TauFlowNet: Revealing latent propagation mechanism of tau aggregates using deep neural transport equations.

Mounting evidence shows that Alzheimer's disease (AD) is characterized by the propagation of tau aggregates throughout the brain in a prion-like manner. Since current pathology imaging technologies only provide a spatial mapping of tau accumulation, computational modeling becomes indispensable in analyzing the spatiotemporal propagation patterns of widespread tau aggregates from the longitudinal data. However, current state-of-the-art works focus on the longitudinal change of focal patterns, lacking a system-level understanding of the tau propagation mechanism that can explain and forecast the cascade of tau accumulation. To address this limitation, we conceptualize that the intercellular spreading of tau pathology forms a dynamic system where each node (brain region) is ubiquitously wired with other nodes while interacting with the build-up of pathological burdens. In this context, we formulate the biological process of tau spreading in a principled potential energy transport model (constrained by brain network topology), which allows us to develop an explainable neural network for uncovering the spatiotemporal dynamics of tau propagation from the longitudinal tau-PET scans. Specifically, we first translate the transport equation into a GNN (graph neural network) backbone, where the spreading flows are essentially driven by the potential energy of tau accumulation at each node. Conventional GNNs employ a l2-norm graph smoothness prior, resulting in nearly equal potential energies across nodes, leading to vanishing flows. Following this clue, we introduce the total variation (TV) into the graph transport model, where the nature of system's Euler-Lagrange equations is to maximize the spreading flow while minimizing the overall potential energy. On top of this min-max optimization scenario, we design a generative adversarial network (GAN-like) to characterize the TV-based spreading flow of tau aggregates, coined TauFlowNet. We evaluate our TauFlowNet on ADNI and OASIS datasets in terms of the prediction accuracy of future tau accumulation and explore the propagation mechanism of tau aggregates as the disease progresses. Compared to the current counterpart methods, our physics-informed deep model yields more accurate and interpretable results, demonstrating great potential in discovering novel neurobiological mechanisms through the lens of machine learning.

Developing Explainable Deep Model for Discovering Novel Control Mechanism of Neuro-Dynamics.

Human brain is a complex system composed of many components that interact with each other. A well-designed computational model, usually in the format of partial differential equations (PDEs), is vital to understand the working mechanisms that can explain dynamic and self-organized behaviors. However, the model formulation and parameters are often tuned empirically based on the predefined domain-specific knowledge, which lags behind the emerging paradigm of discovering novel mechanisms from the unprecedented amount of spatiotemporal data. To address this limitation, we sought to link the power of deep neural networks and physics principles of complex systems, which allows us to design explainable deep models for uncovering the mechanistic role of how human brain (the most sophisticated complex system) maintains controllable functions while interacting with external stimulations. In the spirit of optimal control, we present a unified framework to design an explainable deep model that describes the dynamic behaviors of underlying neurobiological processes, allowing us to understand the latent control mechanism at a system level. We have uncovered the pathophysiological mechanism of Alzheimer's disease to the extent of controllability of disease progression, where the dissected system-level understanding enables higher prediction accuracy for disease progression and better explainability for disease etiology than conventional (black box) deep models.

A Radiation Induced Low-Grade Myofibroblastic Sarcoma in the Retropectoral Area After Breast Conserving Surgery: A Case Report.

Low-grade myofibroblastic sarcoma (LGMFS) is a rare type of sarcoma, and its manifestation as a radiotherapy (RT)-induced sarcoma following RT for breast cancer is even more unusual. To date, only one case of RT-induced mammary myofibroblastic sarcoma (MFS) has been reported. Here we present the case of a 49-year-old woman with LGMFS after undergoing breast-conserving surgery for invasive ductal carcinoma (IDC), and with a history of RT 16 years prior. Due to the rarity of this disease, previous studies have focused primarily on the pathological findings of MFS. In this report however, we present the clinical and radiological features of LGMFS in the retro pectoral area as a rare type of RT-induced sarcoma.

Predictors of Extraprostatic Extension in Patients with Prostate Cancer.

PURPOSE: To identify effective factors predicting extraprostatic extension (EPE) in patients with prostate cancer (PCa). METHODS: This retrospective cohort study recruited 898 consecutive patients with PCa treated with robot-assisted laparoscopic radical prostatectomy. The patients were divided into EPE and non-EPE groups based on the analysis of whole-mount histopathologic sections. Histopathological analysis (ISUP biopsy grade group) and magnetic resonance imaging (MRI) (PI-RADS v2.1 scores [1-5] and the Mehralivand EPE grade [0-3]) were used to assess the prediction of EPE. We also assessed the clinical usefulness of the prediction model based on decision-curve analysis. RESULTS: Of 800 included patients, 235 (29.3%) had EPE, and 565 patients (70.7%) did not (non-EPE). Multivariable logistic regression analysis showed that the biopsy ISUP grade, PI-RADS v2.1 score, and Mehralivand EPE grade were independent risk factors for EPE. In the regression assessment of the models, the best discrimination (area under the curve of 0.879) was obtained using the basic model (age, serum PSA, prostate volume at MRI, positive biopsy core, clinical T stage, and D'Amico risk group) and Mehralivand EPE grade 3. Decision-curve analysis showed that combining Mehralivand EPE grade 3 with the basic model resulted in superior net benefits for predicting EPE. CONCLUSION: Mehralivand EPE grades and PI-RADS v2.1 scores, in addition to basic clinical and demographic information, are potentially useful for predicting EPE in patients with PCa.

A Study on the Effectiveness of Deep Learning-Based Anomaly Detection Methods for Breast Ultrasonography.

In the medical field, it is delicate to anticipate good performance in using deep learning due to the lack of large-scale training data and class imbalance. In particular, ultrasound, which is a key breast cancer diagnosis method, is delicate to diagnose accurately as the quality and interpretation of images can vary depending on the operator's experience and proficiency. Therefore, computer-aided diagnosis technology can facilitate diagnosis by visualizing abnormal information such as tumors and masses in ultrasound images. In this study, we implemented deep learning-based anomaly detection methods for breast ultrasound images and validated their effectiveness in detecting abnormal regions. Herein, we specifically compared the sliced-Wasserstein autoencoder with two representative unsupervised learning models autoencoder and variational autoencoder. The anomalous region detection performance is estimated with the normal region labels. Our experimental results showed that the sliced-Wasserstein autoencoder model outperformed the anomaly detection performance of others. However, anomaly detection using the reconstruction-based approach may not be effective because of the occurrence of numerous false-positive values. In the following studies, reducing these false positives becomes an important challenge.

Oncologic necessity for the complete removal of residual microcalcifications after neoadjuvant chemotherapy for breast cancer.

The surgical range of breast cancer that shows pathologic complete response (pCR) without change in microcalcifications after neoadjuvant chemotherapy (NAC) is controversial. This study examined whole breast specimens to evaluate the necessity of mastectomy in those cases. The viability of cancer cells around the residual microcalcification was assessed using prospectively collected breast samples to confirm the presence or absence of cancer cells. A total of 144 patients with breast cancer and diffuse microcalcifications were classified into the reduced mass with no change in residual microcalcification (RESMIN, n = 49) and non-RESMIN (n = 95) groups. Five specimens were prospectively evaluated to assess the presence of viable cancer cells around the microcalcification. Tumor responses to NAC were significantly better with high pCR rates in the RESMIN group (p = 0.005 and p = 0.002). The incidence of human epidermal growth factor receptor 2-positive and triple-negative breast cancers was significantly high in the RESMIN group (p = 0.007). Although five (10.2%) patients had locoregional recurrence in the RESMIN group, no local recurrence in the breast was reported. Although pCR was highly estimated, residual cancers, including ductal carcinoma in situ, remained in 80% cases. Therefore, given the weak scientific evidence available currently, complete removal of residual microcalcifications should be considered for oncologic safety.

Comparison of Radiation Dose and Image Quality between the 2nd Generation and 3rd Generation Dual-Source Single-Energy and Dual-Source Dual-Energy CT of the Abdomen.

Purpose: We compared the radiation dose and image quality between the 2nd generation and the 3rd generation dual-source single-energy (DSSE) and dual-source dual-energy (DSDE) CT of the abdomen. Materials and Methods: We included patients undergoing follow-up abdominal CT after partial or radical nephrectomy in the first 10 months of 2019 (2nd generation DS CT) and the first 10 months of 2020 (3rd generation DS CT). We divided the 320 patients into 4 groups (A, 2nd generation DSSE CT; B, 2nd generation DSDE CT; C, 3rd generation DSSE CT; and D, 3rd generation DSDE CT) (n = 80 each) matched by sex and body mass index. Radiation dose and image quality (objective and subjective qualities) were compared between the groups. Results: The mean size-specific dose estimation of 3rd generation DSDE CT group was significantly lower than that of the 2nd generation DSSE CT (42.5%, p = 0.013) and 2nd generation DSDE CT (46.9%, p = 0.015) groups. Interobserver agreement was excellent for the overall image quality (intraclass correlation coefficient [ICC]: 0.8867) and image artifacts (ICC: 0.9423). Conclusion: Our results showed a considerable reduction in the radiation dose while maintaining high image quality with 3rd generation DSDE CT as compared to the 2nd generation DSDE CT and 2nd generation DSSE CT.

Ectopic Adrenal Adenoma in Renal Sinus: A Case Report.

The kidney is a rare site of ectopic adrenal adenoma. To the best of our knowledge, some cases of ectopic adrenal adenoma have been found in the kidney, but few of these cases explain the CT and MRI findings of the lesion. We reported a case of ectopic adrenal adenoma in the left renal sinus. A 47-year-old male patient underwent abdominal CT for routine health check-ups, which revealed a 1.2 cm enhancing mass in the left renal sinus. The MRI showed a signal drop of the mass in T1 weighted in- and opposed-phase, which indicates fat components. The mass was confirmed as an ectopic adrenal adenoma after surgery.

Prediction of Implant Size Based on Breast Volume Using Mammography with Fully Automated Measurements and Breast MRI.

BACKGROUND: Determination of implant size is crucial for patients with breast cancer undergoing one-stage breast reconstruction. The purpose of this study is to predict the implant size based on the breast volume measured by mammography (MG) with a fully automated method, and by breast magnetic resonance imaging (MRI) with a semi-automated method, in breast cancer patients with direct-to-implant reconstruction. PATIENTS AND METHODS: This retrospective study included 84 patients with breast cancer who underwent direct-to-implant reconstruction after nipple-sparing or skin-sparing mastectomy and preoperative MG and MRI between April 2015 and April 2019. Breast volume was measured using (a) MG with a fully automated commercial software and (b) MRI with an in-house semi-automated software program. Multivariable regression analyses including breast volume and patient weight (P < 0.05 in univariable analysis) were conducted to predict implant size. RESULTS: MG and MRI breast volume was highly correlated with both implant size (correlation coefficient 0.862 and 0.867, respectively; P values < 0.001) and specimen weight (correlation coefficient 0.802 and 0.852, respectively; P values < 0.001). Mean absolute difference between the MR breast volume and implant size was 160 cc, which was significantly higher than that between the MG breast volume and implant size of 118 cc (P < 0.001). On multivariable analyses, only breast volume measured by both MG and MRI was significantly associated with implant size in any implant type (all P values < 0.001). CONCLUSION: Breast volume measured by MG and MRI can be used to predict appropriate implant size in breast cancer patients undergoing direct-to-implant reconstruction in an efficient and objective manner.

Prediction of the Immune Phenotypes of Bladder Cancer Patients for Precision Oncology.

Bladder cancer (BC) is the most common urinary malignancy; however accurate diagnosis and prediction of recurrence after therapies remain elusive. This study aimed to develop a biosignature of immunotherapy-based responses using gene expression data. Publicly available BC datasets were collected, and machine learning (ML) approaches were applied to identify a novel biosignature to differentiate patient subgroups. Immune phenotyping of BC in the IMvigor210 dataset included three subtypes: inflamed, excluded, and desert immune. Immune phenotypes were analyzed with gene expressions using traditional but powerful classification methods such as random forests, Deep Neural Networks (DNN), Support Vector Machines (SVM) together with boosting and feature selection methods. Specifically, DNN yielded the highest area under the curve (AUC) with precision and recall (PR) curves and receiver operating characteristic (ROC) curves for each phenotype ([Formula: see text] and [Formula: see text], respectively) resulting in the identification of gene expression features useful for immune phenotype classification. Our results suggest significant potential to further develop and utilize machine learning algorithms for analysis of BC and its precaution. In conclusion, the findings from this study present a novel gene expression assay that can accurately discriminate BC patients from controls. Upon further validation in independent cohorts, this gene signature could be developed into a predictive test that can support clinical evaluation and patient care.

Unexpected hyperechoic lesions of the breast and their correlations with pathology: a pictorial essay.

Hyperechoic lesions of the breast encompass a wide range of conditions that are occasionally encountered during breast ultrasonography. Although typical hyperechoic lesions with a distinct fat component on imaging are well known, some hyperechoic lesions are diagnosed as unexpected pathology, making the radiology-pathology correlation difficult. Therefore, understanding the pathology of these lesions and how it correlates with imaging findings can help radiologists accurately diagnose and properly manage a range of related conditions. This article presents a pictorial review of unexpected hyperechoic benign and malignant breast lesions, with a focus on the pathological conditions that give rise to the hyperechoic pattern.

Performing Group Difference Testing on Graph Structured Data From GANs: Analysis and Applications in Neuroimaging.

Generative adversarial networks (GANs) have emerged as a powerful generative model in computer vision. Given their impressive abilities in generating highly realistic images, they are also being used in novel ways in applications in the life sciences. This raises an interesting question when GANs are used in scientific or biomedical studies. Consider the setting where we are restricted to only using the samples from a trained GAN for downstream group difference analysis (and do not have direct access to the real data). Will we obtain similar conclusions? In this work, we explore if "generated" data, i.e., sampled from such GANs can be used for performing statistical group difference tests in cases versus controls studies, common across many scientific disciplines. We provide a detailed analysis describing regimes where this may be feasible. We complement the technical results with an empirical study focused on the analysis of cortical thickness on brain mesh surfaces in an Alzheimer's disease dataset. To exploit the geometric nature of the data, we use simple ideas from spectral graph theory to show how adjustments to existing GANs can yield improvements. We also give a generalization error bound by extending recent results on Neural Network Distance. To our knowledge, our work offers the first analysis assessing whether the Null distribution in "healthy versus diseased subjects" type statistical testing using data generated from the GANs coincides with the one obtained from the same analysis with real data. The code is available at https://github.com/yyxiongzju/GLapGAN.

Clinical adaptation of synthetic MRI-based whole brain volume segmentation in children at 3 T: comparison with modified SPM segmentation methods.

PURPOSE: To validate the use of synthetic magnetic resonance imaging (SyMRI) volumetry by comparing with child-optimized SPM 12 volumetry in 3 T pediatric neuroimaging. METHODS: In total, 106 children aged 4.7-18.7 years who underwent both synthetic and 3D T1-weighted imaging and had no abnormal imaging/neurologic findings were included for the SyMRI vs. SPM T1-only segmentation (SPM T1). Forty of the 106 children who underwent an additional 3D T2-weighted imaging were included for the SyMRI vs. SPM multispectral segmentation (SPM multi). SPM segmentation using an age-appropriate atlas and inverse-transforming template-space intracranial mask was compared with SyMRI segmentation. Volume differences between SyMRI and SPM T1 were plotted against age to evaluate the influence of age on volume difference. RESULTS: Measurements derived from SyMRI and two SPM methods showed excellent agreements and strong correlations except for the CSF volume (CSFV) (intraclass correlation coefficients = 0.87-0.98; r = 0.78-0.96; relative volume difference other than CSFV = 6.8-18.5% [SyMRI vs. SPM T1] and 11.3-22.7% [SyMRI vs. SPM multi]). Dice coefficients of all brain tissues (except CSF) were in the range 0.78-0.91. The Bland-Altman plot and age-related volume difference change suggested that the volume differences between the two methods were influenced by the volume of each brain tissue and subject's age (p < 0.05). CONCLUSION: SyMRI and SPM segmentation results were consistent except for CSFV, which supports routine clinical use of SyMRI-based volumetry in pediatric neuroimaging. However, caution should be taken in the interpretation of the CSF segmentation results.

Weakly-supervised deep learning for ultrasound diagnosis of breast cancer.

Conventional deep learning (DL) algorithm requires full supervision of annotating the region of interest (ROI) that is laborious and often biased. We aimed to develop a weakly-supervised DL algorithm that diagnosis breast cancer at ultrasound without image annotation. Weakly-supervised DL algorithms were implemented with three networks (VGG16, ResNet34, and GoogLeNet) and trained using 1000 unannotated US images (500 benign and 500 malignant masses). Two sets of 200 images (100 benign and 100 malignant masses) were used for internal and external validation sets. For comparison with fully-supervised algorithms, ROI annotation was performed manually and automatically. Diagnostic performances were calculated as the area under the receiver operating characteristic curve (AUC). Using the class activation map, we determined how accurately the weakly-supervised DL algorithms localized the breast masses. For internal validation sets, the weakly-supervised DL algorithms achieved excellent diagnostic performances, with AUC values of 0.92-0.96, which were not statistically different (all Ps > 0.05) from those of fully-supervised DL algorithms with either manual or automated ROI annotation (AUC, 0.92-0.96). For external validation sets, the weakly-supervised DL algorithms achieved AUC values of 0.86-0.90, which were not statistically different (Ps > 0.05) or higher (P = 0.04, VGG16 with automated ROI annotation) from those of fully-supervised DL algorithms (AUC, 0.84-0.92). In internal and external validation sets, weakly-supervised algorithms could localize 100% of malignant masses, except for ResNet34 (98%). The weakly-supervised DL algorithms developed in the present study were feasible for US diagnosis of breast cancer with well-performing localization and differential diagnosis.

Peritumoral edema in breast cancer at preoperative MRI: an interpretative study with histopathological review toward understanding tumor microenvironment.

Peritumoral edema (PE) of breast cancer at T2-weighted MR images is considered a poor prognostic sign and may represent the microenvironment surrounding the tumor; however, its histopathological mechanism remains unclear. The purpose of the study was to identify and describe detailed histopathological characteristics associated with PE at preoperative breast MRI in breast cancer patients. This retrospective study included breast cancer patients who had undergone preoperative MRI and surgery between January 2011 and December 2012. Two radiologists determined the presence of PE in consensus based on the signal intensity surrounding the tumor at T2-weighted images. The following detailed histopathological characteristics were reviewed by two breast pathologists using four-tiered grades; lymphovascular invasion, vessel ectasia, stromal fibrosis, growth pattern, and tumor budding. Tumor necrosis and tumor infiltrating lymphocytes were assessed using a percent scale. Baseline clinicopathological characteristics, including age and histologic grade, were collected. The associations between detailed histopathologic characteristics and PE were examined using multivariable logistic regression with odds ratio (OR) calculation. A total of 136 women (median age, 49 ± 9 years) were assessed; among them 34 (25.0%) had PE. After adjustment of baseline clinicopathological characteristics that were significantly associated with PE (age, T stage, N stage, histologic grade, and subtype, all Ps < 0.05), lymphovascular invasion (P = 0.009), vessel ectasia (P = 0.021), stromal fibrosis (P = 0.024), growth pattern (P = 0.036), and tumor necrosis (P < 0.001) were also associated with PE. In comparison with patients without PE, patients with PE were more likely to have a higher degree of lymphovascular invasion (OR, 2.9), vessel ectasia (OR, 3.3), stromal fibrosis (OR, 2.5), lesser degree of infiltrative growth pattern (OR, 0.4), and higher portion of tumor necrosis (OR, 1.4). PE of breast cancer at MRI is associated with detailed histopathological characteristics of lymphovascular invasion, vessel ectasia, stromal fibrosis, growth pattern, and tumor necrosis, suggesting a relevance for tumor microenvironment.

Delayed unilateral hematoma after reconstructive and aesthetic breast surgery with implants in Asian patients: two case reports.

Hematomas represent one of the postoperative complications in patients undergoing reconstructive or aesthetic breast surgery with a silicone implant. Although there are few reports of intracapsular hematoma, those presenting late hematoma after reconstructive and aesthetic augmentation surgeries are rarer. This study reported two Asian patients with late hematoma after reconstruction and aesthetic breast surgery. A 54-year-old female patient underwent bilateral nipple-sparing mastectomy with immediate breast reconstruction using anatomically shaped textured implant for intraductal carcinoma in August 2019. Contralateral nipple-sparing mastectomy was performed for the BRCA gene mutation on the left breast, which was immediately reconstructed with an anatomically shaped textured implant. In a 1-year postoperative magnetic resonance imaging evaluation, an extracapsular hematoma was found on the right side, which was removed following the removal of both implants. Another case was a 63-year-old female patient who underwent augmentation of both breasts with smooth round implants and experienced right unilateral swelling and painless firmness about 30 years postoperatively. A preoperative magnetic resonance imaging evaluation showed both intracapsular and extracapsular ruptures on the right breast and a bulging implant herniation on the left breast. During the operation, hematoma, implants, and capsule were all removed. The excised capsule was sent for histological evaluation. Slightly dark colored blood was emptied before removing the semisolid-state intracapsular hematoma. In both cases, the patients responded well postoperatively and were discharged to their homes with no postsurgical complications, including seroma, or additional hematoma on the breasts. The etiology of late hematoma following breast augmentation or reconstruction has been poorly characterized. Further reports are needed to clearly establish the reasons for this increase in late hematoma formation.

Value of accurate diagnosis for metastatic supraclavicular lymph nodes in breast cancer: assessment with neck US, CT, and 18F-FDG PET/CT.

PURPOSE: Neck ultrasonography (US), computed tomography (CT), and 18F-fluorodeoxyglucose (FDG) positron emission tomography/computed tomography (PET/CT) are all known to be useful imaging modalities for detecting supraclavicular lymph node (SCN) metastasis in breast cancer. The authors compared the diagnostic values of neck US, CT, and PET/CT in the detection of SCN metastasis in breast cancer. METHODS: SCN metastases identified in neck US, CT, or PET/CT during follow-up visits of patients with breast cancer were pathologically confirmed with the use of US-guided fine-needle aspiration cytology. The clinicopathological factors of the patients were analyzed, and the statistical parameters including sensitivity, specificity, positive and negative predictive values, false-positive and false-negative rates, and accuracy of neck US, CT, and PET/CT were compared. RESULTS: Among 32 cases of suspicious SCNs, 24 were pathologically confirmed as metastasis of breast cancer. The sensitivity of US + CT was 91.7%, which was the same as that of PET/CT, while the sensitivity rates of US alone and CT alone were 87.5% and 83.3%, respectively. Accuracy was 99.8% in PET/CT alone and 98.1% in US + CT. The false-negative rate was 0.1% in US + PET/CT, while it was 0.2% in PET/CT and US + CT, 0.3% in US alone and 0.4% in CT alone. CONCLUSION: PET/CT can be the first choice for detecting SCN metastases in breast cancer. However, if PET/CT is unavailable for any reason, US + CT could be a good second option to avoid false-negative results.

Artificial intelligence in breast ultrasonography.

Although breast ultrasonography is the mainstay modality for differentiating between benign and malignant breast masses, it has intrinsic problems with false positives and substantial interobserver variability. Artificial intelligence (AI), particularly with deep learning models, is expected to improve workflow efficiency and serve as a second opinion. AI is highly useful for performing three main clinical tasks in breast ultrasonography: detection (localization/ segmentation), differential diagnosis (classification), and prognostication (prediction). This article provides a current overview of AI applications in breast ultrasonography, with a discussion of methodological considerations in the development of AI models and an up-to-date literature review of potential clinical applications.