Interpreting Convolutional Neural Networks (CNNs) is critical for safety-sensitive applications such as healthcare and autonomous systems. Popular visual explanation methods like Grad-CAM use a single convolutional layer, potentially missing multi-scale cues and producing unstable saliency maps. We introduce Winsor-CAM, a single-pass gradient-based method that aggregates Grad-CAM maps from all convolutional layers and applies percentile-based Winsorization to attenuate outlier contributions. A user-controllable percentile parameter $p$p enables semantic-level tuning from low-level textures to high-level object patterns. We evaluate Winsor-CAM on six CNN architectures using PASCAL VOC 2012 and PolypGen, comparing localization (IoU, center-of-mass distance) and fidelity (insertion/deletion AUC) against seven baselines including Grad-CAM, Grad-CAM++, LayerCAM, ScoreCAM, AblationCAM, ShapleyCAM, and FullGrad. On DenseNet121 with a subset of Pascal VOC 2012, Winsor-CAM achieves 46.8% IoU and 0.059 CoM distance versus 39.0% and 0.074 for Grad-CAM, with improved insertion AUC (0.656vs. 0.623) and deletion AUC (0.197vs. 0.242). Notably, even the worst-performing fixed $p$p-value configuration outperforms FullGrad across all metrics. An ablation study confirms that incorporating earlier layers improves localization. Similar evaluation on PolypGen polyp segmentation further validates Winsor-CAM's effectiveness in medical imaging contexts. Winsor-CAM provides an efficient, robust, and human-tunable explanation tool for expert-in-the-loop analysis.

The growing reliance on deep learning models in safety-critical domains such as healthcare and autonomous navigation underscores the need for defenses that are both robust to adversarial perturbations and transparent in their decision-making. In this paper, we identify a connection between interpretability and robustness that can be directly leveraged during training. Specifically, we observe that spurious, unstable, or semantically irrelevant features identified through Local Interpretable Model-Agnostic Explanations (LIME) contribute disproportionately to adversarial vulnerability. Building on this insight, we introduce an attribution-guided refinement framework that transforms LIME from a passive diagnostic into an active training signal. Our method systematically suppresses spurious features using feature masking, sensitivity-aware regularization, and adversarial augmentation in a closed-loop refinement pipeline. This approach does not require additional datasets or model architectures and integrates seamlessly into standard adversarial training. Theoretically, we derive an attribution-aware lower bound on adversarial distortion that formalizes the link between explanation alignment and robustness. Empirical evaluations on CIFAR-10, CIFAR-10-C, and CIFAR-100 demonstrate substantial improvements in adversarial robustness and out-of-distribution generalization.

This work introduces SCL-GAN (Spatially-Correlative Lightweight GAN), a novel architecture for facial image reconstruction using thermal face images, designed for efficient execution on edge devices such as the NVIDIA Jetson board. The proposed architecture leverages spatial feature correlations across thermal-visible modalities while maintaining a low parameter count and FLOPs. Experimental results show that SCL-GAN achieves a 68.57% reduction in computational cost (GMac) and a 71.71% reduction in trainable parameters, compared to baseline models. Moreover, we observe consistent improvements in image quality metrics, including a 5.05% increase in SSIM, 4.49% reduction in VGG-FaceLoss, and a 27.83% reduction in FID on the WHU-IIP dataset. On the CVBL-CHILD dataset, SCL-GAN demonstrates an 11.70% SSIM improvement, 18.21% VGG-FaceLoss reduction, and a 47.88% drop in FID. The code is available at: https://github.com/GANGREEK/SCL-GAN.git.

Adversarial examples reveal critical vulnerabilities in deep neural networks by exploiting their sensitivity to imperceptible input perturbations. While adversarial training remains the predominant defense strategy, it often incurs significant computational cost and may compromise clean-data accuracy. In this work, we investigate an architectural approach to adversarial robustness by embedding group-equivariant convolutions—specifically, rotation- and scale-equivariant layers—into standard convolutional neural networks (CNNs). These layers encode symmetry priors that align model behavior with structured transformations in the input space, promoting smoother decision boundaries and greater resilience to adversarial attacks. We propose and evaluate two symmetry-aware architectures: a parallel design that processes standard and equivariant features independently before fusion, and a cascaded design that applies equivariant operations sequentially. Theoretically, we demonstrate that such models reduce hypothesis space complexity, regularize gradients, and yield tighter certified robustness bounds under the CLEVER (Cross Lipschitz Extreme Value for nEtwork Robustness) framework. Empirically, our models consistently improve adversarial robustness and generalization across CIFAR-10, CIFAR-100, and CIFAR-10C under both FGSM and PGD attacks, without requiring adversarial training. These findings underscore the potential of symmetry-enforcing architectures as efficient and principled alternatives to data augmentation-based defenses.

Climate change poses severe risks to the survival of many whale populations, whose habitats and migration patterns are affected by environmental changes. To detect these whales effectively, especially in deep-sea environments, we need to use AI-based techniques to handle the acoustic diversity and variability of different species. However, current methods for whale detection are based on pre- and post-processing steps that reduce their efficiency and generalizability. To address this issue, we present DeepWhaleNet, a novel deep-learning model that automates whale detection in Underwater Passive Acoustic Monitoring datasets. DeepWhaleNet simplifies the detection process by extracting relevant features from raw log-power spectrograms and helps protect these threatened species by supporting conservation efforts. Our model uses a larger short-time Fourier transform as input and a custom ResNet-18 architecture for classification, which enables it to separate whale sounds from noise and capture their temporal and spectral characteristics. We evaluate the performance of DeepWhaleNet and show that it surpasses state-of-the-art methods, achieving an 8.3% improvement in the F-1 score and 21% higher average precision of binary relevance than the baseline method. Moreover, our model demonstrates its versatility and suitability for species-specific retrieval problems through an ablation study on multi-label retrieval problems and a 99.1% recall for Blue Whales.

Designing effective AI models becomes a challenge when dealing with imbalanced/skewed class distributions in datasets. Addressing this, re-sampling techniques often come into play as potential solutions. In this investigation, we delve into the male fertility dataset, exploring 14 re-sampling approaches to understand their impact on enhancing predictive model performance. The research employs conventional AI learners to gauge male fertility potential. Notably, five ensemble AI learners are studied, their performances are compared, and their results are evaluated using four measurement indices. Through comprehensive comparative analysis, we identify substantial enhancement in model effectiveness. Our findings showcase that the LightGBM model with SMOTE-ENN re-sampling stands out, achieving an efficacy of 96.66% and an F1-Score of 95.60% through 5-fold cross-validation. Interestingly, the CatBoost model, without re-sampling, exhibits strong performance, achieving an efficacy of 86.99% and an F1-Score of 93.02%. Furthermore, we benchmark our approach against state-of-the-art methods in male fertility prediction, particularly highlighting the use of re-sampling techniques like SMOTE and ESLSMOTE. Consequently, our proposed model emerges as a robust and efficient computational framework, promising accurate male fertility prediction.

Meta-learning empowers learning systems with the ability to acquire knowledge from multiple tasks, enabling faster adaptation and generalization to new tasks. This review provides a comprehensive technical overview of meta-learning, emphasizing its importance in real-world applications where data may be scarce or expensive to obtain. The article covers the state-of-the-art meta-learning approaches and explores the relationship between meta-learning and multi-task learning, transfer learning, domain adaptation and generalization, self-supervised learning, personalized federated learning, and continual learning. By highlighting the synergies between these topics and the field of meta-learning, the article demonstrates how advancements in one area can benefit the field as a whole, while avoiding unnecessary duplication of efforts. Additionally, the article delves into advanced meta-learning topics such as learning from complex multi-modal task distributions, unsupervised meta-learning, learning to efficiently adapt to data distribution shifts, and continual meta-learning. Lastly, the article highlights open problems and challenges for future research in the field. By synthesizing the latest research developments, this article provides a thorough understanding of meta-learning and its potential impact on various machine learning applications. We believe that this technical overview will contribute to the advancement of meta-learning and its practical implications in addressing real-world problems.

The papers in this special section focus on multimodal learning in medical imaging informatics. Enormous amounts of health-related data are produced daily, such as those from personal devices, e.g., fitness trackers or mobile applications, ambient sensors, clinical data in electronic health records, pathology reports, lab results, medical images, voice recordings, etc. The practice of modern medicine increasingly relies on data from multiple sources to guide better care. Toward this goal, the papers in this section describe the tools and techniques that integrate multiple data types to describe a particular medical event/case toward developing higher confidence in their decision-making and guidance.

Illness due to infectious diseases has been always a global threat. Millions of people die per year due to COVID-19, pneumonia, and Tuberculosis (TB) as all of them infect the lungs. For all cases, early screening/diagnosis can help provide opportunities for better care. To handle this, we develop an application, which we call MobApp4InfectiousDisease that can identify abnormalities due to COVID-19, pneumonia, and TB using Chest X-ray image. In our MobApp4InfectiousDisease, we implemented a customized deep network with a single transfer learning technique. For validation, we offered in-depth experimental study and we achieved, for COVID-19-pneumonia-TB cases, accuracy of 97.72%196.62%199.75%, precision of 92.72%1100.0%199.29%, recall of 98.89%188.54%199.65%, and F1-score of 95.00%194.00%199.00%. Our results are compared with state-of-the-art techniques. To the best of our knowl-edge, this is the first time we deployed our proof-of-the-concept MobApp4InfectiousDisease for a multi-class infec-tious disease classification.

In medical image segmentation, preparing ground truths (or masks) is not trivial as it requires expert clinicians to manually label regions-of-interest. Cervical cytology image segmentation is no exception. In this paper, we propose an unsupervised segmentation framework for cervical cell and whole slide segmentation uses an ensemble of three clustering algorithms namely, K-means, K-means++ and Mean Shift clustering. The final cluster centers obtained from these algorithms are used to initialize cluster points for Fuzzy C-means clustering algorithm. The proposed method is evaluated on multiple standard datasets: HErlev Pap Smear dataset and SIPaKMeD Pap Smear dataset. We also evaluated on a whole slide image dataset (source: CMATER-JU laboratory) and our results are promising and comparable. Overall, our results on multiple benchmark datasets justify the viability of the proposed framework.

Social platforms such as Twitter and Facebook have now become only media to express their thoughts, and due to lack of censorship, it often embellishes themselves as an abode for hate towards minorities. People of color, Asian people, Muslims, women, transgenders, and LGBTQ+ communities are often the target of such online hate and aggression. Though several companies have incorporated considerable algorithms on their platforms, nevertheless due to being rather hard to often detect such comments still make it to the platforms, creating a negative space towards targeted people. This research involves the study and comparison of different hate and aggression detection algorithms with intent on two languages, i.e. English and German including machine learning models (linear SVC, logistic regression, multinomial naive Bayes and random forests) with their variations with feature engineering and bag of words and deep learning (CNN-GRU static, TCN static, Seq2Seq) with their variations vis-à-vis Word2Vec embedding. CNN+GRU static + Word2Vec embedding has outperformed all the other techniques with an accuracy of 68.29%.

Artificial Intelligence (AI) has promoted countless contributions in the field of healthcare and medical imaging. In this paper, we thoroughly analyze peer-reviewed research findings/articles on AI-guided tools for Covid-19 analysis/screening using chest X-ray images in the year 2020. We discuss on how far deep learning algorithms help in decision-making. We identify/address data collections, methodical contributions, promising methods, and challenges. However, a fair comparison is not trivial as dataset sizes vary over time, throughout the year 2020. Even though their unprecedented efforts in building AI-guided tools to detect, localize, and segment Covid-19 cases are limited to education and training, we elaborate on their strengths and possible weaknesses when we consider the need of cross-population train/test models. In total, with search keywords: (Covid-19 OR Coronavirus) AND chest x-ray AND deep learning AND artificial intelligence AND medical imaging in both PubMed Central Repository and Web of Science, we systematically reviewed 58 research articles and performed meta-analysis.

Manifestations of brain tumors can trigger various psychiatric symptoms. Brain tumor detection can efficiently solve or reduce chances of occurrences of diseases, such as Alzheimer's disease, dementia-based disorders, multiple sclerosis and bipolar disorder. In this paper, we propose a segmentation-based approach to detect brain tumors in MRI11. We provide a comparative study between two different U-Net architectures (U-Net: baseline and U-Net: ResNeXt50 backbone) and a Feature Pyramid Network (FPN) that are trained/validated on the TCGA-LGG dataset of size 3, 929 images. U-Net architecture with ResNeXt50 backbone achieves the best Dice coefficient of 0.932, while baseline U-Net and FPN separately achieve Dice coefficients of 0.846 and 0.899, respectively. The results obtained from U-Net with ResNeXt50 backbone outperform previous works.

Gastric malignancy - one of the five most deadliest types of cancer - exceeds annual cases by a million worldwide since 2017. Automated screening tools may help speed up the screening and clinical procedures. In this paper, we propose a binary classification approach to classify gastrointestinal cancer tissues, namely Microsatellite Instable (MSI) and Microsatellite Stable (MSS) through stacked generalization based ensemble Deep Neural Network (DNN). Using a dataset of size 192, 315 images, we achieve an overall accuracy of 94.91% and sensitivity of 95.95%. Our results outperform previous works.

With over 200K cases in the U.S. alone, retinal disorders are the most common cause of irreversible blindness. This serves as a primary aim to analyze automated screening tools to detect retinal disorders. We analyze the OCT dataset (84, 484 images) and enhance the images by using Generative Adversarial Networks (GANs). This work specifically focuses on enhancing the quality of source (training) images for better algorithm validatiorr/testing11Authors contributed equally to the work.. We synthesize super resolution-based images using generators, discriminators and the adversarial nature of the GANs. The performance of the Ret-GAN is validated by PSNR, SSIM, and loss functions. To test the Ret-GAN generated images, we train a convolutional neural network (CNN) with the original dataset images and super-resolution images. We achieve an accuracy of 0.9825 on Ret-GAN generated image data, and 0.9525 on the original data. We statistically analyze the CNN with a number of evaluation metrics to further validate the results. The proposed scheme is compared to benchmark research findings on the same dataset. Our results are encouraging.

In today's world, databases are growing rapidly. Fast automatic detection of anomalous records in these massive databases is a challenging task. Traditional distance-based anomaly detectors are limited to small datasets because of their high time complexities. The univariate histogram-based method is arguably the fastest anomaly detection method. The anomaly score of a data instance is computed as the product of the probability mass of histograms in each dimension. Recent studies proved that such a simple method is comparable with many state-of-the-art methods on several datasets. However, as data features are assumed to be independent, it results in poor performance when features are correlated. Such an issue can be taken care of by using Principal Component (PC) features, which is the primary element of this paper. Our results show that integrating PCs with the original input features improves the performance of histogram-based anomaly detector with no real compromise in computational complexity.

Analysis of irregularities in Covid-19 data could open a new window to learn more about the unprecedented problems of the current global pandemic. Of many, radiographs and clinical records are reliable sources for viral infection investigation and treatment planning. Clinical records help track the Covid-19 pandemic. In this paper, we present a Spike Neural Network (SNN) with supervised synaptic learning to detect abnormalities in Chest X-rays (CXRs) In other words, the proposed SNN can distinguish Covid-19 positive cases from healthy ones. In our decision-making procedure, we introduce clinical practice so Explainable AI (XAI) is possible to carry out. In addition, Support Vector Machine (SVM) with local interpretable model-agnostic explanation (LIME) provides reliable analysis of abnormalities in Covid-19 clinical data.

Binarization plays a crucial role in Optical Character Recognition (OCR) ancillary domains, such as recovery of degraded document images. In Document Image Analysis (DIA), selecting threshold is not trivial since it differs from one problem (dataset) to another. Instead of trying several different thresholds for one dataset to another, we consider noise inherency of document images in our proposed binarization scheme. The proposed stochastic architecture implements the local thresholding technique: Niblack’s binarization algorithm. We introduce a stochastic comparator circuit that works on unipolar stochastic numbers. Unlike the conventional stochastic circuit, it is simple and easy to deploy. We implemented it on the Xilinx Virtex6 XC6VLX760-2FF1760 FPGA platform and received encouraging experimental results. The complete set of results are available upon request. Besides, compared to conventional designs, the proposed stochastic implementation is better in terms of time complexity as well as fault-tolerant capacity.

Audio-based healthcare technologies are among the most significant applications of pattern recognition and Artificial Intelligence. Lately, a major chunk of the World population has been infected with serious respiratory diseases such as COVID-19. Early recognition of lung health abnormalities can facilitate early intervention, and decrease the mortality rate of the infected population. Research has shown that it is possible to automatically monitor lung health abnormalities through respiratory sounds. In this paper, we propose an approach that employs filter bank energy-based features and Random Forests to classify lung problem types from respiratory sounds. The adventitious sounds, crackles and wheezes appear distinct to the human ear. Moreover, different sounds are characterized by different frequency ranges that are dominant. The proposed approach attempts to distinguish the adventitious sounds (crackles and wheezes) by modeling the human auditory perception of these sounds. Specifically, we propose a respiratory sounds representation technique capable of modeling the dominant frequency range present in such sounds. On a publicly available dataset (ICBHI) of size 6898 cycles spanning over 5[Formula: see text]h, our results can be compared with the state-of-the-art results, in distinguishing two different types of adventitious sounds: crackles and wheezes.

Since 1960s, Cardiotocography (CTG) has been considered the primary tool for monitoring fetal health during antepartum and intra-partum periods. It records both Fetal Heart Rate (FHR) and mother's Uterine Contraction Pressure (UCP) simultaneously. However, due to inter and intra-observer variations, the introduction of CTG in fetal care did little to reduce the fetal mortality and morbidity. To ensure that the signs of hypoxia are recognized at the onset it is needed to have a robust and automated clinical decision support system since the visual analysis (clinicians) can be error-prone. In this work, we proposed methods to identify the periodic changes i.e. acceleration and deceleration. Our method detected 987 accelerations and 1755 decelerations from the 556 CTG data. There were 96.6% and 97.3% agreements with the three clinicians estimate for acceleration and deceleration, respectively. Besides, we also proposed a novel method to detect Sinusoidal Heart Rate (SHR) pattern. With Random Forest classifier, the SHR classification accuracy was 93%. The sensitivity and specificity were 93% and 86%, respectively, while both Positive Predictive Value (PPV) and Negative Predictive Value (NPV) were found to be 100%. We conclude that the proposed method can be used as a "gold standard" for SHR identification.

Automated radiological screening is an advancing field in which algorithms and predictive models are used to detect abnormalities in Chest X-rays (CXRs). Traditionally, in machine learning, the exact same dataset has been partitioned into train and test sets, and as a consequence, the validation scores are often biased towards the population it has been trained on. Cross-population test is a measure of how good an algorithm/model performs after training on a data from one region of the world and then evaluating the model on another data from another part of the world, without any additional training or learning on the latter data. To showcase cross-population train/test model, we consider two benchmark CXR (with Tuberculosis) datasets that are made available by the U.S. National Library of Medicine: a) Shenzhen, China; and b) Montgomery County, USA. We used a modified pre-trained deep learning model as our predictive model and achieved a cross-population classification accuracy of 76.05% (0.84, AUC) and 71.47% (0.79, AUC), using each dataset as training and testing data separately. To the best of our knowledge, this is the first cross-population evaluation of a deep learning model being used for abnormality screening using CXRs.

Identifying skin diseases, such as leprosy, Tinea Versicolor, and Vitiligo identification is one of the challenging tasks. Therefore, skin disease identification success rate is comparatively poor as compared to the other computer vision tasks. Traditional Deep Learning (DL) models are not successful in this domain due to the lack of a huge number of data. To address the problem, in the present work, we introduced a customized Generative Adversarial Network (GAN) to generate synthetic data. With data augmentation, we achieved maximum 94.25% recognition accuracy using DensenNet-121, which was 10.95% better than when no augmentation was employed. Source code is publicly available at https://github.com/DVLP-CMATERJU/SkinDiseases_GenerativeAI.git GitHub.

In this paper, we deconstruct and demonstrate a detection framework to classify Retinal Optical Coherence Tomography (OCT) images across three classes namely, Diabetic Macular Edema (DME), Choroidal Neovascularization (CNV), and the DRUSEN from normal Retina. In this research, we developed on a Deep Ensemble Network by the virtue of which we were able to obtain a state-of-the-art accuracy of 98.53% on our test image dataset that was deliberately increased to 12% of the total images. Further, we also took advantage and insight from a feature map obtained from our convolutional layers to build our final model, which we call Optical Coherence Tomography Extended (OCTx). In our experiments, we found that OCTx was more accurate and diverse as compared to previously reported works that were validated on the exact same dataset.

In automated Chest X-Ray (CXR) screening process, foreign objects, such as coins/buttons, medical tubes and devices, and jewelries can adversely impact the performance. In an automated process, conventional machine learning algorithms did not separately consider them into account, and as a consequence, they results in false positive cases. In this paper, we address the use of Deep Neural Network (DNN) to detect circle-like foreign objects of difference sizes in CXRs. We present faster Region-based Convolutional Neural Network (R-CNN) for foreign object detection on a set of 400 publicly available CXR images hosted by LHNCBC, U.S. National Library of Medicine (NLM), National Institutes of Health (NIH). The proposed DNN achieved 97% precision, 90% recall, and 93% F1-score. The results are comparable with the existing techniques.

The motivation for this work is the primary need of screening Tuberculosis (TB) positive patients in the severely resource constrained regions of the world. Chest X-ray (CXR) is considered to be a promising indicator for the onset of TB, but the lack of skilled radiologists in such resource constrained regions degrades the situation. Therefore, several computer aided diagnosis (CAD) systems have been proposed to solve the decision making problem, which includes hand-engineered feature extraction methods to deep learning or Convolutional Neural Network (CNN) based methods. Feature extraction, being a time and resource intensive process, often delays the process of mass screening. Hence, an end to end CNN architecture is proposed in this work to solve the problem. Two benchmark CXR datasets have been used in this work, collected from Shenzhen (China) and Montgomery County (USA), on which the proposed methodology achieved a maximum abnormality detection accuracy (ACC) of 91.7% (0.96 AUC) and 87.47% (0.92 AUC) respectively. Considering these datasets, to the best of our knowledge, the obtained results are superior to the state of the art deep learning based works.

The writing style is a unique characteristic of a human being as it varies from one person to another. Due to such diversity in writing style, handwritten character recognition (HCR) under the purview of pattern recognition is not trivial. Conventional methods used handcrafted features that required a-priori domain knowledge, which is always not feasible. In such a case, extracting features automatically could potentially attract more interests. For this, in the literature, convolutional neural network (CNN) has been a popular approach to extract features from the image data. However, state-of-the-art works do not provide a generic CNN model for character recognition, Devanagari script, for instance. Therefore, in this work, we first study several different CNN models on publicly available handwritten Devanagari characters and numerals datasets. This means that our study is primarily focusing on comparative study by taking trainable parameters, training time and memory consumption into account. Later, we propose and design DevNet, a modified CNN architecture that produced promising results, since computational complexity and memory space are our primary concerns in design.

Speech recognition in multilingual scenario is not trivial in the case when multiple languages are used in one conversation. Language must be identified before we process speech recognition as such tools are language-dependent. We present a language identification system (or AI tool) to distinguish top-seven world languages namely Chinese, Spanish, English, Hindi, Arabic, Bangla and Portuguese [G. F. Simons and C. D. Fennig (eds.), Ethnologue: Laguage of the Americas and the Pacific, Twentieth Edn. (SIL Internatinal, 2017)]. The system uses linear predictive coefficients-based feature, i.e. the line spectral pair–grade ratio (LSP–GR) feature, and ensemble learning for classification. Experiments were performed on more than 200[Formula: see text]h of real-world YouTube data and the highest possible accuracy of 96.95% was received. The results can be compared with other machine learning classifiers.

Offline writer identification is one of the major fields of study in behavioral biometric. It is a process of matching a questioned document with other documents of known writers to find the appropriate writer. In this paper, local handwriting-based attributes are used as features, and multi-layer perceptron and simple logistic classifiers are used for decision making. The method is tested on an unconstrained handwritten Bangla database of 1383 documents with variable number of datasets from 190 writers. Experimental results show the effectiveness of our system, since it outperforms the state-of-the-art methods by approximately 3% (top-3 and top-4 choices). Further, our method is approximately 27 times faster than conventional segmentation-based methods.

Script identification is crucial for automating optical character recognition (OCR) in multi-script documents since OCRs are script-dependent. In this paper, we present a comprehensive survey of the techniques developed for handwritten Indic script identification. Different pre-processing and feature extraction techniques, including classifiers used for script identification, are categorized and their merits and demerits are discussed. We also provide information about some handwritten Indic script datasets. Finally, we highlight the extensions and/or future scope of works together with challenges.

Our primary motivator is the need for screening HIV+ populations in resource-constrained regions for exposure to Tuberculosis, using posteroanterior chest radiographs (CXRs). The proposed method is motivated by the observation that radiological examinations routinely conduct bilateral comparisons of the lung field. In addition, the abnormal CXRs tend to exhibit changes in the lung shape, size, and content (textures), and in overall, reflection symmetry between them. We analyze the lung region symmetry using multi-scale shape features, and edge plus texture features. Shape features exploit local and global representation of the lung regions, while edge and texture features take internal content, including spatial arrangements of the structures. For classification, we have performed voting-based combination of three different classifiers: Bayesian network, multilayer perception neural networks, and random forest. We have used three CXR benchmark collections made available by the U.S. National Library of Medicine and the National Institute of Tuberculosis and Respiratory Diseases, India, and have achieved a maximum abnormality detection accuracy (ACC) of 91.00% and area under the ROC curve (AUC) of 0.96. The proposed method outperforms the previously reported methods by more than 5% in ACC and 3% in AUC.

We present a novel approach for separating Indic scripts with ‘matra’, which is used as a precursor to advance and/or ease subsequent handwritten script identification in multi-script documents. In our study, among state-of-the-art features and classifiers, an optimized fractal geometry analysis and random forest are found to be the best performer to distinguish scripts with ‘matra’ from their counterparts. For validation, a total of 1204 document images are used, where two different scripts with ‘matra’: Bangla and Devanagari are considered as positive samples and the other two different scripts: Roman and Urdu are considered as negative samples. With this precursor, an overall script identification performance can be advanced by more than 5.13% in accuracy and 1.17 times faster in processing time as compared to conventional system.

We present a novel technique to separate panels from stitched multipanel figures appearing in biomedical research articles. Since such figures may comprise images from different imaging modalities, separating them is a crucial first step for effective biomedical content-based image retrieval (CBIR): multimodal biomedical document classification and/or retrieval, for instance. The method applies local line segment detection based on the gray-level pixel changes. It then applies a line vectorization process that connects prominent broken lines along the panel boundaries while eliminating insignificant line segments within the panels. We validated our fully automatic technique on a set of stitched multipanel biomedical figures extracted from articles within the Open Access subset of PubMed Central® repository, and achieved precision and recall of 87.16% and 83.51%, respectively, in less than 0.461[Formula: see text]s per image, on average. We also reported the recent ImageCLEF 2015 competition results that highlight the usefulness of the proposed work.

In biomedical documents/publications, medical images tend to be complex by nature and often contain several regions that are annotated using arrows. In this context, an automated arrowhead detection is a critical precursor to region-of-interest (ROI) labeling and image content analysis. To detect arrowheads, in this paper, images are first binarized using fuzzy binarization technique to segment a set of candidates based on connected component (CC) principle. To select arrow candidates, we use convexity defect-based filtering, which is followed by template matching via dynamic time warping (DTW). The DTW similarity score confirms the presence of arrows in the image. Our test results on biomedical images from imageCLEF 2010 collection shows the interest of the technique, and can be compared with previously reported state-of-the-art results.

We present a novel technique to separate subpanels from stitched multipanel figures appearing in biomedical research articles. Since such figures may comprise images from different imaging modalities, separating them is a critical first step for effective biomedical content-based image retrieval (CBIR). The method applies local line segment detection based on the gray-level pixel changes. It then applies a line vectorization process that connects prominent broken lines along the subpanel boundaries while eliminating insignificant line segments within the subpanels. We have validated our fully automatic technique on a subset of stitched multipanel biomedical figures extracted from articles within the Open Access subset of PubMed Central repository, and have achieved precision and recall of 81.22% and 85.08%, respectively.

We present a novel method for screening pulmonary abnormalities using thoracic edge map in PA chest radiograph (CXR) images. Our particular interest is to aid clinical officers in screening HIV+ populations in resource constrained regions for Tuberculosis (TB). Our work is motivated by the observation that abnormal CXRs tend to exhibit corrupted and/or deformed thoracic edge maps. We study histograms of thoracic edges for all possible orientations of gradients in the range [0, 2π) at different numbers of bins and different pyramid levels. We have used two CXR benchmark collections made available by the U.S. National Library of Medicine, and have achieved a maximum abnormality detection accuracy of 85.92% and area under the ROC curve (AUC) of 0.91 at one second per image, on average, which outperforms the reported state-of-the-art.

We present a novel method for detecting rotated lungs in chest radiographs for quality control and augmenting automated abnormality detection. The method computes a principal rib-orientation measure using a generalized line histogram technique for quality control, and therefore augmenting automated abnormality detection. To compute the line histogram, we use line seed filters as kernels to convolve with edge images, and extract a set of lines from the posterior rib-cage. After convolving kernels in all possible orientations in the range [0°, 180°), we measure the angle with maximum magnitude in the line histogram. This measure provides an approximation of the principal chest rib-orientation for each lung. A chest radiograph is upright if the difference between the orientation angles of both lungs with respect to the horizontal axis is negligible. We validate our method on sets of normal and abnormal images and argue that rib orientation can be used for rotation detection in chest radiographs as an aid in quality control during image acquisition. It can also be used for training and testing data sets for computer aided diagnosis research, for example. In our experiments, we achieve a maximum accuracy of approximately 90%.

We present a generalized line histogram technique to compute global rib-orientation for detecting rotated lungs in chest radiographs. We use linear structuring elements, such as line seed filters, as kernels to convolve with edge images, and extract a set of lines from the posterior rib-cage. After convolving kernels in all possible orientations in the range [0, π], we measure the angle for which the line histogram has maximum magnitude. This measure provides a good approximation of the global chest rib-orientation for each lung. A chest radiograph is said to be upright if the difference between the orientation angles of both lungs with respect to the horizontal axis, is negligible. We validate our method on sets of normal and abnormal images and argue that rib orientation can be used for rotation detection in chest radiographs as aid in quality control during image acquisition, and to discard images from training and testing data sets. In our test, we achieve a maximum accuracy of 90%.

Since OCR engines are usually script-dependent, automatic text recognition in multi-script document requires a pre-processor module that identifies the scripts. Based on this motivation, in this paper, we present a word level handwritten Indian script identification technique. To handle this, words are first segmented by morphological dilation and performed connected component labelling. We then employ the Radon transform, discrete wavelet transform, statistical filters and discrete cosine transform to extract the directional multi-resolution spatial features. We tested the features by using linear discriminant analysis, support vector machine and K-nearest neighbour classifiers over 11 different major Indian scripts (including Roman) in bi-script and tri-script scenario. In our tests, we have achieved maximum accuracies of 98% and 96% for bi-script and tri-scipt respectively.

In this paper, we present a scalable arrow detection technique for biomedical images to support information retrieval systems under the purview of content-based image retrieval (CBIR) and text information retrieval (TIR). The idea primarily follows the criteria based on the geometric properties of the arrow, where we introduce signatures from key points associated with it. To handle this, images are first binarized via a fuzzy binarization tool and several regions of interest are labeled accordingly. Each region is used to generate signatures and then compared with the theoretical ones to check their similarity. Our validation over biomedical images shows the advantage of the technique over the most prominent state-of-the-art methods.

In this paper, we address a new scheme for symbol retrieval based on bag-of-relations (BoRs) which are computed between extracted visual primitives (e.g. circle and corner). Our features consist of pairwise spatial relations from all possible combinations of individual visual primitives. The key characteristic of the overall process is to use topological relation information indexed in BoRs and use this for recognition. As a consequence, directional relation matching takes place only with those candidates having similar topological configurations. A comprehensive study is made by using several different well-known datasets such as GREC, FRESH and SESYD, and includes a comparison with state-of-the-art descriptors. Experiments provide interesting results on symbol spotting and other user-friendly symbol retrieval applications.

This paper presents directional discrete cosine transform (D-DCT) based word level handwritten script identification. The conventional discrete cosine transform (DCT)emphasizes vertical and horizontal energies of an image and de-emphasizes directional edge information, which of course plays a significant role in shape analysis problem, in particular. Conventional DCT however, is not efficient in characterizing the images where directional edges are dominant. In this paper, we investigate two different methods to capture directional edge information, one by performing 1D-DCT along left and right diagonals of an image, and another by decomposing 2D-DCT coefficients in left and right diagonals. The mean and standard deviations of left and right diagonals of DCT coefficients are computed and are used for the classification of words using linear discriminant analysis (LDA) and K-nearest neighbour (K-NN). We validate the method over 9000 words belonging to six different scripts. The classification of words is performed at bi-scripts, triscripts and multi-scripts scenarios and accomplished the identification accuracies respectively as 96.95%, 96.42% and 85.77% in average.

In this paper, we present a model-based document information content extraction approach and perform in-depth evaluation based on clients' relevance. Real-world users i.e., clients first provide a set of key fields from the document image which they think are important. These are used to represent a graph where nodes (i.e., fields) are labelled with dynamic semantics including other features and edges are attributed with spatial relations. Such an attributed relational graph (ARG) is then used to mine similar graphs from a document image that are used to reinforce or update the initial graph iteratively each time we extract them, in order to produce a model. Models therefore, can be employed in the absence of clients. We have validated the concept and evaluated its scientific impact on real-world industrial problem, where table extraction is found to be the best suited application.

In this paper, we address a new scheme for symbol retrieval based on relation bag-of-features (BOFs) which are computed between the extracted visual primitives. Our feature consists of pair wise spatial relations from all possible combinations of individual visual primitives. The key characteristic of the overall process is to use topological information to guide directional relations. Consequently, directional relation matching takes place only with those candidates having similar topological configurations. A comprehensive study is made by using two different datasets. Experimental tests provide interesting results by establishing user-friendly symbol retrieval application.

In this paper, we present a pattern recognition method that uses dynamic programming for the alignment of Radon features. The key characteristic of the method is to use dynamic time warping (DTW) to match corresponding pairs of the Radon features for all possible projections. Thanks to DTW, we avoid compressing the feature matrix into a single vector which would otherwise miss information. To reduce the possible number of matchings, we rely on a initial normalization based on the pattern orientation. A comprehensive study is made using major state-of-the-art shape descriptors over several public datasets of shapes such as graphical symbols (both printed and hand-drawn), handwritten characters and footwear prints. In all tests, the method proves its generic behavior by providing better recognition performance. Overall, we validate that our method is robust to deformed shape due to distortion, degradation and occlusion.

The paper presents a method for isolated off-line character recognition using radon features. The key characteristic of the method is to use DTW algorithm to match corresponding pairs of radon histograms at every projecting angle. Thanks to DTW, it avoids compressing feature matrix into a single vector which may miss information. Comparison has been made with the state-of-the-art of shape descriptors over several different character as well as numeral datasets from different scripts.

In this paper, we present an innovative approach to integrate spatial relations in stroke clustering for handwritten Devanagari character recognition. It handles strokes of any number and order, writer independently. Learnt strokes are hierarchically agglomerated via Dynamic Time Warping based on their location and their number and stored accordingly. We experimentally validate our concept by showing its ability to improve recognition performance on previously published results.

In this paper, we address the use of unified spatial relations for symbol description. We present a topologically guided directional relation signature. It references a unique point set instead of one entity in a pair, thus avoiding problems related to erroneous choices of reference entities and preserves symmetry. We experimentally validate our method on showing its ability to serve in a symbol retrieval application, based only on a spatial relational descriptor that represents the links between the decomposed structural patterns called "vocabulary" in a spatial relational graph.

In this paper, we make an attempt to use inductive logic programming (ILP) to automatically learn non trivial descriptions of symbols, based on a formal description. This work is a first step in this direction and is rather a proof of concept, rather than a fully operational and robust framework. The overall goal of our approach is to express graphic symbols by a number of primitives that may be of any complexity (i.e. not necessarily just lines or points) and connecting relationships that can be deduced from straightforward state-of-the art image treatment and analysis tools. This representation is then used as an input to an ILP solver, in order to deduce non obvious characteristics that may lead to a more semantic related recognition process.