Search

Computational pathology targets the automatic analysis of Whole Slide Images (WSI). WSIs are high-resolution digitized histopathology images, stained with chemical reagents to highlight specific tissue structures and scanned via whole slide scanners. The application of different parameters during WSI acquisition may lead to stain color heterogeneity, especially considering samples collected from several medical centers. Dealing with stain color heterogeneity often limits the robustness of methods developed to analyze WSIs, in particular Convolutional Neural Networks (CNN), the state-of-the-art algorithm for most computational pathology tasks. Stain color heterogeneity is still an unsolved problem, although several methods have been developed to alleviate it, such as Hue-Saturation-Contrast (HSC) color augmentation and stain augmentation methods. The goal of this paper is to present Data-Driven Color Augmentation (DDCA), a method to improve the efficiency of color augmentation methods by increasing the reliability of the samples used for training computational pathology models. During CNN training, a database including over 2 million H&E color variations collected from private and public datasets is used as a reference to discard augmented data with color distributions that do not correspond to realistic data. DDCA is applied to HSC color augmentation, stain augmentation and H&E-adversarial networks in colon and prostate cancer classification tasks. DDCA is then compared with 11 state-of-the-art baseline methods to handle color heterogeneity, showing that it can substantially improve classification performance on unseen data including heterogeneous color variations.

Many “Industry 4.0” applications rely on data-driven methodologies such as Machine Learning and Deep Learning to enable automatic tasks and implement smart factories. Among these applications, the automatic quality control of manufacturing materials is of utmost importance to achieve precision and standardization in production. In this regard, most of the related literature focused on combining Deep Learning with Nondestructive Testing techniques, such as Infrared Thermography, requiring dedicated settings to detect and classify defects in composite materials. Instead, the research described in this paper aims at understanding whether deep neural networks and transfer learning can be applied to plain images to classify surface defects in carbon look components made with Carbon Fiber Reinforced Polymers used in the automotive sector. To this end, we collected a database of images from a real case study, with 400 images to test binary classification (defect vs. no defect) and 1500 for the multiclass classification (components with no defect vs. recoverable vs. non-recoverable). We developed and tested ten deep neural networks as classifiers, comparing ten different pre-trained CNNs as feature extractors. Specifically, we evaluated VGG16, VGG19, ResNet50 version 2, ResNet101 version 2, ResNet152 version 2, Inception version 3, MobileNet version 2, NASNetMobile, DenseNet121, and Xception, all pre-trainined with ImageNet, combined with fully connected layers to act as classifiers. The best classifier, i.e., the network based on DenseNet121, achieved a 97% accuracy in classifying components with no defects, recoverable components, and non-recoverable components, demonstrating the viability of the proposed methodology to classify surface defects from images taken with a smartphone in varying conditions, without the need for dedicated settings. The collected images and the source code of the experiments are available in two public, open-access repositories, making the presented research fully reproducible.

The brain is the organ most studied using Magnetic Resonance (MR). The emergence of 7T scanners has increased MR imaging resolution to a sub-millimeter level. However, there is a lack of automatic segmentation techniques for 7T MR volumes. This research aims to develop a novel deep learning-based algorithm for on-cloud brain extraction and multi-structure segmentation from unenhanced 7T MR volumes. To this aim, a double-stage 3D U-Net was implemented in a cloud service, directing its first stage to the automatic extraction of the brain and its second stage to the automatic segmentation of the grey matter, basal ganglia, white matter, ventricles, cerebellum, and brain stem. The training was performed on the 90% (the 10% of which served for validation) and the test on the 10% of the Glasgow database. A mean test Dice Similarity Coefficient (DSC) of 96.33% was achieved for the brain class. Mean test DSCs of 90.24%, 87.55%, 93.82%, 85.77%, 91.53%, and 89.95% were achieved for the brain structure classes, respectively. Therefore, the proposed double-stage 3D U-Net is effective in brain extraction and multi-structure segmentation from 7T MR volumes without any preprocessing and training data augmentation strategy while ensuring its machine-independent reproducibility.

Knowledge about the anatomical structures of the left heart, specifically the atrium (LA) and ventricle (i.e., endocardium—Vendo—and epicardium—LVepi) is essential for the evaluation of cardiac functionality. Manual segmentation of cardiac structures from echocardiography is the baseline reference, but results are user-dependent and time-consuming. With the aim of supporting clinical practice, this paper presents a new deep-learning (DL)-based tool for segmenting anatomical structures of the left heart from echocardiographic images. Specifically, it was designed as a combination of two convolutional neural networks, the YOLOv7 algorithm and a U-Net, and it aims to automatically segment an echocardiographic image into LVendo, LVepi and LA. The DL-based tool was trained and tested on the Cardiac Acquisitions for Multi-Structure Ultrasound Segmentation (CAMUS) dataset of the University Hospital of St. Etienne, which consists of echocardiographic images from 450 patients. For each patient, apical two- and four-chamber views at end-systole and end-diastole were acquired and annotated by clinicians. Globally, our DL-based tool was able to segment LVendo, LVepi and LA, providing Dice similarity coefficients equal to 92.63%, 85.59%, and 87.57%, respectively. In conclusion, the presented DL-based tool proved to be reliable in automatically segmenting the anatomical structures of the left heart and supporting the cardiological clinical practice.

Following the growing availability of video surveillance cameras and the need for techniques to automatically identify events in video footages, there is an increasing interest towards automatic violence detection in videos. Deep learning-based architectures, such as 3D Convolutional Neural Networks, demonstrated their capability of extracting spatio-temporal features from videos, being effective in violence detection. However, friendly behaviours or fast moves such as hugs, small hits, claps, high fives, etc., can still cause false positives, interpreting a harmless action as violent. To this end, we present three deep learning-based models for violence detection and test them on the AIRTLab dataset, a novel dataset designed to check the robustness of algorithms against false positives. The objective is twofold: on one hand, we compute accuracy metrics on the three proposed models (two are based on transfer learning and one is trained from scratch), building a baseline of metrics for the AIRTLab dataset; on the other hand, we validate the capability of the proposed dataset of challenging the robustness to false positives. The results of the proposed models are in line with the scientific literature, in terms of accuracy, with transfer learning-based networks exhibiting better generalization capabilities than the trained from scratch network. Moreover, the tests highlighted that most of the classification errors concern the identification of non-violent clips, validating the design of the proposed dataset. Finally, to demonstrate the significance of the proposed models, the paper presents a comparison with the related literature, as well as with models based on well-established pre-trained 2D Convolutional Neural Networks (2D CNNs). Such comparison highlights that 3D models get better accuracy performance than time distributed 2D CNNs (merged with a recurrent module) in processing the spatio-temporal features of video clips. The source code of the experiments and the AIRTLab dataset are available in public repositories.

Although face recognition technology is currently integrated into industrial applications, it has open challenges, such as verification and identification from arbitrary poses. Specifically, there is a lack of research about face recognition in surveillance videos using, as reference images, mugshots taken from multiple Points of View (POVs) in addition to the frontal picture and the right profile traditionally collected by national police forces. To start filling this gap and tackling the scarcity of databases devoted to the study of this problem, we present the Face Recognition from Mugshots Database (FRMDB). It includes 28 mugshots and 5 surveillance videos taken from different angles for 39 distinct subjects. The FRMDB is intended to analyze the impact of using mugshots taken from multiple points of view on face recognition on the frames of the surveillance videos. To validate the FRMDB and provide a first benchmark on it, we ran accuracy tests using two CNNs, namely VGG16 and ResNet50, pre-trained on the VGGFace and VGGFace2 datasets for the extraction of face image features. We compared the results to those obtained from a dataset from the related literature, the Surveillance Cameras Face Database (SCFace). In addition to showing the features of the proposed database, the results highlight that the subset of mugshots composed of the frontal picture and the right profile scores the lowest accuracy result among those tested. Therefore, additional research is suggested to understand the ideal number of mugshots for face recognition on frames from surveillance videos.

Fetal heart rate (FHR) monitoring can serve as a benchmark to identify high-risk fetuses. Fetal phonocardiogram (FPCG) is the recording of the fetal heart sounds (FHS) by means of a small acoustic sensor placed on maternal abdomen. Being heavily contaminated by noise, FPCG processing implies mandatory filtering to make FPCG clinically usable. Aim of the present study was to perform a comparative analysis of filters based on Wavelet transform (WT) characterized by different combinations of mothers Wavelet and thresholding settings. By combining three mothers Wavelet (4th-order Coiflet, 4th-order Daubechies and 8th-order Symlet), two thresholding rules (Soft and Hard) and three thresholding algorithms (Universal, Rigorous and Minimax), 18 different WT-based filters were obtained and applied to 37 simulated and 119 experimental FPCG data (PhysioNet/PhysioBank). Filters performance was evaluated in terms of reliability in FHR estimation from filtered FPCG and noise reduction quantified by the signal-to-noise ratio (SNR). The filter obtained by combining the 4th-order Coiflet mother Wavelet with the Soft thresholding rule and the Universal thresholding algorithm was found to be optimal in both simulated and experimental FPCG data, since able to maintain FHR with respect to reference (138.7[137.7; 140.8] bpm vs. 140.2[139.7; 140.7] bpm, P > 0.05, in simulated FPCG data; 139.6[113.4; 144.2] bpm vs. 140.5[135.2; 146.3] bpm, P > 0.05, in experimental FPCG data) while strongly incrementing SNR (25.9[20.4; 31.3] dB vs. 0.7[-0.2; 2.9] dB, P < 10-14, in simulated FPCG data; 22.9[20.1; 25.7] dB vs. 15.6[13.8; 16.7] dB, P < 10-37, in experimental FPCG data). In conclusion, the WT-based filter obtained combining the 4th-order Coiflet mother Wavelet with the thresholding settings constituted by the Soft rule and the Universal algorithm provides the optimal WT-based filter for FPCG filtering according to evaluation criteria based on both noise and clinical features.

The developments in the internet of things (IoT), artificial intelligence (AI), and cyber-physical systems (CPS) are paving the way to the implementation of smart factories in what is commonly recognized as the fourth industrial revolution. In the manufacturing sector, these technological advancements are making Industry 4.0 a reality, with data-driven methodologies based on machine learning (ML) that are capable of extracting knowledge from the data collected by sensors placed on production machines. This is particularly relevant in plastic injection molding, with the objective of monitoring the quality of molded products from the parameters of the production process. In this regard, the main contribution of this paper is the systematic comparison of ML techniques to predict the quality classes of plastic molded products, using real data collected during the production process. Specifically, we compare six different classifiers on the data coming from the production of plastic road lenses. To run the comparison, we collected a dataset composed of the process parameters of 1451 road lenses. On such samples, we tested a multi-class classification, providing a statistical analysis of the results as well as of the importance of the input features. Among the tested classifiers, the ensembles of decision trees, i.e., random forest and gradient-boosted trees (GBT), achieved 95% accuracy in predicting the quality classes of molded products, showing the viability of the use of ML-based techniques for this purpose. The collected dataset and the source code of the experiments are available in a public, open-access repository, making the presented research fully reproducible.

Surface Electromyography (EMG) and Inertial Measurement Unit (IMU) sensors are gaining the attention of the research community as data sources for automatic sign language recognition. In this regard, we provide a dataset of EMG and IMU data collected using the Myo Gesture Control Armband, during the execution of the 26 gestures of the Italian Sign Language alphabet. For each gesture, 30 data acquisitions were executed, composing a total of 780 samples included in the dataset. The gestures were performed by the same subject (male, 24 years old) in lab settings. EMG and IMU data were collected in a 2 seconds time window, at a sampling frequency of 200 Hz.

The automatic detection of violence and crimes in videos is gaining attention, specifically as a tool to unburden security officers and authorities from the need to watch hours of footages to identify event lasting few seconds. So far, most of the available datasets was composed of few clips, in low resolution, often built on too specific cases (e.g. hockey fight). While high resolution datasets are emerging, there is still the need of datasets to test the robustness of violence detection techniques to false positives, due to behaviours which might resemble violent actions. To this end, we propose a dataset composed of 350 clips (MP4 video files, 1920 × 1080 pixels, 30 fps), labelled as non-violent (120 clips) when representing non-violent behaviours, and violent (230 clips) when representing violent behaviours. In particular, the non-violent clips include behaviours (hugs, claps, exulting, etc.) that can cause false positives in the violence detection task, due to fast movements and the similarity with violent behaviours. The clips were performed by non-professional actors, varying from 2 to 4 per clip.

The COVID-19 pandemic exploded at the beginning of 2020, with over four million cases in five months, overwhelming the healthcare sector. Several national governments decided to adopt containment measures, such as lockdowns, social distancing, and quarantine. Among these measures, contact tracing can contribute in bringing under control the outbreak, as quickly identifying contacts to isolate suspected cases can limit the number of infected people. In this paper we present BubbleBox, a system relying on a dedicated device to perform contact tracing. BubbleBox integrates Internet of Things and software technologies into different components to achieve its goal—providing a tool to quickly react to further outbreaks, by allowing health operators to rapidly reach and test possible infected people. This paper describes the BubbleBox architecture, presents its prototype implementation, and discusses its pros and cons, also dealing with privacy concerns.