Analysis of objective quality metrics in computed tomography images affected by metal artifacts.

Image quality (IQ) assessment plays an important role in the medical world. New methods to evaluate image quality have been developed, but their application in the context of computer tomography is yet limited. In this paper the performance of fifteen well-known full reference (FR) IQ metrics is compared with human judgment using images affected by metal artifacts and processed with metal artifact reduction methods from a phantom. Five region of interest with different sizes were selected. IQ was evaluated by seven experienced radiologists completely blinded to the information. To measure the correlation between FR-IQ, and the score assigned by radiologists non-parametric Spearman rank-order correlation coefficient and Kendall's Rank-order Correlation coefficient were used; so as root mean square error and the mean absolute error to measure the prediction accuracy. Cohen's kappa was employed with the purpose of assessing inter-observer agreement. The metrics GMSD, IWMSE, IWPSNR, WSNR and OSS-PSNR were the best ranked. Inter-observer agreement was between 0.596 and 0.954, with p<0.001 in all study. The objective scores predicted by these methods correlate consistently with the subjective evaluations. The application of this metrics will make possible a better evaluation of metal artifact reduction algorithms in future works.

New patch-based strategy for COVID-19 automatic identification using chest x-ray images.

Purpose: The development of a robust model for automatic identification of COVID-19 based on chest x-rays has been a widely addressed topic over the last couple of years; however, the scarcity of good quality images sets, and their limited size, have proven to be an important obstacle to obtain reliable models. In fact, models proposed so far have suffered from over-fitting erroneous features instead of learning lung features, a phenomenon known as shortcut learning. In this research, a new image classification methodology is proposed that attempts to mitigate this problem. Methods: To this end, annotation by expert radiologists of a set of images was performed. The lung region was then segmented and a new classification strategy based on a patch partitioning that improves the resolution of the convolution neural network is proposed. In addition, a set of native images, used as an external evaluation set, is released. Results: The best results were obtained for the 6-patch splitting variant with 0.887 accuracy, 0.85 recall and 0.848 F1score on the external validation set. Conclusion: The results show that the proposed new strategy maintains similar values between internal and external validation, which gives our model generalization power, making it available for use in hospital settings. Supplementary Information: The online version contains supplementary material available at 10.1007/s12553-022-00704-4.

Inpainting-filtering for metal artifact reduction (IMIF-MAR) in computed tomography.

The reduction of metal artifacts remains a challenge in computed tomography because they decrease image quality, and consequently might affect the medical diagnosis. The objective of this study is to present a novel method to correct metal artifacts based solely on the CT-slices. The proposed method consists of four steps. First, metal implants in the original CT-slice are segmented using an entropy based method, producing a metal image. Second, a prior image is acquired using three transformations: Gaussian filter, Parisotto and Schoenlieb inpainting method with the Mumford-Shah image model and L0 Gradient Minimization method (L0GM). Next, based on the projections from the original CT-slice, prior image and metal image, the sinogram is corrected in the traces affected by metal in the process called normalization and denormalization. Finally, the reconstructed image is obtained by FBP and a Nonlocal Means (NLM) filtering. The efficacy of the algorithm is evaluated by comparing five image quality metrics of the images and by inspecting regions of interest (ROI). Phantom data as well as clinical datasets are included. The proposed method is compared with three established metal artifact reduction (MAR) methods. The results from a phantom and clinical dataset show the visible reduction of artifacts. The conclusion is that IMIF-MAR method can reduce streak metal artifacts effectively and avoid new artifacts around metal implants, while preserving the anatomical structures. Considering both clinical and phantom studies, the proposed MAR algorithm improves the quality of clinical images affected by metal artifacts, and could be integrated in clinical setting.

Current limitations to identify covid-19 using artificial intelligence with chest x-ray imaging (part ii). The shortcut learning problem.

Since the outbreak of the COVID-19 pandemic, computer vision researchers have been working on automatic identification of this disease using radiological images. The results achieved by automatic classification methods far exceed those of human specialists, with sensitivity as high as 100% being reported. However, prestigious radiology societies have stated that the use of this type of imaging alone is not recommended as a diagnostic method. According to some experts the patterns presented in these images are unspecific and subtle, overlapping with other viral pneumonias. This report seeks to evaluate the analysis the robustness and generalizability of different approaches using artificial intelligence, deep learning and computer vision to identify COVID-19 using chest X-rays images. We also seek to alert researchers and reviewers to the issue of "shortcut learning". Recommendations are presented to identify whether COVID-19 automatic classification models are being affected by shortcut learning. Firstly, papers using explainable artificial intelligence methods are reviewed. The results of applying external validation sets are evaluated to determine the generalizability of these methods. Finally, studies that apply traditional computer vision methods to perform the same task are considered. It is evident that using the whole chest X-Ray image or the bounding box of the lungs, the image regions that contribute most to the classification appear outside of the lung region, something that is not likely possible. In addition, although the investigations that evaluated their models on data sets external to the training set, the effectiveness of these models decreased significantly, it may provide a more realistic representation as how the model will perform in the clinic. The results indicate that, so far, the existing models often involve shortcut learning, which makes their use less appropriate in the clinical setting.

Current limitations to identify COVID-19 using artificial intelligence with chest X-ray imaging.

The scientific community has joined forces to mitigate the scope of the current COVID-19 pandemic. The early identification of the disease, as well as the evaluation of its evolution is a primary task for the timely application of medical protocols. The use of medical images of the chest provides valuable information to specialists. Specifically, chest X-ray images have been the focus of many investigations that apply artificial intelligence techniques for the automatic classification of this disease. The results achieved to date on the subject are promising. However, some results of these investigations contain errors that must be corrected to obtain appropriate models for clinical use. This research discusses some of the problems found in the current scientific literature on the application of artificial intelligence techniques in the automatic classification of COVID-19. It is evident that in most of the reviewed works an incorrect evaluation protocol is applied, which leads to overestimating the results.

New morphological features based on the Sholl analysis for automatic classification of traced neurons.

BACKGROUND: This article addresses the automatic classification of reconstructed neurons through their morphological features. The purpose was to extend the capabilities of the L-Measure software. METHODS: New morphological features were developed, based on modifications of the conventional Sholl analysis. The lengths of the compartments, as well as their volumes, were added to the features used in the classical analysis in order to improve the results during automatic neuron classification. FSM were used to obtain subsets of lower cardinality from the full feature sets and the usefulness of these subsets was tested through their use in supervised classification tasks. The study was based on two types of neurons belonging to mice: pyramidal and GABAergic interneurons. Furthermore, a set of pyramidal neurons belonging to Later 4 and Layer 5 was analyzed. RESULTS: RF classifier shown the best performance combined with a Wrapper method.U-WNAD set allowed to obtain higher values than WN, A and D in all cases and better results than LM for the filters and wrappers FSM. U-LM-WNAD set, led to the highest AUC values for all the FSM studied. Similar results for different regions of cortex were obtained. Comparison with Existing Methods The new features exhibited high discriminatory power with which the values of AUC and Acc obtained in the experiments exceeded those obtained using only the features provided by L-Measure. CONCLUSIONS: The highest values of AUC and Acc were obtained from the sets U-WNAD and U-LM-WNAD, evidencing the discriminatory power of the new proposed features.

New Features for Neuron Classification.

This paper addresses the problem of obtaining new neuron features capable of improving results of neuron classification. Most studies on neuron classification using morphological features have been based on Euclidean geometry. Here three one-dimensional (1D) time series are derived from the three-dimensional (3D) structure of neuron instead, and afterwards a spatial time series is finally constructed from which the features are calculated. Digitally reconstructed neurons were separated into control and pathological sets, which are related to three categories of alterations caused by epilepsy, Alzheimer's disease (long and local projections), and ischemia. These neuron sets were then subjected to supervised classification and the results were compared considering three sets of features: morphological, features obtained from the time series and a combination of both. The best results were obtained using features from the time series, which outperformed the classification using only morphological features, showing higher correct classification rates with differences of 5.15, 3.75, 5.33% for epilepsy and Alzheimer's disease (long and local projections) respectively. The morphological features were better for the ischemia set with a difference of 3.05%. Features like variance, Spearman auto-correlation, partial auto-correlation, mutual information, local minima and maxima, all related to the time series, exhibited the best performance. Also we compared different evaluators, among which ReliefF was the best ranked.

Segmentación del hígado en imágenes de tomografía computarizada

RESUMEN Fundamento: la segmentación del hígado utilizando datos de tomografía computarizada es el primer paso para el diagnóstico de enfermedades hepáticas. Actualmente la segmentación de estructuras y órganos, basado en imágenes, que se realiza en los hospitales del país, dista de tener los niveles de precisión que se obtienen de los modernos sistemas 3D, por lo que se requiere buscar alternativas viables utilizando el PDI sobre ordenador. Objetivo: determinar una variante eficaz y eficiente desde el punto de vista computacional en condiciones de rutina hospitalaria, para la segmentación de imágenes hepáticas con fines clínicos. Métodos: se compararon dos métodos modernos de segmentación (Graph Cut y EM/MPM) aplicándolos sobre imágenes de tomografía de hígado. Se realizó un análisis evaluativo y estadístico de los resultados obtenidos en la segmentación de las imágenes a partir de los coeficientes de Dice, Vinet y Jaccard. Resultados: con el método Graph Cut, en todos los casos, se segmentó la región deseada, incluso cuando la calidad de las imágenes era baja, se observó gran similitud entre la imagen segmentada y la máscara de referencia. El nivel de detalles visuales es bueno y la reproducción de bordes permanece fiel a la máscara de referencia. La segmentación de las imágenes por el método de EM/MPM, no siempre fue satisfactoria. Conclusiones: el método de segmentación Graph Cut obtuvo mayor precisión para segmentar imágenes de hígado.

ABSTRACT Background: liver segmentation using computed tomography data is the first step for the diagnosis of liver diseases. Currently, the segmentation of structures and organs, based on images, which is carried out in the country's hospitals, is far from having the levels of precision obtained from modern 3D systems, it is necessary to search for viable alternatives using the PDI on a computer. Objective: to determine an effective and efficient variant from the computational point of view in routine hospital conditions, for the segmentation of liver images for clinical purposes. Methods: Two modern segmentation methods (Graph Cut and EM/MPM) were compared by applying them to liver tomography images. An evaluative and statistical analysis of the results obtained in the segmentation of the images from the Dice, Vinet and Jaccard coefficients was carried out. Results: with the Graph Cut method, in all cases, the desired region was segmented, even when the quality of the images was low, great similarity was observed between the segmented image and the reference mask. The level of visual detail is good, and edge reproduction remains true to the reference skin. Image segmentation by the EM/MPM method was not always satisfactory. Conclusions: the Graph Cut segmentation method obtained greater precision to segment liver images.

Conteo de bacterias y levaduras en imágenes digitales

RESUMEN Fundamento: en los laboratorios de microbiología, la identificación y conteo de microorganismos es un procedimiento habitual. Aunque existen en el mercado equipos que posibilitan su realización de manera automática o semiautomática, son muy costosos, por lo cual esta tarea, difícil e irritante para los ojos, la siguen realizando los expertos de manera tradicional mediante la observación de las muestras en los microscopios, con la consiguiente variabilidad entre ellos. Objetivo: proponer un nuevo método para el conteo de bacterias y levaduras en imágenes digitales, bajo diferentes magnificaciones, tomadas a bioproductos de origen microbiano obtenidos por fermentación. Métodos: el sensor empleado para la toma de imágenes de las muestras fue una cámara digital modelo HDCE-X, con un sensor CMOS de ½", con una resolución de 2592 píxeles por 1944 píxeles (5 Mp). Se emplearon dos tipos de magnificaciones: magnificación 40x (PL40, 0.65 apertura numérica and 0.17 de distancia de trabajo) y magnificación 100x (HI plan 100/1.25 con inmersión de aceite). El método propuesto se basa en técnicas de procesamiento digital de imágenes, utilizando herramientas como la detección de contornos, operaciones morfológicas y análisis estadístico, y fue desarrollado en lenguaje Python con empleo de la biblioteca OpenCV. Resultados: la detección y conteo de bacterias se logró con una exactitud y precisión aceptable, en ambos casos por encima de 0,95; no en el caso de las levaduras cuya exactitud y precisión fueron menores, alrededor de 0,78 y 0,86 respectivamente. Se proponen flujos de trabajo basados en técnicas de procesamiento digital de imágenes, fundamentalmente en detección de contornos, operaciones morfológicas y análisis estadístico. Conclusiones: el método posee una efectividad aceptable para el contexto y depende de las características que presenten las imágenes.

ABSTRACT Background: In microbiology laboratories, the identification and counting of microorganisms is a common procedure; and although there is a variety of equipment on the market that possibility to carry out these processes automatically or semi-automatically, it is usually expensive to many laboratories. These are some of the reasons why this arduous and difficult task is still performed in many laboratories by experts in the traditional way, through the observation of samples in microscope, consuming a great time and having variations in the results between experts. Objective: The present work aims to propose a new method for counting bacteria and yeasts in digital images, taken under different magnifications, of microbial bioproducts obtained by fermentation. Methods: The sensor used to take images of the samples was a digital camera model HDCE-X, with a ½" CMOS sensor, with a resolution of 2592 pixels by 1944 pixels (5 Mp). Two types of magnifications were used: 40x magnification (PL40, 0.65 numerical aperture and 0.17 working distance) and 100x magnification (HI plan 100/1.25 with oil immersion). The proposed method is based on digital image processing technics, using tools as contour detection, morphological operations and statistical analysis, and was developed in Python language using the OpenCV library. The work also presents a comparison with the results obtained using ImageJ software for the same purpose. Results: the detection and count of bacteria was achieved with an acceptable accuracy and precision, in both cases above 0.95; not in the case of yeasts whose accuracy and precision was lower, around 0.78 for accuracy and 0.86 for precision. Workflows based on digital image processing techniques are proposed, using tools as contour detection, morphological operations and statistical analysis. Conclusions: the method has an acceptable effectiveness for the context and depends on the characteristics presented by the images.

Methods for Reducing Metal Artifacts in Computerized Tomography / Métodos para reducir los artefactos metálicos en la tomografía computarizada

Abstract Metal artifacts are common in clinical images. Many methods for artifact reduction have been published to overcome this problem. In this work, animage smoothing method for artifact reduction (ISMAR) is proposed for image quality improvement in patients with hip prosthesis and dental fillings, which caused metal artifacts. ISMAR was evaluated and compared with three well-known methods for metal artifact reduction (linear interpolation (LI), normalized metal artifact reduction (NMAR) and frequency split metal artifact reduction (FSMAR)). The new method is based on edge-preserving smoothing via L0 Gradient Minimization filter. Image quality was evaluated by two experienced radiologists completely blinded to the information about if the image was processed or not to suppress the artifacts. They graded image quality in a five points-scale, where zero is an index of clear artifact presence, and five, a whole artifact suppression. The new method had the best results and it was statistically significant respect to the other tested methods (p < 0.05). This new method has a better performance in artifact suppression and tissue feature preservation.

Resumen Los artefactos metálicos son comunes en las imágenes clínicas. Muchos métodos para la reducción de los artefactos han sido publicados para superar este problema. En el presente trabajo, un método de suavizado de imágenes para la reducción de artefactos (ISMAR) es propuesto para mejorar la calidad de la imagen en pacientes con prótesis de cadera y empastes dentales, los cuales causaron artefactos metálicos. ISMAR fue evaluado y comparado con otros tres métodos reconocidos por su desempeño en la reducción de los artefactos metálicos (Interpolación lineal (LI), reducción de artefactos de metal normalizados (NMAR) y reducción de artefactos de metal divididos en frecuencia (FSMAR)). El nuevo método se basa en el suavizado y conservación de bordes, utilizando para ello el filtro de minimización de gradiente L0. La calidad de la imagen fue evaluada por dos radiólogos experimentados completamente ciegos a la información sobre si la imagen fue procesada o no para suprimir los artefactos. Ellos calificaron la calidad de la imagen en una escala de cinco puntos, donde el cero indica la presencia de artefactos, y el cinco, una supresión total de los artefactos. El nuevo método tuvo los mejores resultados y fue estadísticamente significativo con respecto a los otros métodos probados (p < 0.05). Este nuevo método tiene un mejor rendimiento en la supresión de artefactos y en la conservación de las características de los tejidos.