodes than in the inflammatory lymph nodes. With an RI ratio higher than 1.2 as a diagnostic criterion for a metastatic lymph node, the sensitivity, specificity, and positive and negative predictive values of sonography were 71% (10/14), 80% (12/15), 76% (10/13), and 75% (12/16), respectively, with an area under the receiver operating characteristic curve value of 0.824. CONCLUSIONS: This experimental study confirms that RI ratio changes based on high-resolution spectral Doppler sonography are associated with histopathologic changes of metastatic and inflammatory lymph nodes during the initial stage in rabbits.