•Convolutional neural networks are the most used architecture.•End-to-end methods are preferred over hybrid methods.•Recall is the most used metric for detection and localization tasks.•Intersection ...over Union is highly used in segmentation.•A common validation framework, including dataset and metrics, is missing.
Colorectal cancer has a great incidence rate worldwide, but its early detection significantly increases the survival rate. Colonoscopy is the gold standard procedure for diagnosis and removal of colorectal lesions with potential to evolve into cancer and computer-aided detection systems can help gastroenterologists to increase the adenoma detection rate, one of the main indicators for colonoscopy quality and predictor for colorectal cancer prevention. The recent success of deep learning approaches in computer vision has also reached this field and has boosted the number of proposed methods for polyp detection, localization and segmentation. Through a systematic search, 35 works have been retrieved. The current systematic review provides an analysis of these methods, stating advantages and disadvantages for the different categories used; comments seven publicly available datasets of colonoscopy images; analyses the metrics used for reporting and identifies future challenges and recommendations. Convolutional neural networks are the most used architecture together with an important presence of data augmentation strategies, mainly based on image transformations and the use of patches. End-to-end methods are preferred over hybrid methods, with a rising tendency. As for detection and localization tasks, the most used metric for reporting is the recall, while Intersection over Union is highly used in segmentation. One of the major concerns is the difficulty for a fair comparison and reproducibility of methods. Even despite the organization of challenges, there is still a need for a common validation framework based on a large, annotated and publicly available database, which also includes the most convenient metrics to report results. Finally, it is also important to highlight that efforts should be focused in the future on proving the clinical value of the deep learning based methods, by increasing the adenoma detection rate.
In the last few years, there has been a growing expectation created about the analysis of large amounts of data often available in organizations, which has been both scrutinized by the academic world ...and successfully exploited by industry. Nowadays, two of the most common terms heard in scientific circles are Big Data and Deep Learning. In this double review, we aim to shed some light on the current state of these different, yet somehow related branches of Data Science, in order to understand the current state and future evolution within the healthcare area. We start by giving a simple description of the technical elements of Big Data technologies, as well as an overview of the elements of Deep Learning techniques, according to their usual description in scientific literature. Then, we pay attention to the application fields that can be said to have delivered relevant real-world success stories, with emphasis on examples from large technology companies and financial institutions, among others. The academic effort that has been put into bringing these technologies to the healthcare sector are then summarized and analyzed from a twofold view as follows: first, the landscape of application examples is globally scrutinized according to the varying nature of medical data, including the data forms in electronic health recordings, medical time signals, and medical images; second, a specific application field is given special attention, in particular the electrocardiographic signal analysis, where a number of works have been published in the last two years. A set of toy application examples are provided with the publicly-available MIMIC dataset, aiming to help the beginners start with some principled, basic, and structured material and available code. Critical discussion is provided for current and forthcoming challenges on the use of both sets of techniques in our future healthcare.
The high incidence and mortality rate of colorectal cancer require new technologies to improve its early diagnosis. This study aims at extracting the medical needs related to the endoscopic ...technology and the colonoscopy procedure currently used for colorectal cancer diagnosis, essential for designing these demanded technologies.
Semi-structured interviews and an online survey were used.
Six endoscopists were interviewed and 103 were surveyed, obtaining the demanded needs that can be divided into: a) clinical needs, for better polyp detection and classification (especially flat polyps), location, size, margins and penetration depth; b) computer-aided diagnosis (CAD) system needs, for additional visual information supporting polyp characterization and diagnosis; and c) operational/physical needs, related to limitations of image quality, colon lighting, flexibility of the endoscope tip, and even poor bowel preparation.
This study shows some undertaken initiatives to meet the detected medical needs and challenges to be solved. The great potential of advanced optical technologies suggests their use for a better polyp detection and classification since they provide additional functional and structural information than the currently used image enhancement technologies. The inspection of remaining tissue of diminutive polyps (< 5 mm) should be addressed to reduce recurrence rates. Few progresses have been made in estimating the infiltration depth. Detection and classification methods should be combined into one CAD system, providing visual aids over polyps for detection and displaying a Kudo-based diagnosis suggestion to assist the endoscopist on real-time decision making. Estimated size and location of polyps should also be provided. Endoscopes with 360° vision are still a challenge not met by the mechanical and optical systems developed to improve the colon inspection. Patients and healthcare providers should be trained to improve the patient's bowel preparation.
Ovarian cancer (OC) is the second most common gynecological malignancy and the gynecological tumor with the worst prognosis. To try to improve this situation, Data Science technologies could be a ...useful tool to help clinicians to know more about the disease. In our case, we are interested in exploring OC data to discover relationships between clinical and genetic factors and the disease progression. For it, we propose an analysis framework for simple and univariate statistical descriptions of features of different types, based on bootstrap resampling. Foremost, we define the framework for metric, categorical, and dates variables and determine what are the advantages and disadvantages of using different bootstrap resampling strategies, based on their statistical basis. Then, we use it to perform a univariate analysis over an OC dataset that allows to explore how is the disease progression, having platinum-free interval as indicator, in relation to clinical and genetic features of different types. Also, it provides a first set of variables possibly relevant for survival prediction. Results obtained show that some features have led to individual differences between both platinum resistant (<6 months) and platinum sensitive(>6 months) groups. It can be concluded that this could be an indicator that the database could be discriminatory for the hypotheses studied, though it is convenient to make multivariate analyses to check how relationships among features are influenced.
Developments of richer integrative analysis methods for oncological studies are needed for efficiently leveraging the amount of clinical and genetic data available to provide the clinicians with ...better information. However, analyses of this nature often require mixing data of different types, which are not immediate to address jointly with classical methods. In this work, our aim is to find relationships between clinical and genetic features of different types (metric, categorical, and text) and the ovarian cancer (OC) disease progression. To this end, we first propose a univariate statistical method for text type applying bootstrap resampling to Bag of Words and Latent Dirichlet Allocation in order to include as features the free-text fields of the health recordings. Secondly, we extend bootstrap resampling for metric and categorical feature extraction with Principal Component Analysis (PCA) and Multiple Correspondence Analysis (MCA), respectively. We subsequently formulate a novel and integrative method for jointly considering metric, categorical, and text features. Results obtained in text analysis indicate individual differences in some words between two OC patients groups categorised according to their sensitivity to platinum drugs. These results indicate separability between both groups for text features. Also, regarding the multivariate analysis, clinical data results showed separability patterns for the three methods analysed according to the platinum-sensitivity degree. The use of these analytical tools in our OC cohort has allowed us to demonstrate their strengths by confirming the predictive and prognostic role of widely-known clinical and genetic variables (BRCA status, value of adjuvant therapy and optimal resection, or family history) and demonstrating significant associations in other variables whose role in OC development has been studied to a lesser extent (such as PMS1, GPC3, and SLX4 genes). These results highlight the value of implementing these approaches for the identification of novel biomarkers in the context of OC.
Educational disparities in Mathematics performance are a persistent challenge. This study aims to unravel the complex factors contributing to these disparities among students internationally, with a ...focus on the interpretability of the contributing factors.
Utilizing data from the Programme for International Student Assessment (PISA), we conducted rigorous preprocessing and variable selection to prepare for applying binary classification interpretability models. These models were trained using the Stratified K-Fold technique to ensure balanced representation and assessed using six key metrics.
By applying interpretability models such as Shapley Additive Explanations (SHAP) analysis, we identified critical factors impacting student performance, including reading accessibility, critical thinking skills, gender, and geographical location.
Our findings reveal significant disparities linked to resource availability, with students from lower socioeconomic backgrounds possessing fewer books and demonstrating lower performance in Mathematics. The geographical analysis highlighted regional educational disparities, with certain areas consistently underperforming in PISA assessments. Gender also emerged as a determinant, with females contributing differently to performance levels across the spectrum.
The study provides insights into the multifaceted determinants of student Mathematics performance and suggests potential avenues for future research to explore global interpretability models and further investigate the socioeconomic, cultural, and educational factors at play.
Although certain genetic alterations have been defined as predictive and prognostic biomarkers in the context of ovarian cancer (OC), data science methods represent alternative approaches to identify ...novel correlations and define relevant markers in these gynecological tumors. Considering this potential, our work focused both on clinical and genomic data information collected from patients with OC to identify relationships between clinical and genetic factors and disease progression-related variables. For this aim, we proposed two analyses: (1) a nonlinear exploration of an OC dataset using autoencoders, a type of neural network that can be used as a feature extraction tool to represent a dataset in 3-dimensional latent space, so that we could assess whether there are intrinsic or natural nonlinear separability patterns between disease progression groups (in our case, platinum-sensitive and platinum-resistant groups); and (2) the identification of relevant variable relationships by means of an adaptation of the informative variable identifier (IVI), a feature selection method that labels each input feature as informative or noisy with respect to the task at hand, identifies the relationships among features, and builds a ranking of features, allowing us to study which input features and relationships may be most informative for the OC disease progression classification to define new biomarkers involved in disease progression. Our interest has been in clinical and genetic factors and in the combination of clinical features and genetic profile. Results with autoencoders suggest a pattern of separability between disease progression groups in the clinical part and for the combination of genes and clinical features of the OC dataset, that is increased via supervised fine tuning. In the genetic part, this pattern of separability is not observed, but it is more defined when a supervised fine tuning is performed. Results of the IVI-mediated feature selection method show significance for relevant clinical variables (such as type of surgery and neoadjuvant chemotherapy), some mutation genes, and low-risk genetic features. These results highlight the efficacy of the considered approaches to better understand the clinical course of OC.
•Data science methods are suitable for identifying biomarkers in OC.•Feature selection methods show predictive roles of variables in an OC dataset.•Features extraction methods reveal some patterns of separability in an OC dataset.
The problem of Direction of Arrival (DOA) in Antenna Array processing has been widely studied. Nonuniform spacing between consecutive elements of linear arrays gives particular advantages but is also ...challenging. Precedent solutions aimed to deal with it using standard and complex Deep Learning architectures. Given the simplicity of the complex DOA basic equations, we hypothesized that the intrinsic structure of DOA problems should correspond to well-defined manifolds in low-dimensional latent spaces and that this should also hold in nonuniform spatially distributed arrays. We benchmarked a variety of manifold learning estimators for DOA problems, namely, Principal Component Analysis, t-Distributed Stochastic Neighbor Embedding (tSNE), shallow and deep autoencoders (AE), and Uniform Manifold Active Projection (UMAP). The generated latent spaces exhibited intrinsic geometrical structures in one-dimensional manifolds with four branches. On the other hand, DOA recovery built upon fine-tuning these latent spaces exhibited better performance for nonuniform arrays using shallow AE architectures. Our results pave the way for using low-dimensional latent variable spaces in various electromagnetic problems based on the simplicity of the physical equations describing these technologies.
Ovarian cancer (OC) is a deadly disease that affects a large number of women worldwide. Machine Learning (ML) models can help in the early detection of this disease, however, the use of these models ...may be limited by their lack of interpretability and the difficulty to evaluate their performance. In this work, five types of datasets were used, employing clinical features, different types of coding genomic features, and combining both. The use of interpretable ML (IML) models (one linear and one nonlinear model) provided us with better interpretability of the five feature sets. Following this study, nine binary classification models were compared, and the Accuracy, Recall, and Area Under the Curve were analyzed. The results showed that ML models employed the combination of clinical features and genomes with the coding of the position of genes in patients significantly improves the prediction. We demonstrated that the inclusion of different preprocessed patient data and especially through the information provided by IML models, can help clinicians to understand the disease better and make informed treatment decisions.
Hypertrophic Cardiomyopathy (HCM) consists of a thickening of the cardiac muscle, causing fatigue, changes in the cardioelectric system, arrhythmias, and even sudden deaths. Variants in gene MYBPC3 ...are a well-known cause of this illness. Our objective was to find variants in other genes that can cause this pathology. For that purpose, genetic data from a group of patients is analyzed using embedding methods, which allow a lower dimensional representation, which is very helpful for visualization, diagnosis, and personalized therapy. Our results, applying different methods -Principal Component Analyisis (PCA), t-distributed Stochastic Neighbor Embedding (t-SNE), Uniform Manifold Approximation and Projection (UMAP), Orthonormalized Partial Least Squares (OPLS) and Supervised Autoencoders- on genetic data showed a very good separability in the embedded space, allowing us to identify 10 variants that cause that separability. These results may be useful for identifying new HCM cases and implementing new Machine Learning models in those embedded spaces.