J. Imaging, Volume 9, Issue 2 (February 2023) – 35 articles

Improving the reliability of automotive perceptive sensors in degraded weather conditions, including fog, is an important issue for road safety and the development of automated driving. Cerema has designed the PAVIN platform reproducing fog and rain conditions to evaluate optical automotive sensor performance under these conditions. In order to increase the variety of scenarios and technologies under test, the use of digital simulation becomes a major asset. The purpose of this paper is to revive the debate around the realism of the various models underlying the numerical methods. The simulation of the radiative transfer equation by Monte Carlo methods and by simplified noise models is examined. The results of this paper show some gaps in foggy scenes between the ray-tracing method, which is considered to be the most realistic, and simple models for contrast evaluation, which can have a particularly strong impact on obstacle detection algorithms. Full article

Early and accurate tomato disease detection using easily available leaf photos is essential for farmers and stakeholders as it help reduce yield loss due to possible disease epidemics. This paper aims to visually identify nine different infectious diseases (bacterial spot, early blight, Septoria leaf spot, late blight, leaf mold, two-spotted spider mite, mosaic virus, target spot, and yellow leaf curl virus) in tomato leaves in addition to healthy leaves. We implemented EfficientNetB5 with a tomato leaf disease (TLD) dataset without any segmentation, and the model achieved an average training accuracy of 99.84% ± 0.10%, average validation accuracy of 98.28% ± 0.20%, and average test accuracy of 99.07% ± 0.38% over 10 cross folds.The use of gradient-weighted class activation mapping (GradCAM) and local interpretable model-agnostic explanations are proposed to provide model interpretability, which is essential to predictive performance, helpful in building trust, and required for integration into agricultural practice. Full article

LiDAR-based simultaneous localization and mapping (SLAM) and online localization methods are widely used in autonomous driving, and are key parts of intelligent vehicles. However, current SLAM algorithms have limitations in map drift and localization algorithms based on a single sensor have poor adaptability to complex scenarios. A SLAM and online localization method based on multi-sensor fusion is proposed and integrated into a general framework in this paper. In the mapping process, constraints consisting of normal distributions transform (NDT) registration, loop closure detection and real time kinematic (RTK) global navigation satellite system (GNSS) position for the front-end and the pose graph optimization algorithm for the back-end, which are applied to achieve an optimized map without drift. In the localization process, the error state Kalman filter (ESKF) fuses LiDAR-based localization position and vehicle states to realize more robust and precise localization. The open-source KITTI dataset and field tests are used to test the proposed method. The method effectiveness shown in the test results achieves 5–10 cm mapping accuracy and 20–30 cm localization accuracy, and it realizes online autonomous driving in complex scenarios. Full article

The aim of this study was to develop and validate a semi-automated segmentation approach that identifies the round window niche (RWN) and round window membrane (RWM) for use in the development of patient individualized round window niche implants (RNI) to treat inner ear disorders. Twenty cone beam computed tomography (CBCT) datasets of unilateral temporal bones of patients were included in the study. Defined anatomical landmarks such as the RWM were used to develop a customized 3D Slicer™ plugin for semi-automated segmentation of the RWN. Two otolaryngologists (User 1 and User 2) segmented the datasets manually and semi-automatically using the developed software. Both methods were compared in-silico regarding the resulting RWM area and RWN volume. Finally, the developed software was validated ex-vivo in N = 3 body donor implantation tests with additively manufactured RNI. The independently segmented temporal bones of the different Users showed a strong consistency in the volume of the RWN and the area of the RWM. The volume of the semi-automated RWN segmentations were 48 ± 11% smaller on average than the manual segmentations and the area of the RWM of the semi-automated segmentations was 21 ± 17% smaller on average than the manual segmentation. All additively manufactured implants, based on the semi-automated segmentation method could be implanted successfully in a pressure-tight fit into the RWN. The implants based on the manual segmentations failed to fit into the RWN and this suggests that the larger manual segmentations were over-segmentations. This study presents a semi-automated approach for segmenting the RWN and RWM in temporal bone CBCT scans that is efficient, fast, accurate, and not dependent on trained users. In addition, the manual segmentation, often positioned as the gold-standard, actually failed to pass the implantation validation. Full article

Echocardiography is an integral part of the diagnosis and management of cardiovascular disease. The use and application of artificial intelligence (AI) is a rapidly expanding field in medicine to improve consistency and reduce interobserver variability. AI can be successfully applied to echocardiography in addressing variance during image acquisition and interpretation. Furthermore, AI and machine learning can aid in the diagnosis and management of cardiovascular disease. In the realm of echocardiography, accurate interpretation is largely dependent on the subjective knowledge of the operator. Echocardiography is burdened by the high dependence on the level of experience of the operator, to a greater extent than other imaging modalities like computed tomography, nuclear imaging, and magnetic resonance imaging. AI technologies offer new opportunities for echocardiography to produce accurate, automated, and more consistent interpretations. This review discusses machine learning as a subfield within AI in relation to image interpretation and how machine learning can improve the diagnostic performance of echocardiography. This review also explores the published literature outlining the value of AI and its potential to improve patient care. Full article

We present a scoring-based genetic algorithm (SBGA) for wavefront shaping to optimize multiple objectives at a time. The algorithm is able to find one feasible solution despite having to optimize multiple objectives. We employ the algorithm to generate multiple focus points simultaneously and allocate their intensities as desired. We then introduce a third objective to confine light focusing only to desired targets and prevent irradiation in neighboring regions. Through simulations and experiments, we demonstrate the algorithm’s ease of implementation and flexibility to control the search direction. This algorithm can potentially be applied to improve biomedical imaging, optogenetics, and optical trapping. Full article

Aims:Increased left ventricular (LV) wall thickness is frequently encountered in transthoracic echocardiography (TTE). While accurate and early diagnosis is clinically important, given the differences in available therapeutic options and prognosis, an extensive workup is often required to establish the diagnosis. We propose the first echo-based, automated deep learning model with a fusion architecture to facilitate the evaluation and diagnosis of increased left ventricular (LV) wall thickness. Methods and Results: Patients with an established diagnosis of increased LV wall thickness (hypertrophic cardiomyopathy (HCM), cardiac amyloidosis (CA), and hypertensive heart disease (HTN)/others) between 1/2015 and 11/2019 at Mayo Clinic Arizona were identified. The cohort was divided into 80%/10%/10% for training, validation, and testing sets, respectively. Six baseline TTE views were used to optimize a pre-trained InceptionResnetV2 model. Each model output was used to train a meta-learner under a fusion architecture. Model performance was assessed by multiclass area under the receiver operating characteristic curve (AUROC). A total of 586 patients were used for the final analysis (194 HCM, 201 CA, and 191 HTN/others). The mean age was 55.0 years, and 57.8% were male. Among the individual view-dependent models, the apical 4-chamber model had the best performance (AUROC: HCM: 0.94, CA: 0.73, and HTN/other: 0.87). The final fusion model outperformed all the view-dependent models (AUROC: HCM: 0.93, CA: 0.90, and HTN/other: 0.92). Conclusion: The echo-based InceptionResnetV2 fusion model can accurately classify the main etiologies of increased LV wall thickness and can facilitate the process of diagnosis and workup. Full article

A novel method is proposed to estimate surface-spectral reflectance from camera responses using a local optimal reflectance dataset. We adopt a multispectral imaging system that involves an RGB camera capturing multiple images under multiple light sources. A spectral reflectance database is utilized to locally determine the candidates to optimally estimate the spectral reflectance. The proposed estimation method comprises two stages: (1) selecting the local optimal reflectance dataset and (2) determining the best estimate using only the local optimal dataset. In (1), the camera responses are predicted for the respective reflectances in the database, and then the prediction errors are calculated to select the local optimal dataset. In (2), multiple methods are used; in particular, the Wiener and linear minimum mean square error estimators are used to calculate all statistics, based only on the local optimal dataset, and linear and quadratic programming methods are used to solve optimization problems with constraints. Experimental results using different mobile phone cameras show that the estimation accuracy has improved drastically. A much smaller local optimal dataset among spectral reflectance databases is enough to obtain the optimal estimates. The method has potential applications including fields of color science, image science and technology, computer vision, and graphics. Full article

In the past decade, deep neural networks, particularly convolutional neural networks, have revolutionised computer vision. However, all deep learning models may require a large amount of data so as to achieve satisfying results. Unfortunately, the availability of sufficient amounts of data for real-world problems is not always possible, and it is well recognised that a paucity of data easily results in overfitting. This issue may be addressed through several approaches, one of which is data augmentation. In this paper, we survey the existing data augmentation techniques in computer vision tasks, including segmentation and classification, and suggest new strategies. In particular, we introduce a way of implementing data augmentation by using local information in images. We propose a parameter-free and easy to implement strategy, the random local rotation strategy, which involves randomly selecting the location and size of circular regions in the image and rotating them with random angles. It can be used as an alternative to the traditional rotation strategy, which generally suffers from irregular image boundaries. It can also complement other techniques in data augmentation. Extensive experimental results and comparisons demonstrated that the new strategy consistently outperformed its traditional counterparts in, for example, image classification. Full article

Unwanted proteins and metabolic waste in cerebral spinal fluid are cleared from the brain by meningeal and nasal lymphatics and the perineural sheath of cranial nerves; however, the distribution and clearance of cerebral spinal fluid (CSF) along the subarachnoid space of the entire spinal cord is not fully understood. Cryo-fluorescence tomography (CFT) was used to follow the movement of tracers from the ventricular system of the brain down through the meningeal lining of the spinal cord and out to the spinal lymphatic nodes. Isoflurane-anesthetized mice were infused into the lateral cerebroventricle with 5.0 µL of quantum dots [QdotR 605 ITKTM amino (PEG)] over two mins. Mice were allowed to recover (ca 2–3 min) and remained awake and ambulatory for 5, 15, 30, 60, and 120 min after which they were euthanized, and the entire intact body was frozen at −80^°. The entire mouse was sectioned, and white light and fluorescent images were captured after each slice to produce high resolution three-dimensional volumes. Tracer appeared throughout the ventricular system and central canal of the spinal cord and the entire subarachnoid space of the CNS. A signal could be visualized in the nasal cavity, deep cervical lymph nodes, thoracic lymph nodes, and more superficial submandibular lymph nodes as early as 15 min post infusion. A fluorescent signal could be visualized along the dorsal root ganglia and down the proximal extension of the spinal nerves of the thoracic and lumbar segments at 30 min. There was a significant accumulation of tracer in the lumbar and sacral lymph nodes between 15–60 min. The dense fluorescent signal in the thoracic vertebrae noted at 5- and 15-min post infusion was significantly reduced by 30 min. Indeed, all signals in the spinal cord were ostensibly absent by 120 min, except for trace amounts in the coccyx. The brain still had some residual signal at 120 min. These data show that Qdots with a hydrodynamic diameter of 16–20 nm rapidly clear from the brain of awake mice. These data also clearly demonstrate the rapid distribution and efflux of traces along a major length of the vertebral column and the potential contribution of the spinal cord in the clearance of brain waste. Full article

Translational research is aimed at turning discoveries from basic science into results that advance patient treatment. The translation of technical solutions into clinical use is a complex, iterative process that involves different stages of design, development, and validation, such as the identification of unmet clinical needs, technical conception, development, verification and validation, regulatory matters, and ethics. For this reason, many promising technical developments at the interface of technology, informatics, and medicine remain research prototypes without finding their way into clinical practice. Augmented reality is a technology that is now making its breakthrough into patient care, even though it has been available for decades. In this work, we explain the translational process for Medical AR devices and present associated challenges and opportunities. To the best knowledge of the authors, this concept paper is the first to present a guideline for the translation of medical AR research into clinical practice. Full article

Changes in object morphology can be quantified using 3D optical scanning to generate 3D models of an object at different time points. This process requires registration techniques that align target and reference 3D models using mapping functions based on common object features that are unaltered over time. The goal of this study was to determine guidelines when selecting these localized features to ensure robust and accurate 3D model registration. For this study, an object of interest (tibia bone replica) was 3D scanned at multiple time points, and the acquired 3D models were aligned using a simple cubic registration block attached to the object. The size of the registration block and the number of planar block surfaces selected to calculate the mapping functions used for 3D model registration were varied. Registration error was then calculated as the average linear surface variation between the target and reference tibial plateau surfaces. We obtained very low target registration errors when selecting block features with an area equivalent to at least 4% of the scanning field of view. Additionally, we found that at least two orthogonal surfaces should be selected to minimize registration error. Therefore, when registering 3D models to measure multi-temporal morphological change (e.g., mechanical wear), we recommend selecting multiplanar features that account for at least 4% of the scanning field of view. For the first time, this study has provided guidelines for selecting localized object features that can provide accurate 3D model registration for 3D scanned objects. Full article

A rapidly spreading epidemic, COVID-19 had a serious effect on millions and took many lives. Therefore, for individuals with COVID-19, early discovery is essential for halting the infection’s progress. To quickly and accurately diagnose COVID-19, imaging modalities, including computed tomography (CT) scans and chest X-ray radiographs, are frequently employed. The potential of artificial intelligence (AI) approaches further explored the creation of automated and precise COVID-19 detection systems. Scientists widely use deep learning techniques to identify coronavirus infection in lung imaging. In our paper, we developed a novel light CNN model architecture with watershed-based region-growing segmentation on Chest X-rays. Both CT scans and X-ray radiographs were employed along with 5-fold cross-validation. Compared to earlier state-of-the-art models, our model is lighter and outperformed the previous methods by achieving a mean accuracy of 98.8% on X-ray images and 98.6% on CT scans, predicting the rate of 0.99% and 0.97% for PPV (Positive predicted Value) and NPV (Negative predicted Value) rate of 0.98% and 0.99%, respectively. Full article

Smartphones with an in-built camera are omnipresent today in the life of over eighty percent of the world’s population. They are very often used to photograph documents. Document binarization is a key process in many document processing platforms. This paper assesses the quality, file size and time performance of sixty-eight binarization algorithms using five different versions of the input images. The evaluation dataset is composed of deskjet, laser and offset printed documents, photographed using six widely-used mobile devices with the strobe flash off and on, under two different angles and four shots with small variations in the position. Besides that, this paper also pinpoints the algorithms per device that may provide the best visual quality-time, document transcription accuracy-time, and size-time trade-offs. Furthermore, an indication is also given on the “overall winner” that would be the algorithm of choice if one has to use one algorithm for a smartphone-embedded application. Full article

Gender determination is the first step in forensic identification, followed by age and height determination, which are both affected by gender. This study assessed the accuracy of gender estimation using mandibular morphometric indices on panoramic radiographs of an Iranian population. This retrospective study evaluated 290 panoramic radiographs (145 males and 145 females). The maximum and minimum ramus width, coronoid height, condylar height, antegonial angle, antegonial depth, gonial angle, and the superior border of mental foramen were bilaterally measured as well as bicondylar and bigonial breadths using Scanora Lite. Correlation of parameters with gender was analyzed by univariate, multiple, and best models. All indices except for gonial angle were significantly different between males and females and can be used for gender determination according to univariate model. Condylar height, coronoid height, and superior border of mental foramen and ramus were still significantly greater in males than in females after controlling for the effect of confounders (p < 0.05). Based on the best model, a formula including five indices of bicondylar breadth, condylar height, coronoid height, minimum ramus width, and superior border of mental foramen was used for gender determination. Values higher than 56% indicate male gender, while lower values indicate female gender, with 81.38% specificity for correct detection of females and 88.97% sensitivity for correct detection of males. Despite the satisfactory results, future research should focus on larger populations to verify the accuracy of the present findings. Full article

This paper proposed a method for reconstructing floorplans from indoor point clouds. Unlike existing corner and line primitive detection algorithms, this method uses a generative adversarial network to learn the complex distribution of indoor layout graphics, and repairs incomplete room masks into more regular segmentation areas. Automatic learning of the structure information of layout graphics can reduce the dependence on geometric priors, and replacing complex optimization algorithms with Deep Neural Networks (DNN) can improve the efficiency of data processing. The proposed method can retain more shape information from the original data and improve the accuracy of the overall structure details. On this basis, the method further used an edge optimization algorithm to eliminate pixel-level edge artifacts that neural networks cannot perceive. Finally, combined with the constraint information of the overall layout, the method can generate compact floorplans with rich semantic information. Experimental results indicated that the algorithm has robustness and accuracy in complex 3D indoor datasets; its performance is competitive with those of existing methods. Full article

Masked face recognition (MFR) is an interesting topic in which researchers have tried to find a better solution to improve and enhance performance. Recently, COVID-19 caused most of the recognition system fails to recognize facial images since the current face recognition cannot accurately capture or detect masked face images. This paper introduces the proposed method known as histogram-based recurrent neural network (HRNN) MFR to solve the undetected masked face problem. The proposed method includes the feature descriptor of histograms of oriented gradients (HOG) as the feature extraction process and recurrent neural network (RNN) as the deep learning process. We have proven that the combination of both approaches works well and achieves a high true acceptance rate (TAR) of 99 percent. In addition, the proposed method is designed to overcome the underfitting problem and reduce computational burdens with large-scale dataset training. The experiments were conducted on two benchmark datasets which are RMFD (Real-World Masked Face Dataset) and Labeled Face in the Wild Simulated Masked Face Dataset (LFW-SMFD) to vindicate the viability of the proposed HRNN method. Full article

Intracranial hemorrhage is a serious medical problem that requires rapid and often intensive medical care. Identifying the location and type of any hemorrhage present is a critical step in the treatment of the patient. Detection of, and diagnosis of, a hemorrhage that requires an urgent procedure is a difficult and time-consuming process for human experts. In this paper, we propose methods based on EfficientDet’s deep-learning technology that can be applied to the diagnosis of hemorrhages at a patient level and which could, thus, become a decision-support system. Our proposal is two-fold. On the one hand, the proposed technique classifies slices of computed tomography scans for the presence of hemorrhage or its lack of, and evaluates whether the patient is positive in terms of hemorrhage, and achieving, in this regard, 92.7% accuracy and 0.978 ROC AUC. On the other hand, our methodology provides visual explanations of the chosen classification using the Grad-CAM methodology. Full article

(1) Background: The authors developed a new non-invasive dermatological infrared macroimaging analysis technique (MacroIR) that evaluates microvascular, inflammatory, and metabolic changes that may be dermoscopy complimentary, by analyzing different skin and mucosal lesions in a combined way—naked eye, polarized light dermatoscopy (PLD), and MacroIR—and comparing results; (2) Methods: ten cases were evaluated using a smartphone coupled with a dermatoscope and a macro lens integrated far-infrared transducer into specific software to capture and organize high-resolution images in different electromagnetic spectra, and then analyzed by a dermatologist; (3) Results: It was possible to identify and compare structures found in two dermoscopic forms. Visual anatomical changes were correlated with MacroIR and aided skin surface dermatological analysis, presenting studied area microvascular, inflammatory, and metabolic data. All MacroIR images correlated with PLD, naked eye examination, and histopathological findings; (4) Conclusion: MacroIR and clinic dermatologist concordance rates were comparable for all dermatological conditions in this study. MacroIR imaging is a promising method that can improve dermatological diseases diagnosis. The observations are preliminary and require further evaluation in larger studies. Full article

Skin detection involves identifying skin and non-skin areas in a digital image and is commonly used in various applications, such as analyzing hand gestures, tracking body parts, and facial recognition. The process of distinguishing between skin and non-skin regions in a digital image is widely used in a variety of applications, ranging from hand-gesture analysis to body-part tracking to facial recognition. Skin detection is a challenging problem that has received a lot of attention from experts and proposals from the research community in the context of intelligent systems, but the lack of common benchmarks and unified testing protocols has hampered fairness among approaches. Comparisons are very difficult. Recently, the success of deep neural networks has had a major impact on the field of image segmentation detection, resulting in various successful models to date. In this work, we survey the most recent research in this field and propose fair comparisons between approaches, using several different datasets. The main contributions of this work are (i) a comprehensive review of the literature on approaches to skin-color detection and a comparison of approaches that may help researchers and practitioners choose the best method for their application; (ii) a comprehensive list of datasets that report ground truth for skin detection; and (iii) a testing protocol for evaluating and comparing different skin-detection approaches. Moreover, we propose an ensemble of convolutional neural networks and transformers that obtains a state-of-the-art performance. Full article

Deep learning approaches are becoming increasingly important for the estimation of the Remaining Useful Life (RUL) of mechanical elements such as bearings. This paper proposes and evaluates a novel transfer learning-based approach for RUL estimations of different bearing types with small datasets and low sampling rates. The approach is based on an intermediate domain that abstracts features of the bearings based on their fault frequencies. The features are processed by convolutional layers. Finally, the RUL estimation is performed using a Long Short-Term Memory (LSTM) network. The transfer learning relies on a fixed-feature extraction. This novel deep learning approach successfully uses data of a low-frequency range, which is a precondition to use low-cost sensors. It is validated against the IEEE PHM 2012 Data Challenge, where it outperforms the winning approach. The results show its suitability for low-frequency sensor data and for efficient and effective transfer learning between different bearing types. Full article

The present study explores the efficacy of Machine Learning and Artificial Neural Networks in age assessment using the root length of the second and third molar teeth. A dataset of 1000 panoramic radiographs with intact second and third molars ranging from 12 to 25 years was archived. The length of the mesial and distal roots was measured using ImageJ software. The dataset was classified in three ways based on the age distribution: 2–Class, 3–Class, and 5–Class. We used Support Vector Machine (SVM), Random Forest (RF), and Logistic Regression models to train, test, and analyze the root length measurements. The mesial root of the third molar on the right side was a good predictor of age. The SVM showed the highest accuracy of 86.4% for 2–class, 66% for 3–class, and 42.8% for 5–Class. The RF showed the highest accuracy of 47.6% for 5–Class. Overall the present study demonstrated that the Deep Learning model (fully connected model) performed better than the Machine Learning models, and the mesial root length of the right third molar was a good predictor of age. Additionally, a combination of different root lengths could be informative while building a Machine Learning model. Full article

Radiomic analysis allows for the detection of imaging biomarkers supporting decision-making processes in clinical environments, from diagnosis to prognosis. Frequently, the original set of radiomic features is augmented by considering high-level features, such as wavelet transforms. However, several wavelets families (so called kernels) are able to generate different multi-resolution representations of the original image, and which of them produces more salient images is not yet clear. In this study, an in-depth analysis is performed by comparing different wavelet kernels and by evaluating their impact on predictive capabilities of radiomic models. A dataset composed of 1589 chest X-ray images was used for COVID-19 prognosis prediction as a case study. Random forest, support vector machine, and XGBoost were trained (on a subset of 1103 images) after a rigorous feature selection strategy to build-up the predictive models. Next, to evaluate the models generalization capability on unseen data, a test phase was performed (on a subset of 486 images). The experimental findings showed that Bior1.5, Coif1, Haar, and Sym2 kernels guarantee better and similar performance for all three machine learning models considered. Support vector machine and random forest showed comparable performance, and they were better than XGBoost. Additionally, random forest proved to be the most stable model, ensuring an appropriate balance between sensitivity and specificity. Full article

The method of 18F-sodium fluoride (NaF) positron emission tomography/computed tomography (PET/CT) of atherosclerosis was introduced 12 years ago. This approach is particularly interesting because it demonstrates microcalcification as an incipient sign of atherosclerosis before the development of arterial wall macrocalcification detectable by CT. However, this method has not yet found its place in the clinical routine. The more exact association between NaF uptake and future arterial calcification is not fully understood, and it remains unclear to what extent NaF-PET may replace or significantly improve clinical cardiovascular risk scoring. The first 10 years of publications in the field were characterized by heterogeneity at multiple levels, and it is not clear how the method may contribute to triage and management of patients with atherosclerosis, including monitoring effects of anti-atherosclerosis intervention. The present review summarizes findings from the recent 2¾ years including the ability of NaF-PET imaging to assess disease progress and evaluate response to treatment. Despite valuable new information, pertinent questions remain unanswered, not least due to a pronounced lack of standardization within the field and of well-designed long-term studies illuminating the natural history of atherosclerosis and effects of intervention. Full article

In this paper, we propose a framework that constructs two types of image aesthetic assessment (IAA) models with different CNN architectures and improves the performance of image aesthetic score (AS) prediction by the ensemble. Moreover, the attention regions of the models to the images are extracted to analyze the consistency with the subjects in the images. The experimental results verify that the proposed method is effective for improving the AS prediction. The average F1 of the ensemble improves 5.4% over the model of type A, and 33.1% over the model of type B. Moreover, it is found that the AS classification models trained on the XiheAA dataset seem to learn the latent photography principles, although it cannot be said that they learn the aesthetic sense. Full article

Omnidirectional images have drawn great research attention recently thanks to their great potential and performance in various computer vision tasks. However, processing such a type of image requires an adaptation to take into account spherical distortions. Therefore, it is not trivial to directly extend the conventional convolutional neural networks on omnidirectional images because CNNs were initially developed for perspective images. In this paper, we present a general method to adapt perspective convolutional networks to equirectangular images, forming a novel distortion-aware convolution. Our proposed solution can be regarded as a replacement for the existing convolutional network without requiring any additional training cost. To verify the generalization of our method, we conduct an analysis on three basic vision tasks, i.e., semantic segmentation, optical flow, and monocular depth. The experiments on both virtual and real outdoor scenarios show our adapted spherical models consistently outperform their counterparts. Full article

Fast reconstruction of holographic and diffractive optical elements (DOE) can be implemented by binary digital micromirror devices (DMD). Since micromirrors of the DMD have two positions, the synthesized DOEs must be binary. This work studies the possibility of improving the method of synthesis of amplitude binary inline Fresnel holograms in divergent beams. The method consists of the modified Gerchberg–Saxton algorithm, Otsu binarization and direct search with random trajectory technique. To achieve a better quality of reconstruction, various binarization methods were compared. We performed numerical and optical experiments using the DMD. Holograms of halftone image with size up to 1024 × 1024 pixels were synthesized. It was determined that local and several global threshold methods provide the best quality. Compared to the Otsu binarization used in the original method of the synthesis, the reconstruction quality (MSE and SSIM values) is improved by 46% and the diffraction efficiency is increased by 27%. Full article

Background: Previous studies have demonstrated different MRI characteristics in Asian and Western patients with multiple sclerosis (MS). However, the number of studies performed on Thai patients is still limited. Furthermore, these studies were conducted before the revision of the McDonald criteria in 2017. Methods: A retrospective descriptive study was performed on Thai patients diagnosed with MS, according to the McDonald criteria (2017), in a tertiary care hospital in Thailand. Results: Thirty-two patients were included (twenty-seven female and five male patients). The mean age was 37.8 years. Most (28 patients) had relapsing remitting MS. Brain MRIs were available for all 32 patients, all of which showed abnormalities. The most common locations were the periventricular regions (78.1%), juxtacortical regions (75%) and deep white matter (62.5%). Dawson’s fingers were identified in 20 patients (62.5%). Tumefactive MS was noted in two patients. Gadolinium-enhancing brain lesions were noted in nine patients (28.1%). Optic nerve lesions were found in seven patients. Six of the seven patients showed short segmental lesions with predominant posterior-half involvement. Spinal MRIs were available for 26 patients, with abnormalities detected in 23. Most (11 patients) had lesions both in the cervical and in the thoracic spinal cord. In total, 22 patients (95.7%) showed lesions at the periphery, most commonly at the lateral column. Fifteen patients showed lesions shorter than three vertebral segments (65.2%). Enhancing spinal lesions were noted in 14 patients. Dissemination in space was fulfilled in 31 patients (96.9%). Conclusion: Some of the MRI findings in our study were similar to those of previous studies in Thailand and Asia, emphasizing the difference between Asian and Western MS. Full article

Colorectal cancer (CRC) is a leading cause of cancer-related deaths worldwide. The best method to prevent CRC is with a colonoscopy. During this procedure, the gastroenterologist searches for polyps. However, there is a potential risk of polyps being missed by the gastroenterologist. Automated detection of polyps helps to assist the gastroenterologist during a colonoscopy. There are already publications examining the problem of polyp detection in the literature. Nevertheless, most of these systems are only used in the research context and are not implemented for clinical application. Therefore, we introduce the first fully open-source automated polyp-detection system scoring best on current benchmark data and implementing it ready for clinical application. To create the polyp-detection system (ENDOMIND-Advanced), we combined our own collected data from different hospitals and practices in Germany with open-source datasets to create a dataset with over 500,000 annotated images. ENDOMIND-Advanced leverages a post-processing technique based on video detection to work in real-time with a stream of images. It is integrated into a prototype ready for application in clinical interventions. We achieve better performance compared to the best system in the literature and score a F1-score of 90.24% on the open-source CVC-VideoClinicDB benchmark. Full article

This paper proposes a new Hepatocellular Carcinoma (HCC) classification method utilizing a hyperspectral imaging system (HSI) integrated with a light microscope. Using our custom imaging system, we have captured 270 bands of hyperspectral images of healthy and cancer tissue samples with HCC diagnosis from a liver microarray slide. Convolutional Neural Networks with 3D convolutions (3D-CNN) have been used to build an accurate classification model. With the help of 3D convolutions, spectral and spatial features within the hyperspectral cube are incorporated to train a strong classifier. Unlike 2D convolutions, 3D convolutions take the spectral dimension into account while automatically collecting distinctive features during the CNN training stage. As a result, we have avoided manual feature engineering on hyperspectral data and proposed a compact method for HSI medical applications. Moreover, the focal loss function, utilized as a CNN cost function, enables our model to tackle the class imbalance problem residing in the dataset effectively. The focal loss function emphasizes the hard examples to learn and prevents overfitting due to the lack of inter-class balancing. Our empirical results demonstrate the superiority of hyperspectral data over RGB data for liver cancer tissue classification. We have observed that increased spectral dimension results in higher classification accuracy. Both spectral and spatial features are essential in training an accurate learner for cancer tissue classification. Full article