Called Convolutional Neural Networks Research Articles

Abstract P2-11-11: Computer analysis of nuclear morphology with Multiple Instance Learning Predicts Overall Survival for Node Positive Breast Cancer Patients from SWOG S8814: A Blinded Validation Study

Abstract Background: Adjuvant chemotherapy is a common treatment for breast cancer patients with aggressive tumors, but imposes severe side effects. Although lymph node involvement is associated with higher likelihood of mortality, a subset of LN+ patients can survive without relapse, even without adjuvant chemotherapy. SWOG-8814 (S8814) was a randomized clinical trial of hormone receptor positive, LN+ post-menopausal women (pathologic stage T1-3N1-2) to compare patient outcome between endocrine therapy alone (tamoxifen), chemotherapy (cyclophosphamide, doxorubicin, and 5-fluorouracil) followed by 5 years of tamoxifen, and chemotherapy with concurrent tamoxifen. In this work, we evaluated the ability of a machine learning approach called multiple instance learning (MIL) to predict long-term overall (OS) and disease-free survival (DFS) in patients from the S8814 clinical trial via analysis of digitized H&E tissue slides. Methods: The training set (St, n=121) consisted of digitized H&E Whole Slide Images (WSIs) of ER+ LN+ patients from ECOG-2197 - a clinical trial to compare patient outcome under chemotherapy with doxorubicin/docetaxel vs. doxorubicin/cyclophosphamide. For each WSI, the nuclei of ten tumor patches (2000 × 2000 pixels) were automatically segmented by a pretrained machine learning algorithm called Convolutional Neural Networks (CNN). The results of the CNN’s segmentations were used to obtain ten corresponding feature vectors of 3963 features relating to nuclear morphology and spatial arrangement. Five top discriminative features (Haralick texture, Delaunay triangulation, orientation entropy, perimeter kurtosis, architecture) were identified by reliefF-MI feature filtering and forward selection on St. A Normalized Set Kernel Support Vector Machine classifier was trained on St to construct a MIL binary classifier (M+), which generates a continuous survival score to predict which patients would have >10-years of DFS and OS. The score cutoff thatmaximized F1 score on St was used to generate binary predictions of patient outcome. Blinded validation of the risk scores on S8814 was performed by SWOG. Results: M+ was significantly prognostic of OS in univariate analysis (HR=0.72, p=0.020, 95% CI= 0.55 – 0.95) on S8814. M+ stratifications remained statistically significant after multivariable analysis controlling for treatment, number of positive lymph nodes, and tumor size (HR = 0.71, p=0.017, 95% CI=0.54-0.94). M+ predictions did not achieve statistical significance for DFS in univariate analysis (HR=0.81, p=0.081, 95% CI 0.64-1.03) nor multivariable analysis (HR=0.81, p=0.080, 95% CI 0.63-1.03). M+ was not predictive of chemotherapy benefit for OS (p=0.88) and DFS (p=0.65). There was no correlation of the generated continuous score with Recurrence Score (r=-0.05), previously shown to be a predictive factor for chemotherapy benefit. Conclusion: Our findings demonstrate that nuclear morphology provides insight into overall survival in ER+ LN+ breast cancer patients. Future work will involve combining the MIL based morphologic predictor with other clinicopathologic factors to improve DFS risk stratification. Funding: Funding NIH/NCI grants U10CA180888, U10CA180819, U24CA196175; and in part by Genomic Health, INC. (now Exact Sciences Corp.) Citation Format: Daniel Shao, William E. Barlow, Haojia Li, Cheng Lu, Kathy S. Albain, James Rae, Daniel F. Hayes, Andrew K. Godwin, Alastair M. Thompson, Anant Madabhushi, Lajos Pusztai. Computer analysis of nuclear morphology with Multiple Instance Learning Predicts Overall Survival for Node Positive Breast Cancer Patients from SWOG S8814: A Blinded Validation Study [abstract]. In: Proceedings of the 2022 San Antonio Breast Cancer Symposium; 2022 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2023;83(5 Suppl):Abstract nr P2-11-11.

Sustainable Facial Authentication and Expression Prediction using Deep Learning Techniques

To make machines more human-like, computer vision is essential. Computer vision is a field that focuses on mimicking the capabilities of the human visual framework to capture high-level understanding from enhanced images or recordings. Such a computer vision application gives the machine the ability to recognize a person and detect their emotions in order to process them appropriately. Facial verification can be a process of identifying or validating an object through an image, video, or any audio-visual component of their face. It could be a biometric discriminant proof strategy that works directly and confronts measures of discriminating individuals through their facial design and biometric information. The innovation collects a unique set of biometric information from each individual regarding their face and facial expressions to identify an individual. Facial sensory recognition can be an innovation used to analyse estimates from a variety of sources, such as images and recordings. It makes a difference as machines get better the way humans do and treat them according to their emotions. We are using a deep learning computation called Convolutional Neural Networks (CNN) to prepare for this demonstration that determines the sentiment of certain input images. We need to pre-process the images to prepare and test the model. For pre-processing, we do image enhancement combining resizing, equalizing and converting the image to grayscale for the machine to achieve. This demonstration can have multiple applications in both surveillance and feedback systems.

Real-time sign language recognition system

One of the more natural ways of interaction between the machine and human is offered by Human gestures which are useful for sign language recognition (SLR). Sign language is the speaking tongue of a segment of people who are known as Deaf people. they are the most people who benefit from the (SLR) by the Human-Computer Interaction (HCI).Most natural people cannot understand these kinds of languages, so Deaf people fail to communicate with them without this helper which achieves by the (HCI). This study introduced a proposed system that offered SLR in real-time for some gestures from the American sign language (ASL), by using one of the most suitable deep learning-based architectures that were called Convolutional neural networks (CNN) and choosing the Squeezenet module After comparing it with a traditional machine learning system that relied on extracting [HOG] features. Squeezenet presented the best result in accuracy reached to (100%) in the off-time testing and about (97.5%) in the real-time with a competition time of about 3.3sec for capturing the image, predicting it, and converting it to a text and spoken sentence.The system achieved this result without making any preprocessing for the image, which gave it simplicity and low computing time.

Predicting and Classifying Breast Cancer Using Machine Learning

The proposed research work aims to develop a method to predict and classify breast cancer (BC) at an early stage. In this research, three models are developed, and their performance is compared against each other. The first model was built using one of the machine learning algorithms called support vector machine (SVM), the second model was built using a deep learning algorithm called convolutional neural networks (CNNs), and the third model combines CNNs with a transfer learning technique for delivering better results. The data set is provided by the BC Histopathological Image Classification (BreakHis). All models are trained on the training set with two main categories: benign tumor and malignant tumor. The malignant tumor category is divided into subsets of invasive carcinoma tumors and in situ carcinoma tumors. Furthermore, invasive carcinoma tumors are classified into grade 1, grade 2, or grade 3, where grade 3 is the highest and is more aggressive. The results show that the accuracies of biopsy image classification using SVM are 92%, the accuracy of CNN is 94%, and the accuracy of CNN using the transfer learning technique is 97%. The results of this research will be beneficial in the early diagnosis of BC and help doctors in making better decisions and medical interventions.

Genomic prediction of growth traits in scallops using convolutional neural networks

Deep learning neural networks applied to the genomic prediction of complex traits have been of great interest in recent years. Previous studies primarily used simulated phenotypes or/and genotypes in plants and animals. The properties of deep learning models used in genomic selection are not well characterized and not well validated with real datasets. In the present study, we evaluated the performance of a class of deep learning methods called convolutional neural networks (CNNs) in the genomic prediction of four quantitative traits (e.g., shell length, shell height, shell width, and total weight) in a Bay scallop (Argopecten irradians irradians) population. The results were compared with those obtained from two linear models, RR-GBLUP and Bayes B, and multilayer perceptron neural networks (MLPs). One-convolutional layer CNNs with an optimal structure, which was obtained by using AIC or BIC method, had roughly comparable prediction accuracies on the four quantitive traits in the scallop population. Overall, CNNs outperformed RR-GBLUP, Bayes B and MLPs on shell height, shell width and total weight, and performed slightly worse than only Bayes B on shell length. MLPs gave the least accurate predictions on average among the four types of models. Because MLPs had far more parameters to estimate than the two linear models, and their predictions were challenged by the overfitting problem. Genomic prediction accuracy varied with SNP panel size and training population size.The impact of varied marker densities and two GWAS-based scenarios for SNP selection on genomic prediction accuracy was investigated as well. The present results provide evidence that supports the use of convolutional neural networks for genomic prediction of complex traits in scallops, yet the optimal structures of CNNs remained to be exploited in future studies.

Prediction of Heart Disease Using Deep Convolutional Neural Networks

Heart diseases are currently a major cause of death in the world. This problem is severe in developing countries in Africa and Asia. A heart disease predicted at earlier stages not only helps the patients prevent it, but I can also help the medical practitioners learn the major causes of a heart attack and avoid it before its actual occurrence in patient. In this paper, we propose a method named CardioHelp which predicts the probability of the presence of cardiovascular disease in a patient by incorporating a deep learning algorithm called convolutional neural networks (CNN). The proposed method is concerned with temporal data modeling by utilizing CNN for HF prediction at its earliest stage. We prepared the heart disease dataset and compared the results with state-of-the-art methods and achieved good results. Experimental results show that the proposed method outperforms the existing methods in terms of performance evaluation metrics. The achieved accuracy of the proposed method is 97%.

Beach State Recognition Using Argus Imagery and Convolutional Neural Networks

Nearshore morphology is a key driver in wave breaking and the resulting nearshore circulation, recreational safety, and nutrient dispersion. Morphology persists within the nearshore in specific shapes that can be classified into equilibrium states. Equilibrium states convey qualitative information about bathymetry and relevant physical processes. While nearshore bathymetry is a challenge to collect, much information about the underlying bathymetry can be gained from remote sensing of the surfzone. This study presents a new method to automatically classify beach state from Argus daytimexposure imagery using a machine learning technique called convolutional neural networks (CNNs). The CNN processed imagery from two locations: Narrabeen, New South Wales, Australia and Duck, North Carolina, USA. Three different CNN models are examined, one trained at Narrabeen, one at Duck, and one trained at both locations. Each model was tested at the location where it was trained in a self-test, and the single-beach models were tested at the location where it was not trained in a transfer-test. For the self-tests, skill (as measured by the F-score) was comparable to expert agreement (CNN F-values at Duck = 0.80 and Narrabeen = 0.59). For the transfer-tests, the CNN model skill was reduced by 24–48%, suggesting the algorithm requires additional local data to improve transferability performance. Transferability tests showed that comparable F-scores (within 10%) to the self-trained cases can be achieved at both locations when at least 25% of the training data is from each site. This suggests that if applied to additional locations, a CNN model trained at one location may be skillful at new sites with limited new imagery data needed. Finally, a CNN visualization technique (Guided-Grad-CAM) confirmed that the CNN determined classifications using image regions (e.g., incised rip channels, terraces) that were consistent with beach state labelling rules.

CHLOROPHYLL-A CONCENTRATION RETRIEVAL USING CONVOLUTIONAL NEURAL NETWORKS IN LAGUNA LAKE, PHILIPPINES

Abstract. Two existing chlorophyll-a (chl-a) concentration retrieval procedures, which are analytical and empirical, are hindered by the complexity in radiative transfer equation (RTE) and in statistical analyses, respectively. Another promising model in this direction is the use of artificial neural networks (ANN). Mostly, a pixel-to-pixel with one-layer ANN model is used; where in fact that the satellite instrumental errors and man-made objects in water bodies might affect the retrieval and should be taken into account. In this study, the mask-based neural structure, called convolutional neural networks (CNN) model containing both the target and neighborhood pixels, is proposed to reduce the influence of the aforementioned premises. The proposed model is an end-to-end multiple-layer model which integrates band expansion, feature extraction, and chl-a estimation into the structure, leading to an optimal chl-a concentration retrieval. In addition to that, a two-stage training is also proposed to solve the problem of insufficient in-situ samples which happens in most of the time. In the first stage, the proposed model is trained by using the chl-a concentration derived from the water product, provided by satellite agency, and is refined with the in-situ samples in the second stage. Eight Sentinel-3 images from different acquisition time and coincide in-situ measurements over Laguna Lake waters of Philippines were utilized to conduct the model training and testing. Based on quantitative accuracy assessment, the proposed method outperformed the existing dual- and triple- bands combinations in chl-a concentration retrieval.

Human activity recognition by using convolutional neural network

<p>In recent years, many researchers have studied the HAR (Human Activity Recognition) system. HAR using smart home sensor is based on computing in smart environment, and intelligent surveillance system conducts intensive research on peripheral support life. The previous system studied in some of the activities is a fixed motion and the methodology is less accurate. In this paper, vision-based studies using thermal imaging cameras improve the accuracy of motion recognition in intelligent surveillance systems. We use one of the deep learning architectures widely used in image recognition systems called Convolutional Neural Networks (CNN). Therefore, we use CNN and thermal cameras to provide accuracy and many features through the proposed method.</p>

The Unreasonable Effectiveness of Convolutional Neural Networks in Population Genetic Inference.

Population-scale genomic data sets have given researchers incredible amounts of information from which to infer evolutionary histories. Concomitant with this flood of data, theoretical and methodological advances have sought to extract information from genomic sequences to infer demographic events such as population size changes and gene flow among closely related populations/species, construct recombination maps, and uncover loci underlying recent adaptation. To date, most methods make use of only one or a few summaries of the input sequences and therefore ignore potentially useful information encoded in the data. The most sophisticated of these approaches involve likelihood calculations, which require theoretical advances for each new problem, and often focus on a single aspect of the data (e.g., only allele frequency information) in the interest of mathematical and computational tractability. Directly interrogating the entirety of the input sequence data in a likelihood-free manner would thus offer a fruitful alternative. Here, we accomplish this by representing DNA sequence alignments as images and using a class of deep learning methods called convolutional neural networks (CNNs) to make population genetic inferences from these images. We apply CNNs to a number of evolutionary questions and find that they frequently match or exceed the accuracy of current methods. Importantly, we show that CNNs perform accurate evolutionary model selection and parameter estimation, even on problems that have not received detailed theoretical treatments. Thus, when applied to population genetic alignments, CNNs are capable of outperforming expert-derived statistical methods and offer a new path forward in cases where no likelihood approach exists.

Steganalysis via a convolutional neural network using large convolution filters for embedding process with same stego key: A deep learning approach for telemedicine

Summary Introduction Steganography, the art to hide information inside host media like pictures and movies, and steganalysis, its countermeasure attempting to detect the presence of an hidden information within an innocent-looking document, are frequently reported as promising information security techniques for telemedicine. For the past few years, in the race between image steganography and steganalysis, deep learning has emerged as a very promising alternative to steganalyzer approaches based on rich image models combined with ensemble classifiers. A key knowledge of image steganalyzer, which combines relevant image features and innovative classification procedures, can be deduced by a deep learning approach called convolutional neural networks (CNN). This kind of deep learning networks is so well-suited for classification tasks based on the detection of variations in 2D shapes that it is the state-of-the-art in many image recognition problems. Materials and methods We design a CNN-based steganalyzer for images obtained by applying steganography with a unique embedding key. The proposed CNN has a quite different shape compared to the ones resulting from the earlier works, and it is able to provide high detection accuracy for several steganographics tools when the same stego key is reused during the embedding process. The convolutional part of our proposal starts by a global filtering, using a single filter, followed by a second convolutional layer that produces a reduced set of high-level features (256 features for 512 × 512 pixels input images) thanks to the use of large filters. Results The proposed architecture embeds less convolutions, with much larger filters in the final convolutional layer, and is more general: it is able to deal with larger images and lower payloads. For the “same embedding key” scenario, our proposal outperforms all other steganalyzers, in particular the existing CNN-based ones, and defeats many state-of-the-art image steganography schemes. The information encoded by the final vector of features is so discriminating that the classifier part can be reduced to only two output neurons. We finally evaluated the detection ability of the CNN against two spatial domain steganographic schemes and a frequency domain one. More precisely, we designed a perfect steganalyzer for embedding payloads of 0.4 bit per pixel, and for all the steganographic tools investigated in this article (working either in spatial or in frequency domains). Rather interesting results have been obtained too, albeit to a lesser extent, for a payload value of 0.1 bpp. Discussion and conclusions The obtained results are very encouraging, and they outperform all the previous deep learning proposals for steganalysis. A first step in the design of a universal detector has been achieved too, as we are able to detect HUGO based hidden messages even when a WOW steganographier has been used during the training stage. These results allow us to propose to add fragile watermarks on media like pictures or pdf medical documents, to guarantee the authenticity of the material: any attempt of modification of the support will alter the watermark, proving by doing so the modification. Another application is to add personal and medical information inside medical images.

Artificial neural network for bubbles pattern recognition on the images

Two-phase bubble flows have been used in many technological and energy processes as processing oil, chemical and nuclear reactors. This explains large interest to experimental and numerical studies of such flows last several decades. Exploiting of optical diagnostics for analysis of the bubble flows allows researchers obtaining of instantaneous velocity fields and gaseous phase distribution with the high spatial resolution non-intrusively. Behavior of light rays exhibits an intricate manner when they cross interphase boundaries of gaseous bubbles hence the identification of the bubbles images is a complicated problem. This work presents a method of bubbles images identification based on a modern technology of deep learning called convolutional neural networks (CNN). Neural networks are able to determine overlapping, blurred, and non-spherical bubble images. They can increase accuracy of the bubble image recognition, reduce the number of outliers, lower data processing time, and significantly decrease the number of settings for the identification in comparison with standard recognition methods developed before. In addition, usage of GPUs speeds up the learning process of CNN owning to the modern adaptive subgradient optimization techniques.