ABSTRACT Ultra‐short pulse (USP) laser micromachining is a key technology for sustainable manufacturing, offering high precision and minimal thermal damage across a wide range of materials. To enable its effective deployment in industrial environments, it is essential to develop monitoring systems capable of accurately predicting surface roughness at early processing stages, regardless of the initial workpiece condition. Given the high dimensionality of sensor data typically involved, real‐time applicability requires lightweight, interpretable and computationally efficient machine learning (ML) models. This work presents an ML‐based framework that addresses these requirements through explainability‐driven feature reduction. By identifying and selecting the most relevant sensor features, the approach reduces input dimensionality while preserving model performance. Additionally, the computational cost of feature extraction is evaluated to assess the framework's feasibility in real‐time scenarios. Overall, the proposed system is designed to adapt across multiple preprocessing techniques while minimizing processing latency, supporting the deployment of efficient monitoring solutions for industrial USP laser micromachining.

• Our experimentation demonstrates that slide-level zero-shot transfer relying on textual ensemble learning outperforms randomly initialized MIL classification models. • We propose a novel framework termed MIL-Adapter that leverages histopathology vision-language models and enables robust slide-level classification in data- and parameter-efficient settings. • MIL-Adapter couples foundation model feature extraction and lightweight MIL models with vision-language adapters to perform consistent and interpretable prediction of the slides. • We conduct extensive validation of our framework in cancer subtyping tasks across a variety of organs and through diverse configurations of few-shot learning scenarios. Contrastive language-image pretraining has greatly enhanced visual representation learning and enabled zero-shot classification. Vision-language language models (VLM) have succeeded in few-shot learning by leveraging adaptation modules fine-tuned for specific downstream tasks. In computational pathology (CPath), accurate whole-slide image (WSI) prediction is crucial for aiding in cancer diagnosis, and multiple instance learning (MIL) remains essential for managing the gigapixel scale of WSIs. In the intersection between CPath and VLMs, the literature still lacks specific adapters that handle the particular complexity of the slides. To solve this gap, we introduce MIL-Adapter, a novel approach designed to obtain consistent slide-level classification under few-shot learning scenarios. In particular, our framework is the first to combine trainable MIL aggregation functions and lightweight visual-language adapters to improve the performance of histopathological VLMs. MIL-Adapter relies on textual ensemble learning to construct discriminative zero-shot prototypes. It is serves as a solid starting point, surpassing MIL models with randomly initialized classifiers in data-constrained settings. With our experimentation, we demonstrate the value of textual ensemble learning and the robust predictive performance of MIL-Adapter through diverse datasets and configurations of few-shot scenarios, while providing crucial insights on model interpretability. The code is publicly accessible in https://github.com/cvblab/MIL-Adapter .

Prostate cancer is a critical healthcare challenge globally and is one of the most prevalent types of cancer in men. Early and accurate diagnosis is essential for effective treatment and improved patient outcomes. In the existing literature, computer-aided diagnosis (CAD) solutions have been developed to assist pathologists in various tasks, including classification, diagnosis, and prostate cancer grading. Content-based image retrieval (CBIR) techniques provide valuable approaches to enhance these computer-aided solutions. This study evaluates how generative deep learning models can improve the quality of retrievals within a CBIR system. Specifically, we propose applying a Siamese Network approach, which enables us to learn how to encode image patches into latent representations for retrieval purposes. We used the ProGleason-GAN framework trained on the SiCAPv2 dataset to create similar pairs of input patches. Our observations indicate that introducing synthetic patches leads to notable improvements in the evaluated metrics, underscoring the utility of generative models within CBIR tasks. Furthermore, this work is the first in the literature where latent representations optimized for CBIR are used to train an attention mechanism for performing Gleason Scoring of a WSI.

Abstract Many current embedded systems comprise heterogeneous computing components including quite powerful GPUs, which enables their application across diverse sectors. This study demonstrates the efficient execution of a medium-sized self-supervised audio spectrogram transformer (SSAST) model on a low-power system-on-chip (SoC). Through comprehensive evaluation, including real time inference scenarios, we show that GPUs outperform multi-core CPUs in inference processes. Optimization techniques such as adjusting batch size, model compilation with TensorRT, and reducing data precision significantly enhance inference time, energy consumption, and memory usage. In particular, negligible accuracy degradation is observed, with post-training quantization to 8-bit integers showing less than 1% loss. This research underscores the feasibility of deploying transformer neural networks on low-power embedded devices, ensuring efficiency in time, energy, and memory, while maintaining the accuracy of the results.

The rapid advancement of Deep Learning has sparked significant interest in application across various domains, notably in healthcare. However, the integration of Deep Learning models into clinical practice, particularly for sensitive tasks such as cancer diagnosis, presents challenges including model accuracy, interpretability, and data privacy. This last point, data privacy, is a major and challenging problem in healthcare, given the sensitivity of health data to security breaches such as leakage, tampering, and extraction. Strict regulations limit aspects like distribution, storage, and use, increase the difficulty of the clinical use of sophisticated Deep Learning models. These models generally require substantial hardware resources and are often hosted in the cloud, further exacerbating privacy concerns. In response to these challenges, this paper focuses on the optimization and quantification of a U-Net model for cancer tissue segmentation, with the ultimate goal of deploying it on edge devices for efficient inference. Leveraging Deep Learning techniques for histopathological image analysis, particularly in cancer detection, holds immense promise for improving diagnosis outcomes and treatment strategies. The study utilizes a dataset of invasive breast cancer and trains a U-Net segmentation model tailored to mimic biopsy examination carried out by pathologists. Through meticulous optimization using TensorFlow Lite, the model's size and computational complexity are significantly reduced, rendering it suitable for deployment on resource-constrained edge devices such as Jetson Orin nano and Kria KV260. To address concerns related to model complexity, computational costs, and hardware limitations, techniques such as sparsity pruning and post-training quantization are employed. These techniques enable the development of lighter models without compromising accuracy. By overcoming these hurdles, Deep Learning models can potentially revolutionize histopathological image analysis and enhance cancer diagnosis outcomes, paving the way for more efficient and accurate clinical decision-making in the realm of oncology.

Artificial intelligence (AI) agents encounter the problem of catastrophic forgetting when they are trained in sequentially with new data batches. This issue poses a barrier to the implementation of AI-based models in tasks that involve ongoing evolution, such as cancer prediction. Moreover, whole slide images (WSI) play a crucial role in cancer management, and their automated analysis has become increasingly popular in assisting pathologists during the diagnosis process. Incremental learning (IL) techniques aim to develop algorithms capable of retaining previously acquired information while also acquiring new insights to predict future data. Deep IL techniques need to address the challenges posed by the gigapixel scale of WSIs, which often necessitates the use of multiple instance learning (MIL) frameworks. In this paper, we introduce an IL algorithm tailored for analyzing WSIs within a MIL paradigm. The proposed Multiple Instance Class-Incremental Learning (MICIL) algorithm combines MIL with class-IL for the first time, allowing for the incremental prediction of multiple skin cancer subtypes from WSIs within a class-IL scenario. Our framework incorporates knowledge distillation and data rehearsal, along with a novel embedding-level distillation, aiming to preserve the latent space at the aggregated WSI level. Results demonstrate the algorithm's effectiveness in addressing the challenge of balancing IL-specific metrics, such as intransigence and forgetting, and solving the plasticity-stability dilemma.

Abstract Sound classification using neural networks has recently produced very accurate results. A large number of different applications use this type of sound classifiers such as controlling and monitoring the type of activity in a city or identifying different types of animals in natural environments. While traditional acoustic processing applications have been developed on high-performance computing platforms equipped with expensive multi-channel audio interfaces, the Internet of Things (IoT) paradigm requires the use of more flexible and energy-efficient systems. Although software-based platforms exist for implementing general-purpose neural networks, they are not optimized for sound classification, wasting energy and computational resources. In this work, we have used FPGAs to develop an ad hoc system where only the hardware needed for our application is synthesized, resulting in faster and more energy-efficient circuits. The results show that our developments are accelerated by a factor of 35 compared to a software-based implementation on a Raspberry Pi.

Cutaneous spindle cell (CSC) neoplasms are a group of tumors that represent a formidable diagnostic challenge for dermatopathologists. Digital pathology has enabled the application of new methods based on artificial intelligence to reduce the workload of pathologists’ daily practice. In this work, we propose a self-learning framework to detect tumor regions in histological images. The use of a teacher-model paradigm increases the annotated database while avoiding manual annotation. The pre-trained latent space of this model is then used in a second stage by another model to differentiate between leiomyomas (benign cases) and leiomyosarcomas (malignant cases). A contrastive learning approach allows separating the latent space of samples from different classes. This framework was tested on an independent database. This novel approach supposes a step forward in the CSC detection as the obtained results suggest (Acc = 0.90 and 0.8451, respectively).

The advent of artificial intelligence-based tools applied to digital pathology brings the promise of reduced workload for pathologists and enhanced patient care, not to mention medical research progress. Yet, despite its great potential, the field is hindered by the paucity of annotated histological data, a limitation for developing robust deep learning models. To reduce the number of expert annotations needed for training, we introduce a novel framework combining self-training and weakly-supervised learning that uses both annotated and unannotated data samples. Inspired by how pathologists examine biopsies, our method considers whole slide images from a bird’s eye view to roughly localize the tumor area before focusing on its features at a higher magnification level. Notwithstanding the scarcity of the dataset, the experimental results show that the proposed method outperforms models trained with annotated data only and previous works analyzing the same type of lesions, thus demonstrating the efficiency of the approach.

Increasing demands on railway structures have led to a need for new cost-effective maintenance strategies in recent years. Current dynamic railway track monitoring systems are usually based on the analysis of axle-box accelerations to automatically detect track singularities and defects. These methods rely on hand-crafted feature extraction and classifiers for different tasks. However, the low performance shown in previous literature makes it necessary to complement these analyses with in-situ inspections. Very recent works have proposed the use of deep learning systems that allow extracting more generalizable features from time–frequency spectrograms. However, the lack of specific public domain datasets and the finite number of track singularities in a railway structure have limited the development of deep learning based systems. In this paper, we propose a method capable of outstanding in low-data scenarios. In particular, we explore the use of supervised contrastive learning to cluster class embeddings nearly in the encoder latent space, which is used during inference for prototypical distance-based class assignment. We provide comprehensive experiments demonstrating the performance of our method in comparison to previous literature for detecting worn-out crossings.

Current unsupervised anomaly localization approaches rely on generative models to learn the distribution of normal images, which is later used to identify potential anomalous regions derived from errors on the reconstructed images. However, a main limitation of nearly all prior literature is the need of employing anomalous images to set a class-specific threshold to locate the anomalies. This limits their usability in realistic scenarios, where only normal data is typically accessible. Despite this major drawback, only a handful of works have addressed this limitation, by integrating supervision on attention maps during training. In this work, we propose a novel formulation that does not require accessing images with abnormalities to define the threshold. Furthermore, and in contrast to very recent work, the proposed constraint is formulated in a more principled manner, leveraging well-known knowledge in constrained optimization. In particular, the equality constraint on the attention maps in prior work is replaced by an inequality constraint, which allows more flexibility. In addition, to address the limitations of penalty-based functions we employ an extension of the popular log-barrier methods to handle the constraint. Last, we propose an alternative regularization term that maximizes the Shannon entropy of the attention maps, reducing the amount of hyperparameters of the proposed model. Comprehensive experiments on two publicly available datasets on brain lesion segmentation demonstrate that the proposed approach substantially outperforms relevant literature, establishing new state-of-the-art results for unsupervised lesion segmentation, and without the need to access anomalous images.

Abstract Epigenetic alterations have an important role in the development of several types of cancer. Epigenetic studies generate a large amount of data, which makes it essential to develop novel models capable of dealing with large-scale data. In this work, we propose a deep embedded refined clustering method for breast cancer differentiation based on DNA methylation. In concrete, the deep learning system presented here uses the levels of CpG island methylation between 0 and 1. The proposed approach is composed of two main stages. The first stage consists in the dimensionality reduction of the methylation data based on an autoencoder. The second stage is a clustering algorithm based on the soft assignment of the latent space provided by the autoencoder. The whole method is optimized through a weighted loss function composed of two terms: reconstruction and classification terms. To the best of the authors’ knowledge, no previous studies have focused on the dimensionality reduction algorithms linked to classification trained end-to-end for DNA methylation analysis. The proposed method achieves an unsupervised clustering accuracy of 0.9927 and an error rate (%) of 0.73 on 137 breast tissue samples. After a second test of the deep-learning-based method using a different methylation database, an accuracy of 0.9343 and an error rate (%) of 6.57 on 45 breast tissue samples are obtained. Based on these results, the proposed algorithm outperforms other state-of-the-art methods evaluated under the same conditions for breast cancer classification based on DNA methylation data.

Prostate cancer is one of the main diseases affecting men worldwide. The gold standard for diagnosis and prognosis is the Gleason grading system. In this process, pathologists manually analyze prostate histology slides under microscope, in a high time-consuming and subjective task. In the last years, computer-aided-diagnosis (CAD) systems have emerged as a promising tool that could support pathologists in the daily clinical practice. Nevertheless, these systems are usually trained using tedious and prone-to-error pixel-level annotations of Gleason grades in the tissue. To alleviate the need of manual pixel-wise labeling, just a handful of works have been presented in the literature. Furthermore, despite the promising results achieved on global scoring the location of cancerous patterns in the tissue is only qualitatively addressed. These heatmaps of tumor regions, however, are crucial to the reliability of CAD systems as they provide explainability to the system's output and give confidence to pathologists that the model is focusing on medical relevant features. Motivated by this, we propose a novel weakly-supervised deep-learning model, based on self-learning CNNs, that leverages only the global Gleason score of gigapixel whole slide images during training to accurately perform both, grading of patch-level patterns and biopsy-level scoring. To evaluate the performance of the proposed method, we perform extensive experiments on three different external datasets for the patch-level Gleason grading, and on two different test sets for global Grade Group prediction. We empirically demonstrate that our approach outperforms its supervised counterpart on patch-level Gleason grading by a large margin, as well as state-of-the-art methods on global biopsy-level scoring. Particularly, the proposed model brings an average improvement on the Cohen's quadratic kappa ( κ) score of nearly 18% compared to full-supervision for the patch-level Gleason grading task. This suggests that the absence of the annotator's bias in our approach and the capability of using large weakly labeled datasets during training leads to higher performing and more robust models. Furthermore, raw features obtained from the patch-level classifier showed to generalize better than previous approaches in the literature to the subjective global biopsy-level scoring.

Nowadays, glaucoma is the leading cause of blindness worldwide. We propose in this paper two different deep-learning based approaches to address glaucoma detection just from raw circumpapillary OCT images. The first one is based on the development of convolutional neural networks (CNNs) trained from scratch. The second one lies in fine-tuning some of the most common state-of-the-art CNNs architectures. The experiments were performed on a private database composed of 93 glaucomatous and 156 normal B-scans around the optic nerve head of the retina, which were diagnosed by expert ophthalmologists. The validation results evidence that finetuned CNNs outperform the networks trained from scratch when small databases are addressed. Additionally, the VGG family of networks reports the most promising results, with an area under the ROC curve of 0.96 and an accuracy of 0.92, during the prediction of the independent test set.

Worldwide, prostate cancer is one of the main cancers affecting men. The final diagnosis of prostate cancer is based on the visual detection of Gleason patterns in prostate biopsy by pathologists. Computer-aided-diagnosis systems allow to delineate and classify the cancerous patterns in the tissue via computer-vision algorithms in order to support the physicians' task. The methodological core of this work is a U-Net convolutional neural network for image segmentation modified with residual blocks able to segment cancerous tissue according to the full Gleason system. This model outperforms other well-known architectures, and reaches a pixel-level Cohen's quadratic Kappa of 0.52, at the level of previous image-level works in the literature, but providing also a detailed localisation of the patterns.

Color deconvolution aims at separating multi-stained images into single stained ones. In digital histopathological images, true stain color vectors vary between images and need to be estimated to obtain stain concentrations and separate stain bands. These band images can be used for image analysis purposes and, once normalized, utilized with other multi-stained images (from different laboratories and obtained using different scanners) for classification purposes. In this paper we propose the use of Super Gaussian (SG) priors for each stain concentration together with the similarity to a given reference matrix for the color vectors. Variational inference and an evidence lower bound are utilized to automatically estimate all the latent variables. The proposed methodology is tested on real images and compared to classical and state-of-the-art methods for histopathological blind image color deconvolution.

Glaucoma is one of the leading causes of irreversible but preventable blindness in working age populations. Color fundus photography (CFP) is the most cost-effective imaging modality to screen for retinal disorders. However, its application to glaucoma has been limited to the computation of a few related biomarkers such as the vertical cup-to-disc ratio. Deep learning approaches, although widely applied for medical image analysis, have not been extensively used for glaucoma assessment due to the limited size of the available data sets. Furthermore, the lack of a standardize benchmark strategy makes difficult to compare existing methods in a uniform way. In order to overcome these issues we set up the Retinal Fundus Glaucoma Challenge, REFUGE (https://refuge.grand-challenge.org), held in conjunction with MICCAI 2018. The challenge consisted of two primary tasks, namely optic disc/cup segmentation and glaucoma classification. As part of REFUGE, we have publicly released a data set of 1200 fundus images with ground truth segmentations and clinical glaucoma labels, currently the largest existing one. We have also built an evaluation framework to ease and ensure fairness in the comparison of different models, encouraging the development of novel techniques in the field. 12 teams qualified and participated in the online challenge. This paper summarizes their methods and analyzes their corresponding results. In particular, we observed that two of the top-ranked teams outperformed two human experts in the glaucoma classification task. Furthermore, the segmentation results were in general consistent with the ground truth annotations, with complementary outcomes that can be further exploited by ensembling the results.

Generalized nucleus segmentation techniques can contribute greatly to reducing the time to develop and validate visual biomarkers for new digital pathology datasets. We summarize the results of MoNuSeg 2018 Challenge whose objective was to develop generalizable nuclei segmentation techniques in digital pathology. The challenge was an official satellite event of the MICCAI 2018 conference in which 32 teams with more than 80 participants from geographically diverse institutes participated. Contestants were given a training set with 30 images from seven organs with annotations of 21,623 individual nuclei. A test dataset with 14 images taken from seven organs, including two organs that did not appear in the training set was released without annotations. Entries were evaluated based on average aggregated Jaccard index (AJI) on the test set to prioritize accurate instance segmentation as opposed to mere semantic segmentation. More than half the teams that completed the challenge outperformed a previous baseline. Among the trends observed that contributed to increased accuracy were the use of color normalization as well as heavy data augmentation. Additionally, fully convolutional networks inspired by variants of U-Net, FCN, and Mask-RCNN were popularly used, typically based on ResNet or VGG base architectures. Watershed segmentation on predicted semantic segmentation maps was a popular post-processing strategy. Several of the top techniques compared favorably to an individual human annotator and can be used with confidence for nuclear morphometrics.

Recent works show that generative adversarial networks (GANs) can be successfully applied to image synthesis and semi-supervised learning, where, given a small labeled database and a large unlabeled database, the goal is to train a powerful classifier. In this paper, we trained a retinal image synthesizer and a semi-supervised learning method for automatic glaucoma assessment using an adversarial model on a small glaucoma-labeled database and a large unlabeled database. Various studies have shown that glaucoma can be monitored by analyzing the optic disc and its surroundings, and for that reason, the images used in this paper were automatically cropped around the optic disc. The novelty of this paper is to propose a new retinal image synthesizer and a semi-supervised learning method for glaucoma assessment based on the deep convolutional GANs. In addition, and to the best of our knowledge, this system is trained on an unprecedented number of publicly available images (86926 images). This system, hence, is not only able to generate images synthetically but to provide labels automatically. Synthetic images were qualitatively evaluated using t-SNE plots of features associated with the images and their anatomical consistency was estimated by measuring the proportion of pixels corresponding to the anatomical structures around the optic disc. The resulting image synthesizer is able to generate realistic (cropped) retinal images, and subsequently, the glaucoma classifier is able to classify them into glaucomatous and normal with high accuracy (AUC = 0.9017). The obtained retinal image synthesizer and the glaucoma classifier could then be used to generate an unlimited number of cropped retinal images with glaucoma labels.

Prostate cancer is one of the types of cancer with the highest incidence in humans. In particular, prostate cancer is the main cause of death from cancer in men over 70 years of age. The automatic analysis of histological images is nowadays a key factor for helping doctors in the diagnosis task. In this paper, we present granulometries as a novel image descriptor to identify abnormal patterns in the prostatic tissue. The morphological alteration suffered by the main structures of pathological glands are registered by the proposed descriptor and achieved in a feature vector. A committee of SVM classifiers is trained making use of the extracted information with the aim of discriminating between healthy and pathological tissue. The performance of the proposed image descriptor is validated in 45 images provided by the Hospital Clínico of Valencia. Accuracy, sensitivity, specificity and AUC values higher than 0.95±0.02 demonstrate the effectiveness of the method.

Deep Neural Networks (DNN) have become a powerful, widely used, and successful mechanism to solve problems of different nature and varied complexity. Their ability to build models adapted to complex non-linear problems, have made them a technique widely applied and studied. One of the fields where this technique is currently being applied is in the malware classification problem. The malware classification problem has an increasing complexity, due to the growing number of features needed to represent the behaviour of the application as exhaustively as possible. Although other classification methods, as those based on SVM, have been traditionally used, the DNN pose a promising tool in this field. However, the parameters and architecture setting of these DNNs present a serious restriction, due to the necessary time to find the most appropriate configuration. This paper proposes a new genetic algorithm designed to evolve the parameters, and the architecture, of a DNN with the goal of maximising the malware classification accuracy, and minimizing the complexity of the model. This model is tested against a dataset of malware samples, which are represented using a set of static features, so the DNN has been trained to perform a static malware classification task. The experiments carried out using this dataset show that the genetic algorithm is able to select the parameters and the DNN architecture settings, achieving a 91% accuracy.

The high variability in fundus image databases is an important limiting drawback for detecting some retinal pathologies automatically. Age, human retinal pigmentation or lighting conditions affects in the colour of the acquired images. In this paper a colour-normalization method is presented as an initial pre-processing step in order to reduce the heterogeneity of retinal databases. The proposed method is based on geometric transformations applied to the chromaticity diagram of a target image taking into account a reference image. With the aim of quantifying the effect of the proposed colour normalization, a bright lesion detection from pathological images is carried out. A home-made system based on texture analysis and Support Vector Machine classification is used for this purpose. An improvement around a three percent in the detection accuracy demonstrates the importance of a retinal image colour pre-processing before any specific analysis.

This paper addresses the functional localization of intra-patient images of the brain. Functional images of the brain (fMRI and PET) provide information about brain function and metabolism whereas anatomical images (MRI and CT) supply the localization of structures with high spatial resolution. The goal is to find the geometric correspondence between functional and anatomical images in order to complement and fuse the information provided by each imaging modality. The proposed approach is based on a variational formulation of the image registration problem in the frequency domain. It has been implemented as a C/C[Formula: see text] library which is invoked from a GUI. This interface is routinely used in the clinical setting by physicians for research purposes (Inscanner, Alicante, Spain), and may be used as well for diagnosis and surgical planning. The registration of anatomic and functional intra-patient images of the brain makes it possible to obtain a geometric correspondence which allows for the localization of the functional processes that occur in the brain. Through 18 clinical experiments, it has been demonstrated how the proposed approach outperforms popular state-of-the-art registration methods in terms of efficiency, information theory-based measures (such as mutual information) and actual registration error (distance in space of corresponding landmarks).

This paper investigates discrimination capabilities in the texture of fundus images to differentiate between pathological and healthy images. For this purpose, the performance of local binary patterns (LBP) as a texture descriptor for retinal images has been explored and compared with other descriptors such as LBP filtering and local phase quantization. The goal is to distinguish between diabetic retinopathy (DR), age-related macular degeneration (AMD), and normal fundus images analyzing the texture of the retina background and avoiding a previous lesion segmentation stage. Five experiments (separating DR from normal, AMD from normal, pathological from normal, DR from AMD, and the three different classes) were designed and validated with the proposed procedure obtaining promising results. For each experiment, several classifiers were tested. An average sensitivity and specificity higher than 0.86 in all the cases and almost of 1 and 0.99, respectively, for AMD detection were achieved. These results suggest that the method presented in this paper is a robust algorithm for describing retina texture and can be useful in a diagnosis aid system for retinal disease screening.

This work focuses on differentiating between pathological and healthy fundus images. The goal is to distinguish between diabetic retinopathy (DR), age-related macular degeneration (AMD) and normal images by analysing the texture of the retina background. Local Binary Patterns (LBP) are used as texture descriptors. The two class problems DR vs. normal and AMD vs. normal, as well as the three class problem of DR, AMD, and normal, have been tested and have obtained promising results. An average sensitivity and specificity higher than 0.86 in all the cases and almost of 0.96 for AMD detection were achieved with a random forest classifier. These results suggest that LBP is a robust texture descriptor for retinal images and the method proposed in this paper, analysing the retina background directly and avoiding difficult lesion segmentation, can be useful for diagnostic aid.

In this paper, a probability density function of object contours based on the stochastic watershed transform is carried out. The watershed transform produces an over-segmentation of the image due to noise, illumination problems, low contrast, etc., because each regional minimum of the image gives place to a region in the output image. To solve this problem, the efforts are focused on the definition of markers to impose new minima in the image, and enhancing the gradient image. The stochastic watershed performs a probability density function (pdf) of the object contours based on a MonteCarlo simulation of random markers. A variation of the method for defining this pdf based on regional regularization of the image is carried out. The objective is to obtain a pdf of the object contours with less noise and better contrast than that produced by the stochastic watershed to use it as a new gradient image for segmentation purposes.

The goal of the software proposed in this paper is to assist ophthalmologists in diagnosis and disease prevention, helping them to determine cardiovascular risk or other diseases where the vessels can be altered, as well as to monitor the pathology progression and response to different treatments. The performance of the tool has been evaluated by means of a double-blind study where its sensitivity, specificity, and reproducibility to discriminate between health fundus (without cardiovascular risk) and hypertensive patients has been calculated in contrast to an expert ophthalmologist opinion obtained through a visual inspection of the fundus image. An improvement of almost 20% has been achieved comparing the system results with the clinical visual classification.

The algorithm proposed in this paper allows to automatically segment the optic disc from a fundus image. The goal is to facilitate the early detection of certain pathologies and to fully automate the process so as to avoid specialist intervention. The method proposed for the extraction of the optic disc contour is mainly based on mathematical morphology along with principal component analysis (PCA). It makes use of different operations such as generalized distance function (GDF), a variant of the watershed transformation, the stochastic watershed, and geodesic transformations. The input of the segmentation method is obtained through PCA. The purpose of using PCA is to achieve the grey-scale image that better represents the original RGB image. The implemented algorithm has been validated on five public databases obtaining promising results. The average values obtained (a Jaccard's and Dice's coefficients of 0.8200 and 0.8932, respectively, an accuracy of 0.9947, and a true positive and false positive fractions of 0.9275 and 0.0036) demonstrate that this method is a robust tool for the automatic segmentation of the optic disc. Moreover, it is fairly reliable since it works properly on databases with a large degree of variability and improves the results of other state-of-the-art methods.

The impact of PET on radiation therapy is held back by poor methods of defining functional volumes of interest. Many new software tools are being proposed for contouring target volumes but the different approaches are not adequately compared and their accuracy is poorly evaluated due to the illdefinition of ground truth. This paper compares the largest cohort to date of established, emerging and proposed PET contouring methods, in terms of accuracy and variability. We emphasise spatial accuracy and present a new metric that addresses the lack of unique ground truth. 30 methods are used at 13 different institutions to contour functional VOIs in clinical PET/CT and a custom-built PET phantom representing typical problems in image guided radiotherapy. Contouring methods are grouped according to algorithmic type, level of interactivity and how they exploit structural information in hybrid images. Experiments reveal benefits of high levels of user interaction, as well as simultaneous visualisation of CT images and PET gradients to guide interactive procedures. Method-wise evaluation identifies the danger of over-automation and the value of prior knowledge built into an algorithm.

This paper presents an algorithm for a 3D segmentation of the aorta artery in magnetic resonance images (MRI). The purpose is to project the 3D segmented aorta in the patient's abdomen with an augmented reality (AR) system to help the surgeon in laparoscopic interventions. In order to obtain accurate results in the segmentation process a marker-controlled watershed algorithm is used. Since this method requires a robust gradient image and two marker sets, a preprocessing step is carried out in each image. The algorithm is automatic and the results are promising with a Jaccard coefficient (JC) of 0.8107 ± 0.0228.

This paper presents a video-based approach to traffic analysis and monitoring in night light conditions. In this kind of scenarios the headlights of the vehicles are the main features of the image taken from an urban or inter-urban traffic camera. The body of the vehicles is very low contrasted and many of the algorithms used in day-time decrease their performance. In our algorithm, we detect car headlights, and using this information, we obtain the three main magnitudes used in traffic monitoring: number of vehicles per time unit, i.e. intensity, mean speed, and occupancy. Extensive evaluations show that the system exhibits an excellent performance with real-time video sequences from cameras of the Traffic Authority of the city of Valencia, Spain.

The presence of metal artifacts in dental CT prevents the correct exploration and planning of dental interventions. This paper addresses a new paradigm in metal artifact reduction that uses the backprojected data available in the DICOM files. The method, based on variational image registration and morphological lambda reconstruction, enhances the image quality using not only the information of the artifacted image (horizontal approach) but also the information of adjoining images (vertical approach). Some preliminary results involving different CT scanners and patients are presented and discussed.

This paper presents a new content-based method for real-time keyframe selection in H.264 low delay video coding. Based on dynamic thresholds, it is aimed at improving compression efficiency. It has been trained and tested with real sequences taken from movies and commercials, with very different motion and complexity characteristics, and from low to high bitrates, number of frames per second and formats, with a total of 70000 frames and more than 1500 candidate keyframes. Results are presented achieving up to 2.32 dB average peak signal-to-noise ratio improvement, higher than recent methods in the literature, and even processing time gains, preserving the real-time and low delay conditions in real-time coders.

This paper presents a robust motion-detector video sensor. It is intended to operate in surveillance applications for long periods of time with time-varying noise level. It makes use of the fact that whenever there is no motion a similarity measure between frames tends to have similar values.

Deals with an application of image sequence analysis. In particular, it addresses the problem of determining the number of people who get into and out of a train carriage when it is crowded, and background and/or illumination changes. The proposed system analyzes image sequences and processes them using an algorithm based on the use of several morphological tools, which are presented in detail in the paper.

This paper deals with techniques to detect abrupt scene transitions when random brightness variations (flicker) are present. This is normally the case when trying to restore or index old films. The application of conventional techniques in this situation tends to produce a large number of false positive detection of cuts. The paper is intently restricted to techniques which require low computation (no motion estimation).

This paper deals with the reduction of flicker in old films. This artifact appears as global, quick and random variations of the luminance and contrast between consecutive frames of a sequence. Initially, we present a method based on the correction of mean and variance parameters of the sequence. However, although this method reduces mean and variance variation between frames, it does not yield good visual results. Finally, an algorithm based on the histogram matching is proposed. This method provides much better visual results.

The paper deals with an application of image sequence analysis. In particular, it addresses the problem of determining the number of people who get into and out of a train carriage when it is crowded and background and/or illumination might change. The proposed system analyses image sequences and processes them using an algorithm based on the use of several morphological tools and optical flow motion estimation.