RESUMO
[This corrects the article DOI: 10.3389/fonc.2022.951973.].
RESUMO
During image editing, existing deep generative models tend to re-synthesize the entire output from scratch, including the unedited regions. This leads to a significant waste of computation, especially for minor editing operations. In this work, we present Spatially Sparse Inference (SSI), a general-purpose technique that selectively performs computation for edited regions and accelerates various generative models, including both conditional GANs and diffusion models. Our key observation is that users prone to gradually edit the input image. This motivates us to cache and reuse the feature maps of the original image. Given an edited image, we sparsely apply the convolutional filters to the edited regions while reusing the cached features for the unedited areas. Based on our algorithm, we further propose Sparse Incremental Generative Engine (SIGE) to convert the computation reduction to latency reduction on off-the-shelf hardware. With about 1%-area edits, SIGE accelerates DDPM by 3.0× on NVIDIA RTX 3090 and 4.6× on Apple M1 Pro GPU, Stable Diffusion by 7.2× on 3090, and GauGAN by 5.6× on 3090 and 5.2× on M1 Pro GPU. Compared to our conference paper, we enhance SIGE to accommodate attention layers and apply it to Stable Diffusion. Additionally, we offer support for Apple M1 Pro GPU and include more results to substantiate the efficacy of our method.
RESUMO
Background: The prognosis of patients with dilated cardiomyopathy (DCM) is poor and new indicators are urgently needed to predict lethal cardiac events. This study aimed to investigate the value of summed motion score (SMS) in predicting cardiac death of DCM patients using gated single photon emission computed tomography (SPECT) myocardial perfusion imaging (MPI). Methods and results: Eighty-one patients with DCM who underwent 99mTc-MIBI gated SPECT MPI were retrospectively enrolled and were divided into cardiac death and survivor groups. The functional parameters of left ventricle including SMS were measured using quantitative gated SPECT software. During the follow-up period of 44 (25, 54) months, 14 (17.28%) cardiac deaths were observed. Compared with the survivor group, SMS was significantly higher in the cardiac death group. Multivariate cox regression analysis showed that SMS was an independent predictor for cardiac death (HR 1.34, 95% CI 1.02-1.77, P = 0.034). SMS also provided incremental prognostic value over other variables in the multivariate model as determined by likelihood ratio global chi-squared test. In the Kaplan-Meier survival analysis, the event-free survival rate was significantly lower in the high-SMS (HSMS) group than the low-SMS (LSMS) (log-rank P < 0.001). Furthermore, the area under curve (AUC) of SMS was larger than that of LVEF at the 12th month of follow-up (0.85 vs. 0.80, P = 0.045). Conclusion: SMS is an independent predictor of cardiac death in DCM patients and provides incremental prognostic value. SMS might have higher predictive value than LVEF for early cardiac death.
RESUMO
OBJECTIVES: To investigate whether multimodal intratumour and peritumour ultrasound features correlate with specific breast cancer molecular subtypes. METHODS: From March to November 2021, a total of 85 patients with histologically proven breast cancer underwent B-mode, real-time elastography (RTE), colour Doppler flow imaging (CDFI) and contrast-enhanced ultrasound (CEUS). The time intensity curve (TIC) of CEUS was obtained, and the peak and time to peak (TTP) were analysed. Chi-square and binary logistic regression were used to analyse the connection between multimodal ultrasound imaging features and breast cancer molecular subtype. RESULTS: Among 85 breast cancers, the subtypes were as follows: luminal A (36 cases, 42.4%), luminal B (20 cases, 23.5%), human epidermal growth factor receptor-2 positive (HER2+) (16 cases, 18.8%), and triple negative breast cancer (TNBC) (13 cases, 15.3%). Binary logistic regression models showed that RTE (P < 0.001) and CDFI (P = 0.036) were associated with the luminal A cancer subtype (C-index: 0.741), RTE (P = 0.016) and the peak ratio between intratumour and corpus mamma (P = 0.036) were related to the luminal B cancer subtype (C-index: 0.788). The peak ratio between peritumour and intratumour (P = 0.039) was related to the HER2 + cancer subtype (C-index: 0.859), and CDFI (P = 0.002) was associated with the TNBC subtype (C-index: 0.847). CONCLUSIONS: Multimodal ultrasound features could be powerful predictors of specific breast cancer molecular subtypes. The intra- and peritumour CEUS features play assignable roles in separating luminal B and HER2 + breast cancer subtypes.
Assuntos
Neoplasias da Mama , Técnicas de Imagem por Elasticidade , Neoplasias de Mama Triplo Negativas , Humanos , Feminino , Neoplasias da Mama/patologia , Neoplasias de Mama Triplo Negativas/diagnóstico por imagem , Receptor ErbB-2/metabolismo , Mama/diagnóstico por imagem , Ultrassonografia , Biomarcadores Tumorais/metabolismoRESUMO
Objective: The aim of this study was to evaluate whether a predictive model based on a contrast enhanced ultrasound (CEUS)-based nomogram and clinical features (Clin) could differentiate Her-2-overexpressing breast cancers from other breast cancers. Methods: A total of 152 pathology-proven breast cancers including 55 Her-2-overexpressing cancers and 97 other cancers from two units that underwent preoperative CEUS examination, were included and divided into training (n = 102) and validation cohorts (n = 50). Multivariate regression analysis was utilized to identify independent indicators for developing predictive nomogram models. The area under the receiver operating characteristic (AUC) curve was also calculated to establish the diagnostic performance of different predictive models. The corresponding sensitivities and specificities of different models at the cutoff nomogram value were compared. Results: In the training cohort, 7 clinical features (menstruation, larger tumor size, higher CA153 level, BMI, diastolic pressure, heart rate and outer upper quarter (OUQ)) + enlargement in CEUS with P < 0.2 according to the univariate analysis were submitted to the multivariate analysis. By incorporating clinical information and enlargement on the CEUS pattern, independently significant indicators for Her-2-overexpression were used for further predictive modeling as follows: Model I, nomogram model based on clinical features (Clin); Model II, nomogram model combining enlargement (Clin + Enlargement); Model III, nomogram model based on typical clinical features combining enlargement (MC + BMI + diastolic pressure (DP) + outer upper quarter (OUQ) + Enlargement). Model II achieved an AUC value of 0.776 at nomogram cutoff score value of 190, which was higher than that of the other models in the training cohort without significant differences (all P>0.05). In the test cohort, the diagnostic efficiency of predictive model was poor (all AUC<0.6). In addition, the sensitivity and specificity were not significantly different between Models I and II (all P>0.05), in either the training or the test cohort. In addition, Clin exhibited an AUC similar to that of model III (P=0.12). Moreover, model III exhibited a higher sensitivity (70.0%) than the other models with similar AUC and specificity, only in the test cohort. Conclusion: The main finding of the study was that the predictive model based on a CEUS-based nomogram and clinical features could not differentiate Her-2-overexpressing breast cancers from other breast cancers.
RESUMO
The methodology for the synthesis of fluorene-based atropisomers was developed via the strategy of aromatic ring formation. By this strategy, an efficient benzannulation of indene-based diene with benzoylacetonitrile divergently promoted by DABCO and a chiral organocatalyst was established, and various atropisomeric fluorene-based skeletons were generated in good yields, which not only provide a new strategy for the construction of atropisomeric biaryls but also offer a new member to the atropisomeric family.
RESUMO
Background: Continuous contrast-enhanced ultrasound (CEUS) video is a challenging direction for radiomics research. We aimed to evaluate machine learning (ML) approaches with radiomics combined with the XGBoost model and a convolutional neural network (CNN) for discriminating between benign and malignant lesions in CEUS videos with a duration of more than 1 min. Methods: We gathered breast CEUS videos of 109 benign and 81 malignant tumors from two centers. Radiomics combined with the XGBoost model and a CNN was used to classify the breast lesions on the CEUS videos. The lesions were manually segmented by one radiologist. Radiomics combined with the XGBoost model was conducted with a variety of data sampling methods. The CNN used pretrained 3D residual network (ResNet) models with 18, 34, 50, and 101 layers. The machine interpretations were compared with prospective interpretations by two radiologists. Breast biopsies or pathological examinations were used as the reference standard. Areas under the receiver operating curves (AUCs) were used to compare the diagnostic performance of the models. Results: The CNN model achieved the best AUC of 0.84 on the test cohort with the 3D-ResNet-50 model. The radiomics model obtained AUCs between 0.65 and 0.75. Radiologists 1 and 2 had AUCs of 0.75 and 0.70, respectively. Conclusions: The 3D-ResNet-50 model was superior to the radiomics combined with the XGBoost model in classifying enhanced lesions as benign or malignant on CEUS videos. The CNN model was superior to the radiologists, and the radiomics model performance was close to the performance of the radiologists.
RESUMO
BACKGROUND: Diffuse uterine leiomyomatosis (DUL) is a benign uterine smooth muscle neoplasm with unknown etiology. Since DUL is rarely reported, knowledge regarding it is limited. The rate of early diagnosis is low, and DUL is often misdiagnosed as common multiple uterine leiomyomas before surgery. CASE SUMMARY: A 27-year-old patient with no sexual activity presented to the Emergency Department of our hospital complaining of heavy vaginal bleeding. She had a history of uterine fibroids and menorrhagia. Pelvic examination showed a regularly enlarged uterus, similar in size to that associated with a 4-mo pregnancy. Pelvic magnetic resonance imaging (MRI) revealed numerous multiple uterine fibroids, and a transabdominal myomectomy (TM) was performed. Intraoperative exploration revealed that the myometrium was full of myoma nodules of variable sizes. Over 50 leiomyomas were removed. The pathology report confirmed leiomyoma. The patient was discharged and received a gonadotropin-releasing hormone analog (3.75 mg) for 6 mo. Ten months after surgery, the patient presented to the hospital again for abnormal uterine bleeding. MRI showed an irregular mass with a diameter of 5.2 cm without sharp demarcation in the uterine cavity. Submucosal leiomyoma was considered first, and the patient underwent a hysteroscopic myomectomy plus hymen repair. Intraoperative exploration showed that there were several leiomyomatosis masses in the cavity. Postoperative pathological examination confirmed submucosal leiomyoma and necrotic and generative tissue. Although the menstrual cycle was still irregular, the patient did not have symptoms of menorrhagia for a period of 28 mo after the second surgery. CONCLUSION: Individuals with DUL are easily misdiagnosed due to the lack of specific manifestations of this disease. MRI is helpful for early identification and preoperative evaluation. There is currently no unified method of diagnosis. For women who want to preserve fertility, conservative surgery should be made an option. When TM is chosen, a modified new myomectomy should be considered to avoid the drawbacks of traditional TM.
RESUMO
BACKGROUND: A significant proportion of women with preeclampsia (PE) exhibit persistent postpartum hypertension (PHTN) at 3 months postpartum associated with cardiovascular morbidity. This study aimed to screen patients with PE to identify the high-risk population with persistent PHTN. METHODS: This retrospective cohort study enrolled 1,000 PE patients with complete parturient and postpartum blood pressure (BP) profiles at 3 months postpartum. The enrolled patients exhibited new-onset hypertension after 20 weeks of pregnancy, while those with PE superimposed upon chronic hypertension were excluded. Latent class cluster analysis (LCCA), a method of unsupervised learning in machine learning, was performed to ascertain maternal exposure clusters from eight variables and 35 subordinate risk factors. Logistic regression was applied to calculate odds ratios (OR) indicating the association between clusters and PHTN. RESULTS: The 1,000 participants were classified into three exposure clusters (subpopulations with similar characteristics) according to persistent PHTN development: high-risk cluster (31.2%), medium-risk cluster (36.8%), and low-risk cluster (32.0%). Among the 1,000 PE patients, a total of 134 (13.4%) were diagnosed with persistent PHTN, while the percentages of persistent PHTN were24.68%, 10.05%, and 6.25% in the high-, medium-, and low-risk clusters, respectively. Persistent PHTN in the high-risk cluster was nearly five times higher (OR, 4.915; 95% confidence interval (CI), 2.92-8.27) and three times (OR, 2.931; 95% CI, 1.91-4.49) than in the low- and medium-risk clusters, respectively. Persistent PHTN did not differ between the medium- and low-risk clusters. Subjects in the high-risk cluster were older and showed higher BP, poorer prenatal organ function, more adverse pregnancy events, and greater medication requirement than the other two groups. CONCLUSION: Patients with PE can be classified into high-, medium-, and low-risk clusters according to persistent PHTN severity; each cluster has cognizable clinical features. This study's findings stress the importance of controlling persistent PHTN to prevent future cardiovascular disease.
Assuntos
Hipertensão , Pré-Eclâmpsia , Análise por Conglomerados , Feminino , Humanos , Hipertensão/epidemiologia , Período Pós-Parto , Pré-Eclâmpsia/diagnóstico , Pré-Eclâmpsia/epidemiologia , Gravidez , Estudos Retrospectivos , Fatores de RiscoRESUMO
Conditional Generative Adversarial Networks (cGANs) have enabled controllable image synthesis for many vision and graphics applications. However, recent cGANs are 1-2 orders of magnitude more compute-intensive than modern recognition CNNs. For example, GauGAN consumes 281G MACs per image, compared to 0.44G MACs for MobileNet-v3, making it difficult for interactive deployment. In this work, we propose a general-purpose compression framework for reducing the inference time and model size of the generator in cGANs. Directly applying existing compression methods yields poor performance due to the difficulty of GAN training and the differences in generator architectures. We address these challenges in two ways. First, to stabilize GAN training, we transfer knowledge of multiple intermediate representations of the original model to its compressed model and unify unpaired and paired learning. Second, instead of reusing existing CNN designs, our method finds efficient architectures via neural architecture search. To accelerate the search process, we decouple the model training and search via weight sharing. Experiments demonstrate the effectiveness of our method across different supervision settings, network architectures, and learning methods. Without losing image quality, we reduce the computation of CycleGAN by 21×, Pix2pix by 12×, MUNIT by 29×, and GauGAN by 9×, paving the way for interactive image synthesis.
RESUMO
A simple and effective tandem reaction of diynones and allylic alcohols was developed to afford functionalized 3-allyl-4-pyrones in moderate to excellent yields. This protocol underwent a Michael additionâClaisen rearrangementâO-cyclization process, which exhibited broad substrate tolerance, high regioselectivity, and atom economy under a metal-free condition. Moreover, functional transformation of the products was also further studied.
Assuntos
Pironas , Catálise , Ciclização , Estrutura Molecular , FosfinasRESUMO
An efficient cascade reaction of 1-indanylidenemalononitrile with 3-benzylidenebenzofuran-2(3H)-one divergently promoted by DABCO or chiral organocatalyst was developed under mild reaction conditions, and various spiro-dihydrofluorene-benzofuranones were produced in gratifying results, respectively. It is worth noting that both the spiro and axially chiral products can be obtained by tuning the reaction conditions. The mechanism of the transformation was also studied by quantum chemical calculations.
RESUMO
A simple and effective annulation of ynediones and (iso)quinoline N-oxides was developed to afford various functionalized pyrrolo[2,1-a]isoquinolines and pyrrolo[1,2-a]quinolines in moderate to excellent yields. This protocol underwent a tandem [3 + 2] cycloaddition/ring-opening/N-nucleophilic addition, which exhibited high regioselectivity, broad substrate tolerance, and atom economy under catalyst-, additive-free, and air conditions. Moreover, indolizine was also successfully prepared using pyridine N-oxide.
RESUMO
[This corrects the article DOI: 10.1021/acsomega.0c01443.].
RESUMO
Deep neural networks excel at finding hierarchical representations that solve complex tasks over large datasets. How can we humans understand these learned representations? In this work, we present network dissection, an analytic framework to systematically identify the semantics of individual hidden units within image classification and image generation networks. First, we analyze a convolutional neural network (CNN) trained on scene classification and discover units that match a diverse set of object concepts. We find evidence that the network has learned many object classes that play crucial roles in classifying scene classes. Second, we use a similar analytic method to analyze a generative adversarial network (GAN) model trained to generate scenes. By analyzing changes made when small sets of units are activated or deactivated, we find that objects can be added and removed from the output scenes while adapting to the context. Finally, we apply our analytic framework to understanding adversarial attacks and to semantic image editing.
RESUMO
1H-Indazolo[1,2-b]phthalazine-5,10-dione IPDD with an approximate turbine-like spatial structure, primary assembled double-stranded helices at the first level, was predicted by quantum chemical calculations and confirmed by atomic force microscopy. The higher-dimensional hierarchical architectures including fibrils, helical fibers, spherical shells, and porous prismatic structures were observed in sequence by the scanning electron microscopy technique. The final porous prismatic structures sensitive to NH3 vapors have the potential to be applied in gas sensing and absorbing materials.
RESUMO
An effective electrophilic annulation reaction of azacyclic ynones was reported, divergently affording various functionalized 3-iodo-2H-quinolizin-2-ones and 1,3-diiodo-2H-quinolizin-2-ones in moderate to excellent yields with different iodide reagents. This reaction shows high regioselectivity and broad substrate scope under metal-free, room temperature conditions in air. In addition, the products with highly active C-I bonds have an opportunity for further functionalization.
RESUMO
An effective base-controlled divergent annulation reaction of 2-bromophenylacetonitriles and ynones has been developed. Various functionalized 5-cyanobenzoxepines and benzofuro[2,3-b]pyridines were obtained with a broad substrate scope and high regioselectivity in moderate to excellent yield. Of importance, an unexpected O-rearrangement reaction to access benzofuro[2,3-b]pyridines was observed using 1,8-diazabicyclo[5.4.0]undec-7-ene (DBU) as the base, and the possible mechanism was supported by 18O-labeled experiments. In addition, the gram-scale synthesis and further transformation of the product were studied.
RESUMO
Effective transition-metal-free formal [4+2] benzannulation for the preparation of 1-cyano-3-acylnaphthalenes from 2-(2-alkynylphenyl)acetonitriles and alkynones through sequential C-C bond coupling has been developed. This protocol is characterized by mild conditions, excellent functional group tolerance, complete regioselectivity, and atom economy. The plausible mechanism, gram-scale synthesis, and further transformations of the product were studied.
RESUMO
Humans can feel, weigh and grasp diverse objects, and simultaneously infer their material properties while applying the right amount of force-a challenging set of tasks for a modern robot1. Mechanoreceptor networks that provide sensory feedback and enable the dexterity of the human grasp2 remain difficult to replicate in robots. Whereas computer-vision-based robot grasping strategies3-5 have progressed substantially with the abundance of visual data and emerging machine-learning tools, there are as yet no equivalent sensing platforms and large-scale datasets with which to probe the use of the tactile information that humans rely on when grasping objects. Studying the mechanics of how humans grasp objects will complement vision-based robotic object handling. Importantly, the inability to record and analyse tactile signals currently limits our understanding of the role of tactile information in the human grasp itself-for example, how tactile maps are used to identify objects and infer their properties is unknown6. Here we use a scalable tactile glove and deep convolutional neural networks to show that sensors uniformly distributed over the hand can be used to identify individual objects, estimate their weight and explore the typical tactile patterns that emerge while grasping objects. The sensor array (548 sensors) is assembled on a knitted glove, and consists of a piezoresistive film connected by a network of conductive thread electrodes that are passively probed. Using a low-cost (about US$10) scalable tactile glove sensor array, we record a large-scale tactile dataset with 135,000 frames, each covering the full hand, while interacting with 26 different objects. This set of interactions with different objects reveals the key correspondences between different regions of a human hand while it is manipulating objects. Insights from the tactile signatures of the human grasp-through the lens of an artificial analogue of the natural mechanoreceptor network-can thus aid the future design of prosthetics7, robot grasping tools and human-robot interactions1,8-10.