Technical note: Preprocessing of portal images to improve image quality of VMAT-CT

BACKGROUND

The concept of volumetric modulated arc therapy-computed tomography (VMAT-CT) was proposed more than a decade ago. However, its application has been very limited mainly due to the poor image quality. More specifically, the blurred areas in electronic portal imaging device (EPID) images collected during VMAT heavily degrade the image quality of VMAT-CT.

PURPOSE

The goal of this study was to propose systematic methods to preprocess EPID images and improve the image quality of VMAT-CT.

METHODS

Online region-based active contour method was introduced to binarize portal images. Multi-leaf collimator (MLC) motion modeling was developed to remove the MLC motion blur. Outlier filtering was then applied to replace the remaining artifacts with plausible data. To assess the impact of these preprocessing methods on the image quality of VMAT-CT, 44 clinical VMAT plans for several treatment sites (lung, esophagus, and head & neck) were delivered to a Rando phantom, and several real-patient cases were also acquired. VMAT-CT reconstruction was attempted for all the cases, and image quality was evaluated.

RESULTS

All three preprocessing methods could effectively remove the blurred edges of EPID images. The combined preprocessing methods not only saved VMAT-CT from distortions and artifacts, but also increased the percentage of VMAT plans that can be reconstructed.

CONCLUSIONS

The systematic preprocessing of portal images improves the image quality of VMAT-CT significantly, and facilitates the application of VMAT-CT as an effective image guidance tool.

Fetal growth analysis from ultrasound videos based on different biometrics using optimal segmentation and hybrid classifier

Birth defects and their associated deaths, high health and financial costs of maternal care and associated morbidity are major contributors to infant mortality. If permitted by law, prenatal diagnosis allows for intrauterine care, more complicated hospital deliveries, and termination of pregnancy. During pregnancy, a set of measurements is commonly used to monitor the fetal health, including fetal head circumference, crown-rump length, abdominal circumference, and femur length. Because of the intricate interactions between the biological tissues and the US waves mother and fetus, analyzing fetal US images from a specialized perspective is difficult. Artifacts include acoustic shadows, speckle noise, motion blur, and missing borders. The fetus moves quickly, body structures close, and the weeks of pregnancy vary greatly. In this work, we propose a fetal growth analysis through US image of head circumference biometry using optimal segmentation and hybrid classifier. First, we introduce a hybrid whale with oppositional fruit fly optimization (WOFF) algorithm for optimal segmentation of segment fetal head which improves the detection accuracy. Next, an improved U-Net design is utilized for the hidden feature (head circumference biometry) extraction which extracts features from the segmented extraction. Then, we design a modified Boosting arithmetic optimization (MBAO) algorithm for feature optimization to selects optimal best features among multiple features for the reduction of data dimensionality issues. Furthermore, a hybrid deep learning technique called bi-directional LSTM with convolutional neural network (B-LSTM-CNN) for fetal growth analysis to compute the fetus growth and health. Finally, we validate our proposed method through the open benchmark datasets are HC18 (Ultrasound image) and oxford university research archive (ORA-data) (Ultrasound video frames). We compared the simulation results of our proposed algorithm with the existing state-of-art techniques in terms of various metrics.

Automated grape leaf nutrition deficiency disease detection and classification Equilibrium Optimizer with deep transfer learning model

Our approach includes picture preprocessing, feature extraction utilizing the SqueezeNet model, hyperparameter optimisation utilising the Equilibrium Optimizer (EO) algorithm, and classification utilising a Stacked Autoencoder (SAE) model. Each of these processes is carried out in a series of separate steps. During the image preprocessing stage, contrast limited adaptive histogram equalisations (CLAHE) is utilized to improve the contrasts, and Adaptive Bilateral Filtering (ABF) to get rid of any noise that may be present. The SqueezeNet paradigm is utilized to obtain relevant characteristics from the pictures that have been preprocessed, and the EO technique is utilized to fine-tune the hyperparameters. Finally, the SAE model categorises the diseases that affect the grape leaf. The simulation analysis of the EODTL-GLDC technique tested New Plant Diseases Datasets and the results were inspected in many prospects. The results demonstrate that this model outperforms other deep learning techniques and methods that are more often related to machine learning. Specifically, this technique was able to attain a precision of 96.31% on the testing datasets and 96.88% on the training data set that was split 80:20. These results offer more proof that the suggested strategy is successful in automating the detection and categorization of grape leaf diseases.

Dried shiitake mushroom grade recognition using D-VGG network and machine vision

Grading dried shiitake mushrooms is an indispensable production step, as there are large quality differences between different grades, which affect the product's price and marketability. Dried shiitake mushroom samples have irregular shapes, small morphological differences between different grades of the same species, and they may occur in mixed grades, which causes challenges to the automatic grade recognition using machine vision. In this study, a comprehensive method to solve this problem is provided, including image acquisition, preprocessing, dataset creation, and grade recognition. The osprey optimization algorithm (OOA) is used to improve the computational efficiency of Otsu's threshold binarization and obtain complete mushroom contours samples efficiently. Then, a method for dried shiitake mushroom grade recognition based on the improved VGG network (D-VGG) is proposed. The method uses the VGG16 network as the base framework, optimizes the convolutional layer of the network, and uses a global average pooling layer instead of a fully connected layer to reduce the risk of model overfitting. In addition, a residual module and batch normalization are introduced to enhance the learning effect of texture details, accelerate the convergence of the model, and improve the stability of the training process. An improved channel attention network is proposed to enhance the feature weights of different channels and improve the grading performance of the model. The experimental results show that the improved network model (D-VGG) can recognize different dried shiitake mushroom grades with high accuracy and recognition efficiency, achieving a final grading accuracy of 96.21%, with only 46.77 ms required to process a single image. The dried shiitake mushroom grade recognition method proposed in this study provides a new implementation approach for the dried shiitake mushroom quality grading process, as well as a reference for real-time grade recognition of other agricultural products.

Weak Spatial Target Extraction Based on Small-Field Optical System

Compared to wide-field telescopes, small-field detection systems have higher spatial resolution, resulting in stronger detection capabilities and higher positioning accuracy. When detecting by small fields in synchronous orbit, both space debris and fixed stars are imaged as point targets, making it difficult to distinguish them. In addition, with the improvement in detection capabilities, the number of stars in the background rapidly increases, which puts higher requirements on recognition algorithms. Therefore, star detection is indispensable for identifying and locating space debris in complex backgrounds. To address these difficulties, this paper proposes a real-time star extraction method based on adaptive filtering and multi-frame projection. We use bad point repair and background suppression algorithms to preprocess star images. Afterwards, we analyze and enhance the target signal-to-noise ratio (SNR). Then, we use multi-frame projection to fuse information. Subsequently, adaptive filtering, adaptive morphology, and adaptive median filtering algorithms are proposed to detect trajectories. Finally, the projection is released to locate the target. Our recognition algorithm has been verified by real star images, and the images were captured using small-field telescopes. The experimental results demonstrate the effectiveness of the algorithm proposed in this paper. We successfully extracted hip-27066 star, which has a magnitude of about 12 and an SNR of about 1.5. Compared with existing methods, our algorithm has advantages in both recognition rate and false-alarm rate, and can be used as a real-time target recognition algorithm for space-based synchronous orbit detection payloads.

Deep Learning for Dental Diagnosis: A Novel Approach to Furcation Involvement Detection on Periapical Radiographs

Furcation defects pose a significant challenge in the diagnosis and treatment planning of periodontal diseases. The accurate detection of furcation involvements (FI) on periapical radiographs (PAs) is crucial for the success of periodontal therapy. This research proposes a deep learning-based approach to furcation defect detection using convolutional neural networks (CNN) with an accuracy rate of 95%. This research has undergone a rigorous review by the Institutional Review Board (IRB) and has received accreditation under number 202002030B0C505. A dataset of 300 periapical radiographs of teeth with and without FI were collected and preprocessed to enhance the quality of the images. The efficient and innovative image masking technique used in this research better enhances the contrast between FI symptoms and other areas. Moreover, this technology highlights the region of interest (ROI) for the subsequent CNN models training with a combination of transfer learning and fine-tuning techniques. The proposed segmentation algorithm demonstrates exceptional performance with an overall accuracy up to 94.97%, surpassing other conventional methods. Moreover, in comparison with existing CNN technology for identifying dental problems, this research proposes an improved adaptive threshold preprocessing technique that produces clearer distinctions between teeth and interdental molars. The proposed model achieves impressive results in detecting FI with identification rates ranging from 92.96% to a remarkable 94.97%. These findings suggest that our deep learning approach holds significant potential for improving the accuracy and efficiency of dental diagnosis. Such AI-assisted dental diagnosis has the potential to improve periodontal diagnosis, treatment planning, and patient outcomes. This research demonstrates the feasibility and effectiveness of using deep learning algorithms for furcation defect detection on periapical radiographs and highlights the potential for AI-assisted dental diagnosis. With the improvement of dental abnormality detection, earlier intervention could be enabled and could ultimately lead to improved patient outcomes.

Integrated diffusion image operator (iDIO): A pipeline for automated configuration and processing of diffusion MRI data

The preprocessing of diffusion magnetic resonance imaging (dMRI) data involve numerous steps, including the corrections for head motion, susceptibility distortion, low signal-to-noise ratio, and signal drifting. Researchers or clinical practitioners often need to configure different preprocessing steps depending on disparate image acquisition schemes, which increases the technical threshold for dMRI analysis for nonexpert users. This could cause disparities in data processing approaches and thus hinder the comparability between studies. To make the dMRI data processing steps transparent and adapt to various dMRI acquisition schemes for researchers, we propose a semi-automated pipeline tool for dMRI named integrated diffusion image operator or iDIO. This pipeline integrates features from a wide range of advanced dMRI software tools and targets at providing a one-click solution for dMRI data analysis, via adaptive configuration for a set of suggested processing steps based on the image header of the input data. Additionally, the pipeline provides options for post-processing, such as estimation of diffusion tensor metrics and whole-brain tractography-based connectomes reconstruction using common brain atlases. The iDIO pipeline also outputs an easy-to-interpret quality control report to facilitate users to assess the data quality. To keep the transparency of data processing, the execution log and all the intermediate images produced in the iDIO's workflow are accessible. The goal of iDIO is to reduce the barriers for clinical or nonspecialist users to adopt the state-of-art dMRI processing steps.

MR Intensity Normalization Methods Impact Sequence Specific Radiomics Prognostic Model Performance in Primary and Recurrent High-Grade Glioma

PURPOSE

This study investigates the impact of different intensity normalization (IN) methods on the overall survival (OS) radiomics models' performance of MR sequences in primary (pHGG) and recurrent high-grade glioma (rHGG).

METHODS

MR scans acquired before radiotherapy were retrieved from two independent cohorts (rHGG C1: 197, pHGG C2: 141) from multiple scanners (15, 14). The sequences are T1 weighted (w), contrast-enhanced T1w (T1wce), T2w, and T2w-FLAIR. Sequence-specific significant features (SF) associated with OS, extracted from the tumour volume, were derived after applying 15 different IN methods. Survival analyses were conducted using Cox proportional hazard (CPH) and Poisson regression (POI) models. A ranking score was assigned based on the 10-fold cross-validated (CV) concordance index (C-I), mean square error (MSE), and the Akaike information criterion (AICs), to evaluate the methods' performance.

RESULTS

Scatter plots of the 10-CV C-I and MSE against the AIC showed an impact on the survival predictions between the IN methods and MR sequences (C1/C2 C-I range: 0.62-0.71/0.61-0.72, MSE range: 0.20-0.42/0.13-0.22). White stripe showed stable results for T1wce (C1/C2 C-I: 0.71/0.65, MSE: 0.21/0.14). Combat (0.68/0.62, 0.22/0.15) and histogram matching (HM, 0.67/0.64, 0.22/0.15) showed consistent prediction results for T2w models. They were also the top-performing methods for T1w in C2 (Combat: 0.67, 0.13; HM: 0.67, 0.13); however, only HM achieved high predictions in C1 (0.66, 0.22). After eliminating IN impacted SF using Spearman's rank-order correlation coefficient, a mean decrease in the C-I and MSE of 0.05 and 0.03 was observed in all four sequences.

CONCLUSION

The IN method impacted the predictive power of survival models; thus, performance is sequence-dependent.

DenseNet weed recognition model combining local variance preprocessing and attention mechanism

INTRODUCTION

The purpose of this paper is to effectively and accurately identify weed species in crop fields in complex environments. There are many kinds of weeds in the detection area, which are densely distributed.

METHODS

The paper proposes the use of local variance pre-processing method for background segmentation and data enhancement, which effectively removes the complex background and redundant information from the data, and prevents the experiment from overfitting, which can improve the accuracy rate significantly. Then, based on the optimization improvement of DenseNet network, Efficient Channel Attention (ECA) mechanism is introduced after the convolutional layer to increase the weight of important features, strengthen the weed features and suppress the background features.

RESULTS

Using the processed images to train the model, the accuracy rate reaches 97.98%, which is a great improvement, and the comprehensive performance is higher than that of DenseNet, VGGNet-16, VGGNet-19, ResNet-50, DANet, DNANet, and U-Net models.

DISCUSSION

The experimental data show that the model and method we designed are well suited to solve the problem of accurate identification of crop and weed species in complex environments, laying a solid technical foundation for the development of intelligent weeding robots.

Automated Detection of Broncho-Arterial Pairs Using CT Scans Employing Different Approaches to Classify Lung Diseases

Current research indicates that for the identification of lung disorders, comprising pneumonia and COVID-19, structural distortions of bronchi and arteries (BA) should be taken into account. CT scans are an effective modality to detect lung anomalies. However, anomalies in bronchi and arteries can be difficult to detect. Therefore, in this study, alterations of bronchi and arteries are considered in the classification of lung diseases. Four approaches to highlight these are introduced: (a) a Hessian-based approach, (b) a region-growing algorithm, (c) a clustering-based approach, and (d) a color-coding-based approach. Prior to this, the lungs are segmented, employing several image preprocessing algorithms. The utilized COVID-19 Lung CT scan dataset contains three classes named Non-COVID, COVID, and community-acquired pneumonia, having 6983, 7593, and 2618 samples, respectively. To classify the CT scans into three classes, two deep learning architectures, (a) a convolutional neural network (CNN) and (b) a CNN with long short-term memory (LSTM) and an attention mechanism, are considered. Both these models are trained with the four datasets achieved from the four approaches. Results show that the CNN model achieved test accuracies of 88.52%, 87.14%, 92.36%, and 95.84% for the Hessian, the region-growing, the color-coding, and the clustering-based approaches, respectively. The CNN with LSTM and an attention mechanism model results in an increase in overall accuracy for all approaches with an 89.61%, 88.28%, 94.61%, and 97.12% test accuracy for the Hessian, region-growing, color-coding, and clustering-based approaches, respectively. To assess overfitting, the accuracy and loss curves and k-fold cross-validation technique are employed. The Hessian-based and region-growing algorithm-based approaches produced nearly equivalent outcomes. Our proposed method outperforms state-of-the-art studies, indicating that it may be worthwhile to pay more attention to BA features in lung disease classification based on CT images.

An Approach toward Automatic Specifics Diagnosis of Breast Cancer Based on an Immunohistochemical Image

The paper explored the problem of automatic diagnosis based on immunohistochemical image analysis. The issue of automated diagnosis is a preliminary and advisory statement for a diagnostician. The authors studied breast cancer histological and immunohistochemical images using the following biomarkers progesterone, estrogen, oncoprotein, and a cell proliferation biomarker. The authors developed a breast cancer diagnosis method based on immunohistochemical image analysis. The proposed method consists of algorithms for image preprocessing, segmentation, and the determination of informative indicators (relative area and intensity of cells) and an algorithm for determining the molecular genetic breast cancer subtype. An adaptive algorithm for image preprocessing was developed to improve the quality of the images. It includes median filtering and image brightness equalization techniques. In addition, the authors developed a software module part of the HIAMS software package based on the Java programming language and the OpenCV computer vision library. Four molecular genetic breast cancer subtypes could be identified using this solution: subtype Luminal A, subtype Luminal B, subtype HER2/neu amplified, and basalt-like subtype. The developed algorithm for the quantitative characteristics of the immunohistochemical images showed sufficient accuracy in determining the cancer subtype "Luminal A". It was experimentally established that the relative area of the nuclei of cells covered with biomarkers of progesterone, estrogen, and oncoprotein was more than 85%. The given approach allows for automating and accelerating the process of diagnosis. Developed algorithms for calculating the quantitative characteristics of cells on immunohistochemical images can increase the accuracy of diagnosis.

Coupled Gold Nanoparticles with Aptamers Colorimetry for Detection of Amoxicillin in Human Breast Milk Based on Image Preprocessing and BP-ANN

Antibiotic residues in breast milk can have an impact on the intestinal flora and health of babies. Amoxicillin, as one of the most used antibiotics, affects the abundance of some intestinal bacteria. In this study, we developed a convenient and rapid process that used a combination of colorimetric methods and artificial intelligence image preprocessing, and back propagation-artificial neural network (BP-ANN) analysis to detect amoxicillin in breast milk. The colorimetric method derived from the reaction of gold nanoparticles (AuNPs) was coupled to aptamers (ssDNA) with different concentrations of amoxicillin to produce different color results. The color image was captured by a portable image acquisition device, and image preprocessing was implemented in three steps: segmentation, filtering, and cropping. We decided on a range of detection from 0 µM to 3.9 µM based on the physiological concentration of amoxicillin in breast milk and the detection effect. The segmentation and filtering steps were conducted by Hough circle detection and Gaussian filtering, respectively. The segmented results were analyzed by linear regression and BP-ANN, and good linear correlations between the colorimetric image value and concentration of target amoxicillin were obtained. The R2 and MSE of the training set were 0.9551 and 0.0696, respectively, and those of the test set were 0.9276 and 0.1142, respectively. In prepared breast milk sample detection, the recoveries were 111.00%, 98.00%, and 100.20%, and RSDs were 6.42%, 4.27%, and 1.11%. The result suggests that the colorimetric process combined with artificial intelligence image preprocessing and BP-ANN provides an accurate, rapid, and convenient way to achieve the detection of amoxicillin in breast milk.

Quantitative Assessment of Breast-Tumor Stiffness Using Shear-Wave Elastography Histograms

Purpose: Shear-wave elastography (SWE) measures tissue elasticity using ultrasound waves. This study proposes a histogram-based SWE analysis to improve breast malignancy detection. Methods: N = 22/32 (patients/tumors) benign and n = 51/64 malignant breast tumors with histological ground truth. Colored SWE heatmaps were adjusted to a 0−180 kPa scale. Normalized, 250-binned RGB histograms were used as image descriptors based on skewness and area under curve (AUC). The histogram method was compared to conventional SWE metrics, such as (1) the qualitative 5-point scale classification and (2) average stiffness (SWEavg)/maximal tumor stiffness (SWEmax) within the tumor B-mode boundaries. Results: The SWEavg and SWEmax did not discriminate malignant lesions in this database, p > 0.05, rank sum test. RGB histograms, however, differed between malignant and benign tumors, p < 0.001, Kolmogorov−Smirnoff test. The AUC analysis of histograms revealed the reduction of soft-tissue components as a significant SWE biomarker (p = 0.03, rank sum). The diagnostic accuracy of the suggested method is still low (Se = 0.30 for Se = 0.90) and a subject for improvement in future studies. Conclusions: Histogram-based SWE quantitation improved the diagnostic accuracy for malignancy compared to conventional average SWE metrics. The sensitivity is a subject for improvement in future studies.

A High-Accuracy Detection System: Based on Transfer Learning for Apical Lesions on Periapical Radiograph

Apical Lesions, one of the most common oral diseases, can be effectively detected in daily dental examinations by a periapical radiograph (PA). In the current popular endodontic treatment, most dentists spend a lot of time manually marking the lesion area. In order to reduce the burden on dentists, this paper proposes a convolutional neural network (CNN)-based regional analysis model for spical lesions for periapical radiographs. In this study, the database was provided by dentists with more than three years of practical experience, meeting the criteria for clinical practical application. The contributions of this work are (1) an advanced adaptive threshold preprocessing technique for image segmentation, which can achieve an accuracy rate of more than 96%; (2) a better and more intuitive apical lesions symptom enhancement technique; and (3) a model for apical lesions detection with an accuracy as high as 96.21%. Compared with existing state-of-the-art technology, the proposed model has improved the accuracy by more than 5%. The proposed model has successfully improved the automatic diagnosis of apical lesions. With the help of automation, dentists can focus more on technical and medical diagnoses, such as treatment, tooth cleaning, or medical communication. This proposal has been certified by the Institutional Review Board (IRB) with the certification number 202002030B0.

A Robust Framework Combining Image Processing and Deep Learning Hybrid Model to Classify Cardiovascular Diseases Using a Limited Number of Paper-Based Complex ECG Images

Heart disease can be life-threatening if not detected and treated at an early stage. The electrocardiogram (ECG) plays a vital role in classifying cardiovascular diseases, and often physicians and medical researchers examine paper-based ECG images for cardiac diagnosis. An automated heart disease prediction system might help to classify heart diseases accurately at an early stage. This study aims to classify cardiac diseases into five classes with paper-based ECG images using a deep learning approach with the highest possible accuracy and the lowest possible time complexity. This research consists of two approaches. In the first approach, five deep learning models, InceptionV3, ResNet50, MobileNetV2, VGG19, and DenseNet201, are employed. In the second approach, an integrated deep learning model (InRes-106) is introduced, combining InceptionV3 and ResNet50. This model is developed as a deep convolutional neural network capable of extracting hidden and high-level features from images. An ablation study is conducted on the proposed model altering several components and hyperparameters, improving the performance even further. Before training the model, several image pre-processing techniques are employed to remove artifacts and enhance the image quality. Our proposed hybrid InRes-106 model performed best with a testing accuracy of 98.34%. The InceptionV3 model acquired a testing accuracy of 90.56%, the ResNet50 89.63%, the DenseNet201 88.94%, the VGG19 87.87%, and the MobileNetV2 achieved 80.56% testing accuracy. The model is trained with a k-fold cross-validation technique with different k values to evaluate the robustness further. Although the dataset contains a limited number of complex ECG images, our proposed approach, based on various image pre-processing techniques, model fine-tuning, and ablation studies, can effectively diagnose cardiac diseases.

Tobacco shred varieties classification using Multi-Scale-X-ResNet network and machine vision

The primary task in calculating the tobacco shred blending ratio is identifying the four tobacco shred types: expanded tobacco silk, cut stem, tobacco silk, and reconstituted tobacco shred. The classification precision directly affects the subsequent determination of tobacco shred components. However, the tobacco shred types, especially expanded tobacco silk and tobacco silk, have no apparent differences in macro-scale characteristics. The tobacco shreds have small size and irregular shape characteristics, creating significant challenges in their recognition and classification based on machine vision. This study provides a complete set of solutions aimed at this problem for screening tobacco shred samples, taking images, image preprocessing, establishing datasets, and identifying types. A block threshold binarization method is used for image preprocessing. Parameter setting and method performance are researched to obtain the maximum number of complete samples with acceptable execution time. ResNet50 is used as the primary classification and recognition network structure. By increasing the multi-scale structure and optimizing the number of blocks and loss function, a new tobacco shred image classification method is proposed based on the MS-X-ResNet (Multi-Scale-X-ResNet) network. Specifically, the MS-ResNet network is obtained by fusing the multi-scale Stage 3 low-dimensional and Stage 4 high-dimensional features to reduce the overfitting risk. The number of blocks in Stages 1-4 are adjusted from the original 3:4:6:3 to 3:4:N:3 (A-ResNet) and 3:3:N:3 (B-ResNet) to obtain the X-ResNet network, which improves the model's classification performance with lower complexity. The focal loss function is selected to reduce the impact of identification difficulty for different sample types on the network and improve its performance. The experimental results show that the final classification accuracy of the network on a tobacco shred dataset is 96.56%. The image recognition of a single tobacco shred requires 103 ms, achieving high classification accuracy and efficiency. The image preprocessing and deep learning algorithms for tobacco shred classification and identification proposed in this study provide a new implementation approach for the actual production and quality detection of tobacco and a new way for online real-time type identification of other agricultural products.

BreastNet18: A High Accuracy Fine-Tuned VGG16 Model Evaluated Using Ablation Study for Diagnosing Breast Cancer from Enhanced Mammography Images

Simple Summary

Breast cancer diagnosis at an early stage using mammography is important, as it assists clinical specialists in treatment planning to increase survival rates. The aim of this study is to construct an effective method to classify breast images into four classes with a low error rate. Initially, unwanted regions of mammograms are removed, the quality is enhanced, and the cancerous lesions are highlighted with different artifacts removal, noise reduction, and enhancement techniques. The number of mammograms is increased using seven augmentation techniques to deal with over-fitting and under-fitting problems. Afterwards, six fine-tuned convolution neural networks (CNNs), originally developed for other purposes, are evaluated, and VGG16 yielded the highest performance. We propose a BreastNet18 model based on the fine-tuned VGG16, changing different hyper parameters and layer structures after experimentation with our dataset. Performing an ablation study on the proposed model and selecting suitable parameter values for preprocessing algorithms increases the accuracy of our model to 98.02%, outperforming some existing state-of-the-art approaches. To analyze the performance, several performance metrics are generated and evaluated for every model and for BreastNet18. Results suggest that accuracy improvement can be obtained through image pre-processing techniques, augmentation, and ablation study. To investigate possible overfitting issues, a k-fold cross validation is carried out. To assert the robustness of the network, the model is tested on a dataset containing noisy mammograms. This may help medical specialists in efficient and accurate diagnosis and early treatment planning.

Abstract

Background: Identification and treatment of breast cancer at an early stage can reduce mortality. Currently, mammography is the most widely used effective imaging technique in breast cancer detection. However, an erroneous mammogram based interpretation may result in false diagnosis rate, as distinguishing cancerous masses from adjacent tissue is often complex and error-prone. Methods: Six pre-trained and fine-tuned deep CNN architectures: VGG16, VGG19, MobileNetV2, ResNet50, DenseNet201, and InceptionV3 are evaluated to determine which model yields the best performance. We propose a BreastNet18 model using VGG16 as foundational base, since VGG16 performs with the highest accuracy. An ablation study is performed on BreastNet18, to evaluate its robustness and achieve the highest possible accuracy. Various image processing techniques with suitable parameter values are employed to remove artefacts and increase the image quality. A total dataset of 1442 preprocessed mammograms was augmented using seven augmentation techniques, resulting in a dataset of 11,536 images. To investigate possible overfitting issues, a k-fold cross validation is carried out. The model was then tested on noisy mammograms to evaluate its robustness. Results were compared with previous studies. Results: Proposed BreastNet18 model performed best with a training accuracy of 96.72%, a validating accuracy of 97.91%, and a test accuracy of 98.02%. In contrast to this, VGGNet19 yielded test accuracy of 96.24%, MobileNetV2 77.84%, ResNet50 79.98%, DenseNet201 86.92%, and InceptionV3 76.87%. Conclusions: Our proposed approach based on image processing, transfer learning, fine-tuning, and ablation study has demonstrated a high correct breast cancer classification while dealing with a limited number of complex medical images.

Superpixel Segmentation Based on Grid Point Density Peak Clustering

Superpixel segmentation is one of the key image preprocessing steps in object recognition and detection methods. However, the over-segmentation in the smoothly connected homogenous region in an image is the key problem. That would produce redundant complex jagged textures. In this paper, the density peak clustering will be used to reduce the redundant superpixels and highlight the primary textures and contours of the salient objects. Firstly, the grid pixels are extracted as feature points, and the density of each feature point will be defined. Secondly, the cluster centers are extracted with the density peaks. Finally, all the feature points will be clustered by the density peaks. The pixel blocks, which are obtained by the above steps, are superpixels. The method is carried out in the BSDS500 dataset, and the experimental results show that the Boundary Recall (BR) and Achievement Segmentation Accuracy (ASA) are 95.0% and 96.3%, respectively. In addition, the proposed method has better performance in efficiency (30 fps). The comparison experiments show that not only do the superpixel boundaries have good adhesion to the primary textures and contours of the salient objects, but they can also effectively reduce the redundant superpixels in the homogeneous region.

An Improved Adaptive Spatial Preprocessing Method for Remote Sensing Images

Since remote sensing images are one of the main sources for people to obtain required information, the quality of the image becomes particularly important. Nevertheless, noise often inevitably exists in the image, and the targets are usually blurred by the acquisition of the imaging system, resulting in the degradation of quality of the images. In this paper, a novel preprocessing algorithm is proposed to simultaneously smooth noise and to enhance the edges, which can improve the visual quality of remote sensing images. It consists of an improved adaptive spatial filter, which is a weighted filter integrating functions of both noise removal and edge sharpness. Its processing parameters are flexible and adjustable relative to different images. The experimental results confirm that the proposed method outperforms the existing spatial algorithms both visually and quantitatively. It can play an important role in the remote sensing field in order to achieve more information of interested targets.

Robust Korean License Plate Recognition Based on Deep Neural Networks

With the rapid rise of private vehicles around the world, License Plate Recognition (LPR) plays a vital role in supporting the government to manage vehicles effectively. However, an introduction of new types of license plate (LP) or slight changes in the LP format can break previous LPR systems, as they fail to recognize the LP. Moreover, the LPR system is extremely sensitive to the conditions of the surrounding environment. Thus, this paper introduces a novel deep learning-based Korean LPR system that can effectively deal with existing challenges. The main contributions of this study include (1) a robust LPR system with the integration of three pre-processing techniques (defogging, low-light enhancement, and super-resolution) that can effectively recognize the LP under various conditions, (2) the establishment of two original Korean LPR approaches for different scenarios, including whole license plate recognition (W-LPR) and single-character license plate recognition (SC-LPR), and (3) the introduction of two Korean LPR datasets (synthetic data and real data) involving a new type of LP introduced by the Korean government. Through several experiments, the proposed LPR framework achieved the highest recognition accuracy of 98.94%.

Improvement of Image Binarization Methods Using Image Preprocessing with Local Entropy Filtering for Alphanumerical Character Recognition Purposes

Automatic text recognition from the natural images acquired in uncontrolled lighting conditions is a challenging task due to the presence of shadows hindering the shape analysis and classification of individual characters. Since the optical character recognition methods require prior image binarization, the application of classical global thresholding methods in such case makes it impossible to preserve the visibility of all characters. Nevertheless, the use of adaptive binarization does not always lead to satisfactory results for heavily unevenly illuminated document images. In this paper, the image preprocessing methodology with the use of local image entropy filtering is proposed, allowing for the improvement of various commonly used image thresholding methods, which can be useful also for text recognition purposes. The proposed approach was verified using a dataset of 140 differently illuminated document images subjected to further text recognition. Experimental results, expressed as Levenshtein distances and F-Measure values for obtained text strings, are promising and confirm the usefulness of the proposed approach.

Automatic Hotspot and Sun Glint Detection in UAV Multispectral Images

Last advances in sensors, photogrammetry and computer vision have led to high-automation levels of 3D reconstruction processes for generating dense models and multispectral orthoimages from Unmanned Aerial Vehicle (UAV) images. However, these cartographic products are sometimes blurred and degraded due to sun reflection effects which reduce the image contrast and colour fidelity in photogrammetry and the quality of radiometric values in remote sensing applications. This paper proposes an automatic approach for detecting sun reflections problems (hotspot and sun glint) in multispectral images acquired with an Unmanned Aerial Vehicle (UAV), based on a photogrammetric strategy included in a flight planning and control software developed by the authors. In particular, two main consequences are derived from the approach developed: (i) different areas of the images can be excluded since they contain sun reflection problems; (ii) the cartographic products obtained (e.g., digital terrain model, orthoimages) and the agronomical parameters computed (e.g., normalized vegetation index-NVDI) are improved since radiometric defects in pixels are not considered. Finally, an accuracy assessment was performed in order to analyse the error in the detection process, getting errors around 10 pixels for a ground sample distance (GSD) of 5 cm which is perfectly valid for agricultural applications. This error confirms that the precision in the detection of sun reflections can be guaranteed using this approach and the current low-cost UAV technology.