Medical Image Analysis

Breast cancer ranks as the most prevalent form of cancer diagnosed in women. Diagnosis faces several challenges, such as changes in the size, shape, and appearance of the breast, dense breast tissue, and lumps or thickening, especially if present in only one breast. The major challenge in the deep learning (DL) diagnosis of breast cancer is its non-uniform shape, size, and position, particularly with malignant tumors. Researchers strive through computer-aided diagnosis (CAD) systems and other methods to assist in detecting and classifying tumor types. This work proposes a DL system for analyzing medical images that improves the accuracy of breast cancer detection and classification from ultrasound (US) images. It reaches an accuracy of 99.29%, exceeding previous work. First, image processing is applied to enhance the quality of input images. Second, image segmentation is performed using the U-Net architecture. Third, many features are extracted using Mobilenet. Finally, classification is performed using visual geometry group 16 (VGG16). The accuracy of detection and classification using the proposed system was evaluated.

The Western Ghat (Sahyadri, A mountain range along the western coast of the Indian peninsula) is a hotspot for biodiversity. The region is also subject to land use and land cover (LULC) changes. If areas are to be properly planned and urbanized, then it is important to know what kinds of LULC exist thereon. As the Western Ghats are rich both environmentally and socioeconomically , monitoring land cover changes within its domain becomes crucial. This study adopts multiclass semantic segmentation within Western Ghats focusing more on Idukki (Western Ghats) of Kerala. We used high-resolution spatial images obtained from Google satellites through QGIS. The dataset has 1564 pre-processed images and 1564 corresponding masks. We have studied U-Net, FCN, SegNet, PSPNet, Mask R-CNN and DeepLabV3+ and used the DeepLabV3+ Model, a convolutional neural network architecture, to classify the satellite images into four classes: 1. Settlements 2. Forests 3. Water bodies 4. Vegetated area. The segmentation provides insights into the spatial distribution. This research aims at advancing knowledge about LULC dynamics within the Western Ghats. Scope for conservation planning, environmental policy formulation and socioeconomic development in the region are possibilities that can also come out as a result from these findings.

Skin cancer is one of the most prevalent cancers worldwide, with early diagnosis being critical for improving survival rates. Dermoscopy, a non-invasive imaging tool, is widely used for identifying pigmented skin lesions. However, its accuracy is heavily dependent on expert interpretation, which introduces variability and limits accessibility in resourceconstrained settings. This highlighted the need for automated solutions to enhance diagnostic consistency and aid in early detection. The aim of this study was to develop a refined machine-learning framework for classifying pigmented skin lesions using dermoscopy images. We employed an enhanced Inception-V3 model, a state-of-the-art convolutional neural network, integrated with a simplified soft-attention mechanism, advanced data augmentation techniques, and Bayesian hyperparameter tuning. These innovations improved the model's ability to accurately focus on and identify relevant lesion features, marking a significant advancement in the field. Using the ISIC-2019 dataset, a publicly available resource containing dermoscopy images classified into eight diagnostic categories, we implemented preprocessing steps such as resizing, cleaning, and data balancing. Additionally, ImageNet transfer learning and Bayesian optimization were applied to refine the model. The inclusion of a soft-attention mechanism further enhanced the model's capacity to identify patterns within lesion images. Our model exhibited outstanding performance on the ISIC-2019 dataset, achieving a sensitivity of 98.5%, specificity of 99.62%, precision of 97.42%, accuracy of 97.38%, an F1 score of 97.34%, and an area under the curve (AUC) of 0.99. These metrics underscored the model's superior capability in accurate and reliable classification of pigmented skin lesions, surpassing current benchmarks and demonstrating significant advancements over existing methodologies.

Quantum-Enhanced Federated Learning for Real-Time Medical Image Analysis with Ethical AI Governance is a nascent approach combining the principles of federated learning and quantum computing to transform the medical image analysis sector with due regard to the most critical ethical principles. Reflecting on the definition, federated learning is an approach that allows several institutions to jointly train machine learning models without the need to share any patient's data. Therefore, FL is the industry-institution strategic tool that allows for a higher level of privacy security in healthcare. On the other hand, by embracing quantum computing, the developed approach includes more sophisticated computational skills, permitting improved data performance and enhanced model accurateness, a pivotal factor for real-time medical diagnosis. Consequently, the significance of the approach lies in the potential to boost the medical image analysis sector. Quantum-enhanced FL will help comply with the most rigid ethical requirements by involving decentralized data performance and exploiting various datasets characteristic of healthcare

Introduction: Augmented Reality (AR) is revolutionizing surgical practice through improved visualization, accuracy, and training. This paper examines approaches to integrating AR and its potential future applications. Methods: Literature review and expert critique were used to determine AR effectiveness, challenges, and how to implement it in surgery. Results: Evidence suggests that AR enhances surgical precision and training. The adoption is impeded by costs, technical restrictions, and specialist training requirements. Future development will involve the use of AI and robotics to develop more advanced capabilities. Conclusion: Although useful, AR's application in surgery demands overcoming financial and technical constraints. Strategic planning and additional research must be conducted to achieve its greatest potential in practice.

Empfohlene Zitierung / Suggested Citation: Burri, R. (2003). Digitalisieren, disziplinieren: soziotechnische Anatomie und die Konstitution des Körpers in medizinischen Bildgebungsverfahren (TUTS Working Papers, 3-2003). Berlin: Technische Universität Berlin, Fak. VI Planen, Bauen, Umwelt, Institut für Soziologie Fachgebiet Techniksoziologie. https://nbn-resolving.org/ urn:nbn:de:0168-ssoar-11594

Effective diagnosis and therapy of brain tumors depend on their identification and classification. Using the Brain Tumor dataset, this study makes use of sophisticated transfer learning models and deep convolutional neural networks (DCNNs). When models like DCNN, ResNet152, EfficientNetB2, Xception, and NasNetMobile were tested, an ensemble of Xception and NasNetMobile produced the best accuracy (98.1%). Grade 0 (no malignancy) to Grade III (big tumor) were the four grades into which tumors were divided. With a mean average precision (mAP) of 78.9%, YOLOv9 fared better than other models for anomaly detection. A Flask-based interactive interface was created for safe and easy access in order to improve usage.

Circular data, i.e., data in the form of 'natural' directions or angles are very common in a number of dierent areas such as biological, meteorological, geological, and political sciences. Clustering circular data is not an easy task due to the circular geometry of the data space. Some clustering approaches, such as the spherical k-means, use the cosine distance instead of the euclidean distance in order to measure the dierence between points. In this paper, we propose a variation of the randomized gravitational clustering algorithm in order to deal with circular data. Basically, we use the cosine distance, we modify the gravitational law in order to use the cosine distance and we use geodesics ('straight' lines in curved spaces) in order to move points according to the gravitational dynamic. Our initial experiments indicate that the spherical gravitational clustering algorithm is able to nd clusters in noisy circular data.

We discuss the potentially misleading effect of software techniques for elaborating low-contrast images. In particular, we present the example of the stains embedded into one of the most studied archaeological objects in history, the Shroud of Turin. We show for the first time that image processing of both old and recent photographs of the Shroud may lead some researchers to perceive inscriptions and patterns that do not actually exist, confirming that there is a narrow boundary between image enhancement and manipulation.

Breast cancer is one of the leading causes of cancer death in women around the world. In order to reduce the death rate, the tumors have to be detected at the early stage. CAD system are considered to be the best technique for screening. This paper presents a survey of various staging techniques used for early and accurate detection of tumor, which aids in early medication by reducing false positive and false negative discrepancies. The several stages involved in the detection of tumor are pre-processing, segmentation, feature extraction and classification lastly staging of cancer, if confirmed for malignancy. Pre-processing stage improves the quality of the image by removing artifacts, and enhancing the image. Segmentation deals with removal of the pectoral muscle and segmenting the tumor part. Classification is used to classify the detected tumor as Malignant or Benign. The tumors identified as Malignant are then processed using various parameters like size, shape, energy levels and several other features for staging.

Objective: The objective of this research is to automate the calculation of differential blood count in blood smear photomicrographsusing image quantization.Material and Methods: A series of image processing steps were used for the detection of White Blood Cells (WBCs),Red Blood Cells (RBCs) and platelets as: image acquisition, separating the channels of RGB and applying wiener filteron each channel for smoothing the image. The purpose is to enhance the visual interpretation of the image, recombiningthe channels and applying the quantization over the wiener output.Results: The accuracy of this technique is very close to that of the hematologists’ manual calculation. It was 85% forRed blood cells and 98% for White blood cells.Conclusion: This proposed technique gives precise results under varied luminance conditions such as darkness,brightness and low contrast images, it gives reliable results for all the images in the image sets having different qualityof images

Automated ventricle volume estimation (AVVE) on cardiac magnetic resonance (CMR) images is very important for clinical cardiac disease diagnosis. However, current AVVE methods ignore the error correction for the estimated volume. This results in clinically intolerable ventricle volume estimation error and further leads to wrong ejection fraction (EF) assessment, which significantly limits the application potential of AVVE methods. The objective of this paper is to address this problem with AVVE and further make it more clinically applicable. We proposed a dynamically constructed network to achieve accurate AVVE. First, we introduced a novel dynamically constructed deep learning framework, that evolves a single model into a bi-model volume estimation network. In this way, the EF correlation can be built directly based on the bi-model network. Second, we proposed an error correction strategy using dynamically created residual nodes, which is based on stochastic configurations with an EF correlation constraint. Finally, we formulated the proposed method into an end-to-end joint optimization framework for accurate ventricle volume estimation with effective error correction. Experiments and comparisons on large-scale cardiac

We proposed a novel efficient method for 3D left ventricle (LV) segmentation on echocardiography, which is important for cardiac disease diagnosis. The proposed method effectively overcame the 3D echocardiography's challenges: high dimensional data, complex anatomical environments, and limited annotation data. First, we proposed a deep atlas network, which integrated LV atlas into

A holistic multitask regression approach was implemented to tackle the limitations of clinical image analysis. Standard practice requires identifying multiple anatomic structures in multiple planes from multiple anatomic regions using multiple modalities. The proposed novel holistic multitask regression network (HMR-Net) formulates organ segmentation as a multitask learning problem. Multitask learning leverages the strength of joint task problem solving from capturing task correlations. HMR-Net performs multitask regression by estimating an organ's class, regional location, and precise contour coordinates. The estimation of each coordinate point also corresponds to another regression task. HMR-Net leverages hierarchical multiscale and fused organ features to handle nonlinear relationships between image appearance and distinct organ properties. Simultaneously, holistic shape information is captured by encoding coordinate correlations. The multitask pipeline enables the capturing of holistic organ information (e.g. class, location, shape) to perform shape regression for medical image segmentation. HMR-Net was validated on eight representative datasets obtained from a total of 222 subjects. A mean average precision and dice score reaching up to 0.81 and 0.93, respectively, was achieved on the representative multiapplication database. The generalized model demonstrates comparable or superior performance compared to state-of-the-art algorithms. The high-performance accuracy demonstrates our model as an effective general framework to perform organ shape regression in

Alzheimer's disease (AD) early diagnosis plays a pivotal role in effective intervention and patient care. With limited and noisy medical imaging datasets, GANs are utilized to generate synthetic brain images, aiding in the augmentation of existing data. The selected classification algorithms are well-established in computer vision and have demonstrated efficacy in image classification tasks. The GAN is employed for data augmentation, creating synthetic images representative of AD-associated features. Subsequently, the augmented dataset is utilized to train and evaluate the performance of multiple classification algorithms, providing a comprehensive analysis of their effectiveness in AD detection. This research contributes to the field of Alzheimer's disease diagnosis by integrating GANs for data augmentation and evaluating the performance of ten diverse classification algorithms, offering insights into their suitability for early detection. In this study, we leverage Generative Adversarial Networks (GANs) for data augmentation in medical imaging, enhancing the quality and diversity of brain images associated with AD.

Motor symptoms greatly diminish the quality of life for those living with Parkinson's disease (PD), a degenerative neurological condition. An early and precise diagnosis is essential for treatment and therapy to be successful. Here, we see a state-of-the-art method for PD prediction based on spiral drawing picture categorization. The preprocessing stage utilizes the Enhanced Residual Noise Elimination Neural Network algorithm, substantially enhancing image quality by effectively reducing unwanted noise. The Improved Mask R-CNN is employed for segmentation, offering superior delineation of relevant structures within the spiral drawings. Feature extraction uses a Convolutional Neural Network (CNN) with AlexNet, which captures intricate patterns and essential details from the images. Finally, classification is performed using IVGG19 Networks, an architecture tailored explicitly for high-precision image classification tasks. Integrating these advanced techniques aims to improve the accuracy of Parkinson's disease prediction, providing a reliable tool for early diagnosis and aiding in better clinical decision-making.

This paper presents a mass preserving image registration algorithm for lung CT images. To account for the local change in lung tissue intensity during the breathing cycle, a tissue appearance model based on the principle of preservation of total lung mass is proposed. This model is incorporated into a standard image registration framework with a composition of a global affine and several free-form B-Spline transformations with increasing grid resolution. The proposed mass preserving registration method is compared to registration using the sum of squared intensity differences as a similarity function on four groups of data: 44 pairs of longitudinal inspiratory chest CT scans with small difference in lung volume; 44 pairs of longitudinal inspiratory chest CT scans with large difference in lung volume; 16 pairs of expiratory and inspiratory CT scans; and 5 pairs of images extracted at end exhale and end inhale phases of 4D-CT images. Registration errors, measured as the average distance between vessel tree centerlines in the matched images, are significantly lower for the proposed mass preserving image registration method in the second, third and fourth group, while there is no statistically significant difference between the two methods in the first group. Target registration error, assessed via a set of manually annotated landmarks in the last group, was significantly smaller for the proposed registration method.

The increased use of image-guided surgery systems during neurosurgery has brought to prominence the inaccuracies of conventional intraoperative navigation systems caused by shape changes such as those due to brain shift. We propose a method to track the deformation of the brain and update preoperative images using intraoperative MR images acquired at different crucial time points during surgery. We use a deformable surface matching algorithm to capture the deformation of boundaries of key structures (cortical surface, ventricles, and tumor) throughout the neurosurgical procedure, and a linear finite element elastic model to infer a volumetric deformation. The boundary data is extracted from intraoperative MR images using a realtime intraoperative segmentation algorithm. The algorithm has been applied to a sequence of intraoperative MR images of the brain exhibiting brain shift and tumor resection. Our results characterize the brain shift after opening of the dura and at the different stages of tumor resection, and brain swelling afterwards. Analysis of the average deformation capture was assessed by comparing landmarks identified manually and the results indicate an accuracy of ¥ §¦ © ¥ ¦ mm (mean © standard deviation) for boundary surface landmarks, of ¥ ¦ © ¥ §¦ mm for landmarks inside the boundary surfaces, and ¦ © ¥ ¦ mm for landmarks in the vicinity of the tumor.

In this paper, we focus on joint regression and classification for Alzheimer's disease diagnosis and propose a new feature selection method by embedding the relational information inherent in the observations into a sparse multi-task learning framework. Specifically, the relational information includes three kinds of relationships (such as feature-feature relation, response-response relation, and sample-sample relation), for preserving three kinds of the similarity, such as for the features, the response variables, and the samples, respectively. To conduct feature selection, we first formulate the objective function by imposing these three relational characteristics along with an ℓ2,1-norm regularization term, and further propose a computationally efficient algorithm to optimize the proposed objective function. With the dimension-reduced data, we train two support vector regression models to predict the clinical scores of ADAS-Cog and MMSE, respectively, and also a support vecto...

Contemporary healthcare techniques
involve maintaining a physical record of
doctor visits, prescriptions, medical
findings, tests, etc. Receiving our medical
reports on WhatsApp, calling our doctor, or
searching for medicines online is the closest
we have come to digitization of healthcare.
Coupled down by the fact that a lot of
individuals know little to know about their
medical history, prescription makes it
challenging to treat them in critical
conditions. Fragmentation of medical
records, leading to misdiagnoses, redundant
tests, and treatment inefficiencies.

This paper presents a system to support medical diagnosis and detection of abnormal lesions by processing endoscopic images. Endoscopic images possess rich information expressed by texture. Texture information can be efficiently extracted from medium scales of the wavelet transform. The set of features proposed in this paper to encode textural information is named color wavelet covariance (CWC). CWC coefficients are based on the covariances of second order textural measures, an optimum subset of them is proposed. The proposed approach is supported by a classifier based on multilayer perceptron network for the characterization of the image regions along the video frames. The whole methodology has been applied on real data containing 6 full endoscopic exams and reached 87% specificity and 97.4% sensitivity.

Instrument segmentation plays a vital role in 3D ultrasound (US) guided cardiac intervention. Efficient and accurate segmentation during the operation is highly desired since it can facilitate the operation, reduce the operational complexity, and therefore improve the outcome. Nevertheless, current image-based instrument segmentation methods are not efficient nor accurate enough for clinical usage. Lately, fully convolutional neural networks (FCNs), including 2D and 3D FCNs, have been used in different volumetric segmentation tasks. However, 2D FCN cannot exploit the 3D contextual information in the volumetric data, while 3D FCN requires high computation cost and a large amount of training data. Moreover, with limited computation resources, 3D FCN is commonly applied with a patch-based strategy, which is therefore not efficient for clinical applications. To address these, we propose a POI-FuseNet, which consists of a patch-of-interest (POI) selector and a FuseNet. The POI selector can efficiently select the interested regions containing the instrument, while FuseNet can make use of 2D and 3D FCN features to hierarchically exploit contextual information. Furthermore, we propose a hybrid loss function, which consists of a contextual loss and a class-balanced focal loss, to improve the segmentation performance of the network. With the collected challenging ex-vivo dataset on RF-ablation catheter, our method achieved a Dice score of 70.5%, superior to the state-of-the-art methods. In addition, based on the pre-trained model from exvivo dataset, our method can be adapted to the in-vivo dataset on guidewire and achieves a Dice score of 66.5% for a different cardiac operation. More crucially, with POI-based strategy, segmentation efficiency is reduced to around 1.3 seconds per volume, which shows the proposed method is promising for clinical use.

On the analysis of the Diffusion-Weighted Magnetic Resonance Images, multi-compartment models overcome the limitations of the well-known Diffusion Tensor model for fitting in vivo brain axonal orientations at voxels with fiber crossings, branching, kissing or bifurcations. Some successful multi-compartment methods are based on diffusion dictionaries. The diffusion dictionary-based methods assume that the observed Magnetic Resonance signal at each voxel is a linear combination of the fixed dictionary elements (dictionary atoms). The atoms are fixed along different orientations and diffusivity profiles. In this work, we present a sparse and adaptive diffusion dictionary method based on the Diffusion Basis Functions Model to estimate in vivo brain axonal fiber populations. Our proposal overcomes the following limitations of the diffusion dictionary-based methods: the limited angular resolution and the fixed shapes for the atom set. We propose to iteratively re-estimate the orientations and the diffusivity profile of the atoms independently at each voxel by using a simplified and easier-to-solve mathematical approach. As a result, we improve the fitting of the Diffusion-Weighted Magnetic Resonance signal. The advantages with respect to the former Diffusion Basis Functions method are demonstrated on the synthetic data-set used on the 2012 HARDI Reconstruction Challenge and in vivo human data. We demonstrate that improvements obtained in the intra-voxel fiber structure estimations benefit brain research allowing to obtain better tractography estimations. Hence, these improvements result in an accurate computation of the brain connectivity patterns.

This paper describes an evaluation framework that allows a standardized and quantitative comparison of IVUS lumen and media segmentation algorithms. This framework has been introduced at the MICCAI 2011 Computing and Visualization for (Intra)Vascular Imaging (CVII) workshop, comparing the results of eight teams that participated. We describe the available data-base comprising of multi-center, multi-vendor and multi-frequency IVUS datasets, their acquisition, the creation of the reference standard and the evaluation measures. The approaches address segmentation of the lumen, the media, or both borders; semi-or fully-automatic operation; and 2-D vs. 3-D methodology. Three performance measures for quantitative analysis have been proposed. The results of the evaluation indicate that segmentation of the vessel lumen and media is possible with an accuracy that is comparable to manual annotation when semi-automatic methods are used, as well as encouraging results can be obtained also in case of fully-automatic segmentation. The analysis performed in this paper also highlights the challenges in IVUS segmentation that remains to be solved.

: CRT (Cardiac Resynchronization Therapy) LBBB (Left bundle branch block) LV (Left ventricle) RV (Right ventricle) EAM (Electroanatomical mapping) CE CMR (Contrast enhanced cardiac magnetic resonance) LAT (Local activation time) CS (Coronary sinus) DT MRI (Diffusion tensor magnetic resonance imaging)

Aiming at the security protection of remote sensing images, a robust watermarking algorithm based on SURF (Speeded Up Robust Features) feature on selective regions is proposed. The algorithm first extracts the SURF feature points of the carrier, and then performs a 5/3 integer wavelet transform on the carrier image to filter out the low-frequency coefficients of the ROI and the intermediate frequency coefficients of the non-interest area (ROB); With sampling pyramid decomposition, the near subband after watermark decomposition is embedded in the lowfrequency subband of the region of interest, and the residual subband is embedded in the intermediate frequency coefficient of the non-interesting region. Experimental data show that the algorithm can resist conventional geometric attacks. The similarity of the watermark is high, and the NC value is kept above 0.89, which has good reversibility and robustness.

Dynamical contrast enhanced (DCE) imaging allows non invasive access to tissue micro-vascularization. It appears as a promising tool to build imaging biomarkers for diagnostic, prognosis or anti-angiogenesis treatment monitoring of cancer. However, quantitative analysis of DCE image sequences suffers from low signal

Dynamical contrast enhanced (DCE) imaging allows non invasive access to tissue micro-vascularization. It appears as a promising tool to build imaging biomarkers for diagnostic, prognosis or anti-angiogenesis treatment monitoring of cancer. However, quantitative analysis of DCE image sequences suffers from low signal

Automated left ventricular (LV) segmentation is crucial for efficient quantification of cardiac function and morphology to aid subsequent management of cardiac pathologies. In this paper, we parameterize the complete (all short axis slices and phases) LV segmentation task in terms of the radial distances between the LV centerpoint and the endo-and epicardial contours in polar space. We then utilize convolutional neural network regression to infer these parameters. Utilizing parameter regression, as opposed to conventional pixel classification, allows the network to inherently reflect domain-specific physical constraints. We have benchmarked our approach primarily against the publicly-available left ventricle segmentation challenge (LVSC) dataset, which consists of 100 training and 100 validation cardiac MRI cases representing a heterogeneous mix of cardiac pathologies and imaging parameters across multiple centers. Our approach attained a 0.77 Jaccard index, which is the highest published overall result in comparison to other automated algorithms. To test general applicability, we also evaluated against the Kaggle Second Annual Data Science Bowl, where the evaluation metric was the indirect clinical measures of LV volume rather than direct myocardial contours. Our approach attained a Continuous Ranked Probability Score (CRPS) of 0.0124, which would have ranked tenth in the original challenge. With this we demonstrate the effectiveness of convolutional neural network regression paired with domain-specific features in clinical segmentation.

Given the enormous number of mammograms performed during last years, computerbased diagnosis of breast cancer turned into a necessity. In particular, the diagnosis of breast masses and their classification currently arouse great interest. Indeed, the complexity of the processed forms and the difficulty encountered in order to discern them require the use of appropriate descriptors. This article is placed in the context of evaluating the results of supervised classification algorithms and their comparison. In this work, we conduct some experiments using the Wisconsin diagnosis Breast Cancer (WDBC) dataset in order to classify the dataset samples to be either benign or malignant. We show that the best results are obtained using our new proposed neighboring Support Vector Classifier (NSVC).

Generative Adversarial Networks are a class of artificial intelligence algorithms that consist of a generator and a discriminator trained simultaneously through adversarial training. GANs have found crucial applications in various fields, including medical imaging. In healthcare, GANs contribute by generating synthetic medical images, enhancing data quality, and aiding in image segmentation, disease detection, and medical image synthesis. Their importance lies in their ability to generate realistic images, facilitating improved diagnostics, research, and training for medical professionals. Understanding its applications, algorithms, current advancements, and challenges is imperative for further advancement in the medical imaging domain. However, no study explores the recent state-of-the-art development of GANs in medical imaging. To overcome this research gap, in this extensive study, we began by exploring the vast array of applications of GANs in medical imaging, scrutinizing them within recent research. We then dive into the prevalent datasets and pre-processing techniques to enhance comprehension. Subsequently, an in-depth discussion of the GAN algorithms, elucidating their respective strengths and limitations, is provided. After that, we meticulously analyzed the results and experimental details of some recent cutting-edge research to obtain a more comprehensive understanding of the current development of GANs in medical imaging. Lastly, we discussed the diverse challenges encountered and future research directions to mitigate these concerns. This systematic review offers a complete overview of GANs in medical imaging, encompassing their application domains, models, state-of-the-art results analysis, challenges, and research directions, serving as a valuable resource for multidisciplinary studies.

Medical imaging is undeniably a very important field in the provision of
effective health care. It allows for early detection, diagnosis, and treatment of different kinds of diseases. On the other hand, preparation of labeled data from the biomedical domain poses difficulties, since it requires heavy privacy considerations and the expert radiologist to perform manual labeling of that data. Due to that reason, efforts to utilize advancements in deep learning algorithms are often hampered. Generating expert annotation of medical images at scale is difficult, costly, and time-consuming. This challenge is even greater in developing countries, especially in Africa, where there is a massive shortage of radiologists to annotate these images.  Another related issue in medical imaging is the scarcity of datasets, particularly in Africa, where there is a severe shortage of digital imaging modalities. This becomes more obvious when studying a particular disease, for instance. Hence, the scarcity of large,
well-labelled datasets is the main constraint for deep learning applications in medical imaging. To solve this problem, we devise a self-supervised learning approach. The proposed method used inpainting as the pretext task and U-Net for segmentation. The pre-trained network was then fine-tuned on a varied sample of annotated datasets. Our results performance improvement with 50% of labeled data compared to supervise learning. This approach has the potential to significantly improve the applicability of deep learning for medical imaging in resource-constrained settings.

: CRT (Cardiac Resynchronization Therapy) LBBB (Left bundle branch block) LV (Left ventricle) RV (Right ventricle) EAM (Electroanatomical mapping) CE CMR (Contrast enhanced cardiac magnetic resonance) LAT (Local activation time) CS (Coronary sinus) DT MRI (Diffusion tensor magnetic resonance imaging)

One of the main uses of the Intravascular Ultrasound (IVUS) images is tissue classification. Some of the most important tissues are calcium, fibrotic, and lipid plaque. Usually, this task is achieved using DICOM images. Here we exposed the use of reconstructed images from RF data to improve the classification process, because it is given the advantage of normalizing the images to a fixed parameter set. In this approach, 6 different data sets are generated using 150 tissue sections obtained from 54 images of 6 patients. Then, the images are classified employing adaboost classifiers with Error Correcting Output Codes (ECOC) using 85% of the data for training and the rest for testing. The pixel classification rate for DICOM images is: 76.11% composed of 72% for calcified, 76.13% for fibrotic, and 80.21% for lipid tissue. For reconstructed images the hit rate is 79.03%, from which 77.6% of calcium, 69% of fibrosis and 90.5% of lipid tissue are well detected.

Automated segmentation of Lungs plays a crucial role in the computer-aided diagnosis of chest X-Ray (CXR) images. Developing an efficient Lung segmentation model is challenging because of difficulties such as the presence of several edges at the rib cage and clavicle, inconsistent lung shape among different individuals, and the appearance of the lung apex. In this paper, we propose a robust model for Lung segmentation in Chest Radiographs. Our model learns to ignore the irrelevant regions in an input Chest Radiograph while highlighting regions useful for lung segmentation. The proposed model is evaluated on two public chest X-Ray datasets (Montgomery County, MD, USA, and Shenzhen No. 3 People's Hospital in China). The experimental result with a DICE score of 98.6% demonstrates the robustness of our proposed lung segmentation approach.

Nephrolithiasis is a scientific term that refers to kidney stones and means the formation of crystal concretions in the kidney. It is considered a widespread situation that affects millions of people worldwide. Those stones can cause serious discomfort to infected people, especially when they traverse the urinary system, although, the big stones may need a surgical intervention. Various systems are already in use to address kidney stones, including ultrasound imaging for detection, extracorporeal shock wave lithotripsy (ESWL) for non-invasive stone fragmentation, and ureteroscopy for surgical removal, showcasing the advances in medical technology for managing this condition. This study presents an approach for detecting stones in the affected kidney. A public dataset has been employed in this work, containing (2370) images of healthy and affected kidneys. The dataset was utilized to train the proposed approach for the aim of stone detection. To achieve high detection accuracy, we implemented two key phases before classification. The preprocessing phase enhances image quality by reducing noise using a median filter and improving contrast through contrast stretching and tone enhancement. The segmentation phase follows, accurately identifying the kidney's edges and regions of interest for effective feature extraction. The Local Binary Pattern (LBP) technique, combined with the support vector machine (SVM) algorithm serves as the primary components of the proposed model. The feature extraction comes into action through the LBP technique as a preparation step for the SVM classifier to complete the stone detection process. The approach introduced in this paper has the potential to enhance detection accuracy and efficiency. Furthermore, it could be used as an early detection tool to identify potential cases, thereby helping to prevent complications and adverse outcomes. This method aims to improve on the traditional manual process employed by radiologists, which could be described as time and effort consumption rather than the exposure of the interpretations. The obtained results were compared with the most relevant approaches in the field of kidney stone detection, demonstrating the model's effectiveness in achieving the desired goal with a diagnostic accuracy of 96.37% for kidney stones.

Images of the retina acquired using optical coherence tomography (OCT) often suffer from intensity inhomogeneity problems that degrade both the quality of the images and the performance of automated algorithms utilized to measure structural changes. This intensity variation has many causes, including off-axis acquisition, signal attenuation, multi-frame averaging, and vignetting, making it difficult to correct the data in a fundamental way. This paper presents a method for inhomogeneity correction by acting to reduce the variability of intensities within each layer. In particular, the N3 algorithm, which is popular in neuroimage analysis, is adapted to work for OCT data. N3 works by sharpening the intensity histogram, which reduces the variation of intensities within different classes. To apply it here, the data are first converted to a standardized space called macular flat space (MFS). MFS allows the intensities within each layer to be more easily normalized by removing the natural curvature of the retina. N3 is then run on the MFS data using a modified smoothing model, which improves the efficiency of the original algorithm. We show that our method more accurately corrects gain fields on synthetic OCT data when compared to running N3 on non-flattened data. It also reduces the overall variability of the intensities within each layer, without sacrificing contrast between layers, and improves the performance of registration between OCT images.

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Skin cancer is a prevalent and deadly cancer, and early detection is crucial for improving treatment success. Intelligent technologies are currently being used to classify skin lesions. The fundamental goal of this experimental research is to investigate biomedical skin cancer datasets to develop an effective approach for determining whether a cancer is malignant or benign. Well-known deep learning classification models (convolutional neural network (CNN) (sequential), ResNet50, InceptionV3, and Xception) are employed to train and categorize the dataset images. Two large and balanced datasets are collected and employed in this research. One is used to compare the performance of the employed model algorithms. Next, the selected model(s) are again trained on the second dataset for validation and generalization purposes. It turns out that the performance of the Xception model is superior and can be generalized. The performance results obtained from various simulations are tabulated and graphed. Comparative results are also presented.

Multi-compartment modeling of diffusion-weighted magnetic resonance imaging measurements is necessary for accurate brain connectivity analysis. Existing methods for estimating the number and orientations of fascicles in an imaging voxel either depend on non-convex optimization techniques that are sensitive to initialization and measurement noise, or are prone to predicting spurious fascicles. In this paper, we propose a machine learning-based technique that can accurately estimate the number and orientations of fascicles in a voxel. Our method can be trained with either simulated or real diffusion-weighted imaging data. Our method estimates the angle to the closest fascicle for each direction in a set of discrete directions uniformly spread on the unit sphere. This information is then processed to extract the number and orientations of fascicles in a voxel. On realistic simulated phantom data with known ground truth, our method predicts the number and orientations of crossing fascicles more accurately than several existing methods. It also leads to more accurate tractography. On real data, our method is better than or compares favorably with standard methods in terms of robustness to measurement down-sampling and also in terms of expert quality assessment of tractography results.

Background: Obtaining accurate left ventricular (LV) endocardium segmentation requires the exclusion of the papillary muscles from the cardiac wall, which is a significant, albeit challenging, step in cardiac image analysis. Most medical imaging systems suffer from noise in that it affects the processing procedure. We herein introduce a segmentation algorithm, which improves segmentation accuracy by excluding the papillary muscles and suppressing noise effects. Methods: We proposed a hybrid fuzzy-based level set method (LSM) to segment the cardiac wall in magnetic resonance imaging image series. In this approach, we applied improved spatial fuzzy c-means (FCM) on the given image and then utilized the obtained result as the initial contour for the LSM method to obtain more accurate segmentation results. Results: We compared the obtained results with those obtained via manual segmentation as the gold standard vis-à-vis accuracy, Jaccard coefficient, dice coefficient, and false positive ratio. Also, the robustness of the proposed method to the noise was tested by adding the Gaussian noise with different variances to the original image. The obtained results showed 96 ± 1.3% accuracy in normal images and 89 ± 4% accuracy in noisy images with a signal-to-noise ratio of -2.2 to -1. Our results demonstrated that our proposed method was able to exclude the papillary muscles from the cardiac wall. Moreover, our hybrid method showed better accuracy than the 2 methods of FCM and LSM alone.

The arterial lumen is modeled by a spatially continuous right generalized cylinder with piece-wise constant parameters. The method is the identifies the parameters of each cylinder piece from a series of planar contours extracted along an approximate axis of the artery. This curve is defined by a minimal path between the artery end-points. The contours are extracted by use of a 2D Fast Marching algorithm. The identification of the axial parameters is based on a geometrical analogy with piece-wise helical curves, while the identification of the surface parameters uses the Fourier series decomposition of the contours. Thus identified parameters are used as observations in a Kalman optimal estimation scheme that manages the spatial consistency from each piece to another. The method was was evaluated on 15 training and 31 testing datasets from the MICCAI 3D Segmentation in the Clinic Grand Challenge: Carotid Bifurcation Lumen Segmentation and Stenosis Grading (http://cls2009.bigr.nl/).

Pulmonary nodule detection, false positive reduction and segmentation represent three of the most common tasks in the computer aided analysis of chest CT images. Methods have been proposed for each task with deep learning based methods heavily favored recently. However training deep learning models to solve each task separately may be sub-optimal -resource intensive and without the benefit of feature sharing. Here, we propose a new end-to-end 3D deep convolutional neural net (DCNN), called NoduleNet, to solve nodule detection, false positive reduction and nodule segmentation jointly in a multi-task fashion. To avoid friction between different tasks and encourage feature diversification, we incorporate two major design tricks: 1) decoupled feature maps for nodule detection and false positive reduction, and 2) a segmentation refinement subnet for increasing the precision of nodule segmentation. Extensive experiments on the large-scale LIDC dataset demonstrate that the multi-task training is highly beneficial, improving the nodule detection accuracy by 10.27%, compared to the baseline model trained to only solve the nodule detection task. We also carry out systematic ablation studies to highlight contributions from each of the added components. Code is available at .

Artificial intelligence (AI) is transforming the field of radiology. Among various radiologic subspecialties, thoracic imaging has seen a significant rise in demand due to the global increase in heart, vascular, lung, and thoracic diseases such as lung cancer, pneumonia, pulmonary embolism, and cardiovascular diseases. AI promises to revolutionize radiologic diagnostics by enhancing detection, improving accuracy, and reducing the time required to interpret images. It leverages deep learning algorithms, particularly convolutional neural networks, which are increasingly integrated into thoracic imaging workflows to assist radiologists in diagnosing and evaluating heart, vascular, lung, and thoracic diseases. AI systems can help radiologists identify subtle findings that might otherwise be overlooked, thereby increasing efficiency and reducing diagnostic errors. Studies have shown that several AI algorithms have been trained to detect acute chest conditions such as pulmonary embolism, aortic dissection, pneumonia, rib fractures, and lung nodules with high sensitivity and specificity, offering substantial benefits in emergency and high-workload environments. This review article focuses on acute conditions presenting as acute chest syndrome or trauma in emergency settings. It provides an overview of AI applications in thoracic imaging, focusing on advancements in screening, early disease detection, triage and prioritization, automated image analysis, and workflow optimization. These points are supported by review articles published on the subject, including our own publications. We further explore challenges such as regulatory barriers, interpretability, and the need for large, diverse datasets. Finally, we discuss future directions for AI in thoracic imaging, highlighting its potential to enhance patient outcomes and healthcare system efficiencies.

The purpose of this study is to raise the number of aberrant red blood cells that can be recognized
via image processing. To identify the data, this study used the Random Forest Technique as a
machine learning model. As a result, ten aberrant red blood cells were discovered and
categorized by the system. The system's pictures were captured from previous hospital patients.
In addition, the slides are taken using a camera. After that, the image was loaded into the
programme. The image was analyzed and classified by the system. The system's outputs, in
practice, reveal the names of the aberrant red blood cells discovered in the image, as well as a
soft copy of the list.
Keywords: Random Forest Algorithm, Aberrant Red Blood Cells, Classification

With a point matching distance measure which is invariant under translation, rotation and permutation, we learn 2-D point-set objects, by clustering noisy point-set images. Unlike traditional clustering methods which use distance measures that operate on feature vectors -a representation common to most problem domains -this object-based clustering technique employs a distance measure specific to a type of object within a problem domain. Formulating the clustering problem as two nested objective functions, we derive optimization dynamics similar to the Expectation-Maximization algorithm used in mixture models.