•We present a multi-context ensemble of convolutional neural networks for the automated detection of small lesions in medical images.•The ensemble consists of different CNNs, aiming at learning ...different levels of image spatial context.•Multiple-depth CNNs are individually trained on image patches of different dimensions and then combined together.•We tested our method on microcalcification detection in mammograms and microaneurysm detection in fundus images.•We obtained statistically significantly better detection performance with respect to other approaches in the literature.
In this paper, we propose a novel method for the detection of small lesions in digital medical images. Our approach is based on a multi-context ensemble of convolutional neural networks (CNNs), aiming at learning different levels of image spatial context and improving detection performance. The main innovation behind the proposed method is the use of multiple-depth CNNs, individually trained on image patches of different dimensions and then combined together. In this way, the final ensemble is able to find and locate abnormalities on the images by exploiting both the local features and the surrounding context of a lesion. Experiments were focused on two well-known medical detection problems that have been recently faced with CNNs: microcalcification detection on full-field digital mammograms and microaneurysm detection on ocular fundus images. To this end, we used two publicly available datasets, INbreast and E-ophtha. Statistically significantly better detection performance were obtained by the proposed ensemble with respect to other approaches in the literature, demonstrating its effectiveness in the detection of small abnormalities.
•Water contaminant detection based on machine learning.•Comparison among different machine learning algorithms on a real problem.•End-to-end solution integrating sensors, data processing and ...classification.•A framework for selecting a suitable machine learning solution for edge computing.•SENSIPLUS, a low power micro-analytical sensing platform.
Water pollution caused by human activities poses a serious global threat to human health. Sensor technologies enabling water monitoring are an important tool that can help facing this problem. In this work, we propose an embedded IoT-ready system based on a proprietary sensor technology for the detection and recognition of six water contaminants. The system architecture is composed of two layers: (i) a sensing layer based on the SENSIPLUS chip, a proprietary Micro-Analytical Sensing Platform with six interdigitated electrodes metalized through different materials; and (ii) a data collection, communication, and classification layer with both hardware and software components. Being classification the most computationally and resource intensive operation, we evaluated nine machine learning solutions of different complexity and analyzed the trade-off between recognition accuracy, processing time, and memory usage to find a solution suitable to be implemented on an edge node. The highest average accuracy of 95.4% was achieved with K-nearest neighbor classification without constraints on processing time and memory usage, which confirms the potentiality of the system. When such constraints are taken into consideration, the best performance dropped to 86.4% offered by Multi Layer Perceptron.
Is it safe to perform controlled ovarian stimulation (COS) for fertility preservation before starting anticancer therapies or ART after treatments in young breast cancer patients?
Performing COS ...before, or ART following anticancer treatment in young women with breast cancer does not seem to be associated with detrimental prognostic effect in terms of breast cancer recurrence, mortality or event-free survival (EFS).
COS for oocyte/embryo cryopreservation before starting chemotherapy is standard of care for young women with breast cancer wishing to preserve fertility. However, some oncologists remain concerned on the safety of COS, particularly in patients with hormone-sensitive tumors, even when associated with aromatase inhibitors. Moreover, limited evidence exists on the safety of ART in breast cancer survivors for achieving pregnancy after the completion of anticancer treatments.
The present systematic review and meta-analysis was carried out by three blinded investigators using the keywords 'breast cancer' and 'fertility preservation'; keywords were combined with Boolean operators. Eligible studies were identified by a systematic literature search of Medline, Web of Science, Embase and Cochrane library with no language or date restriction up to 30 June 2021.
To be included in this meta-analysis, eligible studies had to be case-control or cohort studies comparing survival outcomes of women who underwent COS or ART before or after breast cancer treatments compared to breast cancer patients not exposed to these strategies. Survival outcomes of interest were cancer recurrence rate, relapse rate, overall survival and number of deaths. Adjusted relative risk (RR) and hazard ratio (HR) with 95% CI were extracted. When the number of events for each group were available but the above measures were not reported, HRs were estimated using the Watkins and Bennett method. We excluded case reports or case series with <10 patients and studies without a control group of breast cancer patients who did not pursue COS or ART. Quality of data and risk of bias were assessed using the Newcastle-Ottawa Assessment Scale.
A total of 1835 records were retrieved. After excluding ineligible publications, 15 studies were finally included in the present meta-analysis (n = 4643). Among them, 11 reported the outcomes of breast cancer patients who underwent COS for fertility preservation before starting chemotherapy, and 4 the safety of ART following anticancer treatment completion. Compared to women who did not receive fertility preservation at diagnosis (n = 2386), those who underwent COS (n = 1594) had reduced risk of recurrence (RR 0.58, 95% CI 0.46-0.73) and mortality (RR 0.54, 95% CI 0.38-0.76). No detrimental effect of COS on EFS was observed (HR 0.76, 95% CI 0.55-1.06). A similar trend of better outcomes in terms of EFS was observed in women with hormone-receptor-positive disease who underwent COS (HR 0.36, 95% CI 0.20-0.65). A reduced risk of recurrence was also observed in patients undergoing COS before neoadjuvant chemotherapy (RR 0.22, 95% CI 0.06-0.80). Compared to women not exposed to ART following completion of anticancer treatments (n = 540), those exposed to ART (n = 123) showed a tendency for better outcomes in terms of recurrence ratio (RR 0.34, 95% CI 0.17-0.70) and EFS (HR 0.43, 95% CI 0.17-1.11).
This meta-analysis is based on abstracted data and most of the studies included are retrospective cohort studies. Not all studies had matching criteria between the study population and the controls, and these criteria often differed between the studies. Moreover, rate of recurrence is reported as a punctual event and it is not possible to establish when recurrences occurred and whether follow-up, which was shorter than 5 years in some of the included studies, is adequate to capture late recurrences.
Our results demonstrate that performing COS at diagnosis or ART following treatment completion does not seem to be associated with detrimental prognostic effect in young women with breast cancer, including among patients with hormone receptor-positive disease and those receiving neoadjuvant chemotherapy.
Partially supported by the Associazione Italiana per la Ricerca sul Cancro (AIRC; grant number MFAG 2020 ID 24698) and the Italian Ministry of Health-5 × 1000 funds 2017 (no grant number). M.L. acted as consultant for Roche, Pfizer, Novartis, Lilly, AstraZeneca, MSD, Exact Sciences, Gilead, Seagen and received speaker honoraria from Roche, Pfizer, Novartis, Lilly, Ipsen, Takeda, Libbs, Knight, Sandoz outside the submitted work. F.S. acted as consultant for Novartis, MSD, Sun Pharma, Philogen and Pierre Fabre and received speaker honoraria from Roche, Novartis, BMS, MSD, Merck, Sun Pharma, Sanofi and Pierre Fabre outside the submitted work. I.D. has acted as a consultant for Roche, has received research grants from Roche and Ferring, has received reagents for academic clinical trial from Roche diagnostics, speaker's fees from Novartis, and support for congresses from Theramex and Ferring outside the submitted work. L.D.M. reported honoraria from Roche, Novartis, Eli Lilly, MSD, Pfizer, Ipsen, Novartis and had an advisory role for Roche, Eli Lilly, Novartis, MSD, Genomic Health, Pierre Fabre, Daiichi Sankyo, Seagen, AstraZeneca, Eisai outside the submitted work. The other authors declare no conflict of interest. The funding organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript and decision to submit the manuscript for publication.
N/A.
This paper is the editorial of the virtual special issue (VSI) “Artificial Intelligence for Distributed Smart Systems” (AI4DSS), of which the authors of this paper have been the guest editors. It ...aims to bring together the work of experts from the fields of artificial intelligence and that of smart sensing. Smart Sensing and, more generally, Smart Cyber Physical Systems are nowadays significantly impacting the everyday life of citizens and, in perspective, they will become pervasive in every aspect of human life from public health and well-being to home, infrastructures and environment management. Another important issue is related to the possibility of exploiting collaborative approaches through Distributed Architectures. In this kind of applications, smart sensors are spread into the environment of interest where some kind of “social intelligence” is generated. The papers included in this special issue allowed us to highlight the advances on this subject in three areas that are: Smart Cities, Smart Industries, Smart Healthcare.
•End-to-end writer recognition system on Avila Bible in three steps.•Row detection, row classification, page classification.•Row detection in transfer learning with MobileNetV2.•Transfer learning vs. ...from scratch for row classification.•Five models trained in fine tuning and from scratch with small labelled dataset.
This paper presents an end-to-end system to identify writers in medieval manuscripts. The proposed system consists in a three-step model for detection and classification of lines in the manuscript and page writer identification. The first two steps are based on deep neural networks trained with transfer learning techniques and specialized to solve the task in hand. The third stage is a weighted majority vote row-decision combiner that assigns to each page a writer. The main goal of this paper is to study the applicability of deep learning in this context when a relatively small training dataset is available. We tested our system with several state-of-the-art deep architectures on a digitized manuscript known as the Avila Bible, using only 9.6% of the total pages for training. Our approach proves to be very effective in identifying page writers, reaching a peak of 96.48% of accuracy and 96.56% of F1 score.
•We propose a multiclass-to-binary decomposition strategy founded on Coding Theory.•We use Low-Density Parity-Check codes, a very effective family of binary block codes.•Exploiting the algebraic ...properties of the code, we handle both coding and decoding.•Two decoding rules are proposed that provide many advantages over known strategies.•Several experiments are performed that show significant performance improvements.
A powerful strategy for the classification of multiple classes is to create a classifier ensemble that decomposes the polychotomy into several dichotomies. The central issue when designing a multiclass-to-binary decomposition scheme is the definition of both the coding matrix and the decoding algorithm. In this study, we propose a new classification system based on low-density parity-check codes, which is a very effective class of binary block codes. The main idea is to exploit the algebraic properties of the codes to generate the codewords for the coding matrix and to define two decoding approaches, which allow us to detect and recover possible errors or rejects produced by the dichotomizers. Experiments based on benchmark datasets demonstrated that the proposed approach provides a statistically significant improvement in terms of the classification performance compared with state-of-the-art decomposition strategies.
•Object detection is frequently a complex, severely unbalanced classification problem.•A cascade of node classifiers allows us to efficiently handle the complexity.•In our proposal, each node ...classifier is trained with a ranking-based algorithm.•Ranking effectively faces the imbalance between object and non-object patches.•Our method is effective if compared to other learning strategies for skewed classes.
To distinguish objects from non-objects in images under computational constraints, a suitable solution is to employ a cascade detector that consists of a sequence of node classifiers with increasing discriminative power. However, among the millions of image patches generated from an input image, only very few contain the searched object. When trained on these highly unbalanced data sets, the node classifiers tend to have poor performance on the minority class. Thus, we propose a learning strategy aimed at maximizing the node classifiers ranking capability rather than their accuracy. We also provide an efficient implementation yielding the same time complexity of the original Viola–Jones cascade training. Experimental results on highly unbalanced real problems show that our approach is both efficient and effective when compared to other node training strategies for skewed classes.
ECOC is a widely used and successful technique, which implements a multi-class classification system by decomposing the original problem into several two-class problems. In this paper, we study the ...possibility to provide ECOC systems with a tailored reject option carried out through different schemes that can be grouped under two different categories: an external and an internal approach. The first one is based on the reliability of the entire system output and does not require any change in its structure. The second scheme, instead, estimates the reliability of the internal dichotomizers and implies a slight modification in the decoding stage. Experimental results on popular benchmark data sets are reported to show the behavior of the different schemes.
► ECOC implements a multi-class classification system. ► ECOC works by decomposing the original problem into several two-class problems. ► An analysis of the behavior of ECOC systems with a reject option. ► Hamming distance or classifier output to design the reject rule. ► Error reject curve to evaluate ECOC systems.
The majority of the available classification systems focus on the minimization of the classification error rate. This is not always a suitable metric specially when dealing with two-class problems ...with skewed classes and cost distributions. In this case, an effective criterion to measure the quality of a decision rule is the area under the Receiver Operating Characteristic curve (AUC) that is also useful to measure the ranking quality of a classifier as required in many real applications. In this paper we propose a nonparametric linear classifier based on the maximization of AUC. The approach lies on the analysis of the Wilcoxon–Mann–Whitney statistic of each single feature and on an iterative pairwise coupling of the features for the optimization of the ranking of the combined feature. By the pairwise feature evaluation the proposed procedure is essentially different from other classifiers using AUC as a criterion. Experiments performed on synthetic and real data sets and comparisons with previous approaches confirm the effectiveness of the proposed method.