Applied Sciences

Research

12 pages, 3763 KiB

Open AccessArticle

Knee Osteoarthritis Classification Using 3D CNN and MRI

by Carmine Guida, Ming Zhang and Juan Shan

Appl. Sci. 2021, 11(11), 5196; https://doi.org/10.3390/app11115196 - 03 Jun 2021

Cited by 12 | Viewed by 6095

Osteoarthritis (OA) is the most common form of arthritis and can often occur in the knee. While convolutional neural networks (CNNs) have been widely used to study medical images, the application of a 3-dimensional (3D) CNN in knee OA diagnosis is limited. This [...] Read more.

Osteoarthritis (OA) is the most common form of arthritis and can often occur in the knee. While convolutional neural networks (CNNs) have been widely used to study medical images, the application of a 3-dimensional (3D) CNN in knee OA diagnosis is limited. This study utilizes a 3D CNN model to analyze sequences of knee magnetic resonance (MR) images to perform knee OA classification. An advantage of using 3D CNNs is the ability to analyze the whole sequence of 3D MR images as a single unit as opposed to a traditional 2D CNN, which examines one image at a time. Therefore, 3D features could be extracted from adjacent slices, which may not be detectable from a single 2D image. The input data for each knee were a sequence of double-echo steady-state (DESS) MR images, and each knee was labeled by the Kellgren and Lawrence (KL) grade of severity at levels 0–4. In addition to the 5-category KL grade classification, we further examined a 2-category classification that distinguishes non-OA (KL ≤ 1) from OA (KL ≥ 2) knees. Clinically, diagnosing a patient with knee OA is the ultimate goal of assigning a KL grade. On a dataset with 1100 knees, the 3D CNN model that classifies knees with and without OA achieved an accuracy of 86.5% on the validation set and 83.0% on the testing set. We further conducted a comparative study between MRI and X-ray. Compared with a CNN model using X-ray images trained from the same group of patients, the proposed 3D model with MR images achieved higher accuracy in both the 5-category classification (54.0% vs. 50.0%) and the 2-category classification (83.0% vs. 77.0%). The result indicates that MRI, with the application of a 3D CNN model, has greater potential to improve diagnosis accuracy for knee OA clinically than the currently used X-ray methods. Full article

(This article belongs to the Special Issue Recent Developments in Machine Learning Techniques for Medical Image Analysis)

► Show Figures

Figure 1

16 pages, 4115 KiB

Open AccessArticle

Segmentation of Brain Tumors from MRI Images Using Convolutional Autoencoder

by Milica M. Badža and Marko Č. Barjaktarović

Appl. Sci. 2021, 11(9), 4317; https://doi.org/10.3390/app11094317 - 10 May 2021

Cited by 19 | Viewed by 3510

Abstract

The use of machine learning algorithms and modern technologies for automatic segmentation of brain tissue increases in everyday clinical diagnostics. One of the most commonly used machine learning algorithms for image processing is convolutional neural networks. We present a new convolutional neural autoencoder [...] Read more.

The use of machine learning algorithms and modern technologies for automatic segmentation of brain tissue increases in everyday clinical diagnostics. One of the most commonly used machine learning algorithms for image processing is convolutional neural networks. We present a new convolutional neural autoencoder for brain tumor segmentation based on semantic segmentation. The developed architecture is small, and it is tested on the largest online image database. The dataset consists of 3064 T1-weighted contrast-enhanced magnetic resonance images. The proposed architecture’s performance is tested using a combination of two different data division methods, and two different evaluation methods, and by training the network with the original and augmented dataset. Using one of these data division methods, the network’s generalization ability in medical diagnostics was also tested. The best results were obtained for record-wise data division, training the network with the augmented dataset. The average accuracy classification of pixels is 99.23% and 99.28% for 5-fold cross-validation and one test, respectively, and the average dice coefficient is 71.68% and 72.87%. Considering the achieved performance results, execution speed, and subject generalization ability, the developed network has great potential for being a decision support system in everyday clinical practice. Full article

(This article belongs to the Special Issue Recent Developments in Machine Learning Techniques for Medical Image Analysis)

► Show Figures

Figure 1

19 pages, 23632 KiB

Open AccessArticle

Deep Learning Techniques Applied to Predict and Measure Finger Movement in Patients with Multiple Sclerosis

by Dmitry Viatkin, Begonya Garcia-Zapirain and Amaia Méndez Zorrilla

Appl. Sci. 2021, 11(7), 3137; https://doi.org/10.3390/app11073137 - 01 Apr 2021

Cited by 3 | Viewed by 2322

Abstract

This research focuses on the development of a system for measuring finger joint angles based on camera image and is intended for work within the field of medicine to track the movement and limits of hand mobility in multiple sclerosis. Measuring changes in [...] Read more.

This research focuses on the development of a system for measuring finger joint angles based on camera image and is intended for work within the field of medicine to track the movement and limits of hand mobility in multiple sclerosis. Measuring changes in hand mobility allows the progress of the disease and its treatment process to be monitored. A static RGB camera without depth vision was used in the system developed, with the system receiving only the image from the camera and no other input data. The research focuses on the analysis of each image in the video stream independently of other images from that stream, and 12 measured hand parameters were chosen as follows: 3 joint angles for the index finger, 3 joint angles for the middle finger, 3 joint angles for the ring finger, and 3 joint angles for the pinky finger. Convolutional neural networks were used to analyze the information received from the camera, and the research considers neural networks based on different architectures and their combinations as follows: VGG16, MobileNet, MobileNetV2, InceptionV3, DenseNet, ResNet, and convolutional pose machine. The final neural network used for image analysis was a modernized neural network based on MobileNetV2, which obtained the best mean absolute error value of 4.757 degrees. Additionally, the mean square error was 67.279 and the root mean square error was 8.202 degrees. This neural network analyzed a single image from the camera without using other sensors. For its part, the input image had a resolution of 512 by 512 pixels, and was processed by the neural network in 7–15 ms by GPU Nvidia 2080ti. The resulting neural network developed can measure finger joint angle values for a hand with non-standard parameters and positions. Full article

(This article belongs to the Special Issue Recent Developments in Machine Learning Techniques for Medical Image Analysis)

► Show Figures

Figure 1

12 pages, 1044 KiB

Open AccessArticle

Spider U-Net: Incorporating Inter-Slice Connectivity Using LSTM for 3D Blood Vessel Segmentation

by Kyeorye Lee, Leonard Sunwoo, Tackeun Kim and Kyong Joon Lee

Appl. Sci. 2021, 11(5), 2014; https://doi.org/10.3390/app11052014 - 25 Feb 2021

Cited by 16 | Viewed by 3255

Abstract

Blood vessel segmentation (BVS) of 3D medical imaging such as computed tomography and magnetic resonance angiography (MRA) is an essential task in the clinical field. Automation of 3D BVS using deep supervised learning is being researched, and U-Net-based approaches, which are considered as [...] Read more.

Blood vessel segmentation (BVS) of 3D medical imaging such as computed tomography and magnetic resonance angiography (MRA) is an essential task in the clinical field. Automation of 3D BVS using deep supervised learning is being researched, and U-Net-based approaches, which are considered as standard for medical image segmentation, are proposed a lot. However, the inherent characteristics of blood vessels, e.g., they are complex and narrow, as well as the resolution and sensitivity of the imaging modalities increases the difficulty of 3D BVS. We propose a novel U-Net-based model named Spider U-Net for 3D BVS that considers the connectivity of the blood vessels between the axial slices. To achieve this, long short-term memory (LSTM), which can capture the context of the consecutive data, is inserted into the baseline model. We also propose a data feeding strategy that augments data and makes Spider U-Net stable. Spider U-Net outperformed 2D U-Net, 3D U-Net, and the fully convolutional network-recurrent neural network (FCN-RNN) in dice coefficient score (DSC) by 0.048, 0.077, and 0.041, respectively, for our in-house brain MRA dataset and also achieved the highest DSC for two public datasets. The results imply that considering inter-slice connectivity with LSTM improves model performance in the 3D BVS task. Full article

(This article belongs to the Special Issue Recent Developments in Machine Learning Techniques for Medical Image Analysis)

► Show Figures

Figure 1

13 pages, 8928 KiB

Open AccessArticle

Fully Leveraging Deep Learning Methods for Constructing Retinal Fundus Photomontages

by Jooyoung Kim, Sojung Go, Kyoung Jin Noh, Sang Jun Park and Soochahn Lee

Appl. Sci. 2021, 11(4), 1754; https://doi.org/10.3390/app11041754 - 16 Feb 2021

Viewed by 3550

Abstract

Retinal photomontages, which are constructed by aligning and integrating multiple fundus images, are useful in diagnosing retinal diseases affecting peripheral retina. We present a novel framework for constructing retinal photomontages that fully leverage recent deep learning methods. Deep learning based object detection is [...] Read more.

Retinal photomontages, which are constructed by aligning and integrating multiple fundus images, are useful in diagnosing retinal diseases affecting peripheral retina. We present a novel framework for constructing retinal photomontages that fully leverage recent deep learning methods. Deep learning based object detection is used to define the order of image registration and blending. Deep learning based vessel segmentation is used to enhance image texture to improve registration performance within a two step image registration framework comprising rigid and non-rigid registration. Experimental evaluation demonstrates the robustness of our montage construction method with an increased amount of successfully integrated images as well as reduction of image artifacts. Full article

(This article belongs to the Special Issue Recent Developments in Machine Learning Techniques for Medical Image Analysis)

► Show Figures

Figure 1

16 pages, 2327 KiB

Open AccessArticle

Synthesize and Segment: Towards Improved Catheter Segmentation via Adversarial Augmentation

by Ihsan Ullah, Philip Chikontwe, Hongsoo Choi, Chang Hwan Yoon and Sang Hyun Park

Appl. Sci. 2021, 11(4), 1638; https://doi.org/10.3390/app11041638 - 11 Feb 2021

Cited by 3 | Viewed by 2869

Abstract

Automatic catheter and guidewire segmentation plays an important role in robot-assisted interventions that are guided by fluoroscopy. Existing learning based methods addressing the task of segmentation or tracking are often limited by the scarcity of annotated samples and difficulty in data collection. In [...] Read more.

Automatic catheter and guidewire segmentation plays an important role in robot-assisted interventions that are guided by fluoroscopy. Existing learning based methods addressing the task of segmentation or tracking are often limited by the scarcity of annotated samples and difficulty in data collection. In the case of deep learning based methods, the demand for large amounts of labeled data further impedes successful application. We propose a synthesize and segment approach with plug in possibilities for segmentation to address this. We show that an adversarially learned image-to-image translation network can synthesize catheters in X-ray fluoroscopy enabling data augmentation in order to alleviate a low data regime. To make realistic synthesized images, we train the translation network via a perceptual loss coupled with similarity constraints. Then existing segmentation networks are used to learn accurate localization of catheters in a semi-supervised setting with the generated images. The empirical results on collected medical datasets show the value of our approach with significant improvements over existing translation baseline methods. Full article

(This article belongs to the Special Issue Recent Developments in Machine Learning Techniques for Medical Image Analysis)

► Show Figures

Figure 1

12 pages, 9206 KiB

Open AccessArticle

Quantitative Assessment of Shape Deformation of Regional Cranial Bone for Evaluation of Surgical Effect in Patients with Craniosynostosis

by Min Jin Lee, Helen Hong and Kyu Won Shim

Appl. Sci. 2021, 11(3), 990; https://doi.org/10.3390/app11030990 - 22 Jan 2021

Cited by 1 | Viewed by 3828

Abstract

Surgery in patients with craniosynostosis is a common treatment to correct the deformed skull shape, and it is necessary to verify the surgical effect of correction on the regional cranial bone. We propose a quantification method for evaluating surgical effects on regional cranial [...] Read more.

Surgery in patients with craniosynostosis is a common treatment to correct the deformed skull shape, and it is necessary to verify the surgical effect of correction on the regional cranial bone. We propose a quantification method for evaluating surgical effects on regional cranial bones by comparing preoperative and postoperative skull shapes. To divide preoperative and postoperative skulls into two frontal bones, two parietal bones, and the occipital bone, and to estimate the shape deformation of regional cranial bones between the preoperative and postoperative skulls, an age-matched mean-normal skull surface model already divided into five bones is deformed into a preoperative skull, and a deformed mean-normal skull surface model is redeformed into a postoperative skull. To quantify the degree of the expansion and reduction of regional cranial bones after surgery, expansion and reduction indices of the five cranial bones are calculated using the deformable registration as deformation information. The proposed quantification method overcomes the quantification difficulty when using the traditional cephalic index(CI) by analyzing regional cranial bones and provides useful information for quantifying the surgical effects of craniosynostosis patients with symmetric and asymmetric deformities. Full article

(This article belongs to the Special Issue Recent Developments in Machine Learning Techniques for Medical Image Analysis)

► Show Figures

Figure 1

17 pages, 5682 KiB

Open AccessArticle

Robust Resolution-Enhanced Prostate Segmentation in Magnetic Resonance and Ultrasound Images through Convolutional Neural Networks

by Oscar J. Pellicer-Valero, Victor Gonzalez-Perez, Juan Luis Casanova Ramón-Borja, Isabel Martín García, María Barrios Benito, Paula Pelechano Gómez, José Rubio-Briones, María José Rupérez and José D. Martín-Guerrero

Appl. Sci. 2021, 11(2), 844; https://doi.org/10.3390/app11020844 - 18 Jan 2021

Cited by 2 | Viewed by 2521

Abstract

Prostate segmentations are required for an ever-increasing number of medical applications, such as image-based lesion detection, fusion-guided biopsy and focal therapies. However, obtaining accurate segmentations is laborious, requires expertise and, even then, the inter-observer variability remains high. In this paper, a robust, accurate [...] Read more.

Prostate segmentations are required for an ever-increasing number of medical applications, such as image-based lesion detection, fusion-guided biopsy and focal therapies. However, obtaining accurate segmentations is laborious, requires expertise and, even then, the inter-observer variability remains high. In this paper, a robust, accurate and generalizable model for Magnetic Resonance (MR) and three-dimensional (3D) Ultrasound (US) prostate image segmentation is proposed. It uses a densenet-resnet-based Convolutional Neural Network (CNN) combined with techniques such as deep supervision, checkpoint ensembling and Neural Resolution Enhancement. The MR prostate segmentation model was trained with five challenging and heterogeneous MR prostate datasets (and two US datasets), with segmentations from many different experts with varying segmentation criteria. The model achieves a consistently strong performance in all datasets independently (mean Dice Similarity Coefficient -DSC- above 0.91 for all datasets except for one), outperforming the inter-expert variability significantly in MR (mean DSC of 0.9099 vs. 0.8794). When evaluated on the publicly available Promise12 challenge dataset, it attains a similar performance to the best entries. In summary, the model has the potential of having a significant impact on current prostate procedures, undercutting, and even eliminating, the need of manual segmentations through improvements in terms of robustness, generalizability and output resolution. Full article

(This article belongs to the Special Issue Recent Developments in Machine Learning Techniques for Medical Image Analysis)

► Show Figures

Graphical abstract

14 pages, 2916 KiB

Open AccessArticle

Topology-Aware Retinal Artery–Vein Classification via Deep Vascular Connectivity Prediction

by Seung Yeon Shin, Soochahn Lee, Il Dong Yun and Kyoung Mu Lee

Appl. Sci. 2021, 11(1), 320; https://doi.org/10.3390/app11010320 - 31 Dec 2020

Cited by 4 | Viewed by 2376

Abstract

Retinal artery–vein (AV) classification is a prerequisite for quantitative analysis of retinal vessels, which provides a biomarker for neurologic, cardiac, and systemic diseases, as well as ocular diseases. Although convolutional neural networks have presented remarkable performance on AV classification, it often comes with [...] Read more.

Retinal artery–vein (AV) classification is a prerequisite for quantitative analysis of retinal vessels, which provides a biomarker for neurologic, cardiac, and systemic diseases, as well as ocular diseases. Although convolutional neural networks have presented remarkable performance on AV classification, it often comes with a topological error, like an abrupt class flipping on the same vessel segment or a weakness for thin vessels due to their indistinct appearances. In this paper, we present a new method for AV classification where the underlying vessel topology is estimated to give consistent prediction along the actual vessel structure. We cast the vessel topology estimation as iterative vascular connectivity prediction, which is implemented as deep-learning-based pairwise classification. In consequence, a whole vessel graph is separated into sub-trees, and each of them is classified as an artery or vein in whole via a voting scheme. The effectiveness and efficiency of the proposed method is validated by conducting experiments on two retinal image datasets acquired using different imaging techniques called DRIVE and IOSTAR. Full article

(This article belongs to the Special Issue Recent Developments in Machine Learning Techniques for Medical Image Analysis)

► Show Figures

Figure 1

13 pages, 3085 KiB

Open AccessArticle

Screening Patients with Early Stage Parkinson’s Disease Using a Machine Learning Technique: Measuring the Amount of Iron in the Basal Ganglia

by Seon Lee, Se-Hong Oh, Sun-Won Park, Chaewon Shin, Jeehun Kim, Jung-Hyo Rhim, Jee-Young Lee and Joon-Yul Choi

Appl. Sci. 2020, 10(23), 8732; https://doi.org/10.3390/app10238732 - 06 Dec 2020

Viewed by 1976

Abstract

The purpose of this study was to determine whether a support vector machine (SVM) model based on quantitative susceptibility mapping (QSM) can be used to differentiate iron accumulation in the deep grey matter of early Parkinson’s disease (PD) patients from healthy controls (HC) [...] Read more.

The purpose of this study was to determine whether a support vector machine (SVM) model based on quantitative susceptibility mapping (QSM) can be used to differentiate iron accumulation in the deep grey matter of early Parkinson’s disease (PD) patients from healthy controls (HC) and Non-Motor Symptoms Scale (NMSS) scores in early PD patients. QSM values on magnetic resonance imaging (MRI) were obtained for 24 early PD patients and 27 age-matched HCs. The mean QSM values in deep grey matter areas were used to construct SVM and logistic regression (LR) models to differentiate between early PD patients and HCs. Additional SVM and LR models were constructed to differentiate between low and high NMSS scores groups. A paired t-test was used to assess the classification results. For the differentiation between early PD patients and HCs, SVM had an accuracy of 0.79 ± 0.07, and LR had an accuracy of 0.73 ± 0.03 (p = 0.027). SVM for NMSS classification had a fairly high accuracy of 0.79 ± 0.03, while LR had 0.76 ± 0.04. An SVM model based on QSM offers competitive accuracy for screening early PD patients and evaluates non-motor symptoms, which may offer clinicians the ability to assess the progression of motor symptoms in the patient population. Full article

(This article belongs to the Special Issue Recent Developments in Machine Learning Techniques for Medical Image Analysis)

► Show Figures

Figure 1

16 pages, 6743 KiB

Open AccessArticle

A Novel Hybrid Machine Learning Classification for the Detection of Bruxism Patients Using Physiological Signals

by Md Belal Bin Heyat, Faijan Akhtar, Asif Khan, Alam Noor, Bilel Benjdira, Yumna Qamar, Syed Jafar Abbas and Dakun Lai

Appl. Sci. 2020, 10(21), 7410; https://doi.org/10.3390/app10217410 - 22 Oct 2020

Cited by 44 | Viewed by 5948

Abstract

Bruxism is a sleep disorder in which the patient clinches and gnashes their teeth. Bruxism detection using traditional methods is time-consuming, cumbersome, and expensive. Therefore, an automatic tool to detect this disorder will alleviate the doctor workload and give valuable help to patients. [...] Read more.

Bruxism is a sleep disorder in which the patient clinches and gnashes their teeth. Bruxism detection using traditional methods is time-consuming, cumbersome, and expensive. Therefore, an automatic tool to detect this disorder will alleviate the doctor workload and give valuable help to patients. In this paper, we targeted this goal and designed an automatic method to detect bruxism from the physiological signals using a novel hybrid classifier. We began with data collection. Then, we performed the analysis of the physiological signals and the estimation of the power spectral density. After that, we designed the novel hybrid classifier to enable the detection of bruxism based on these data. The classification of the subjects into “healthy” or “bruxism” from the electroencephalogram channel (C4-A1) obtained a maximum specificity of 92% and an accuracy of 94%. Besides, the classification of the sleep stages such as the wake (w) stage and rapid eye movement (REM) stage from the electrocardiogram channel (ECG1-ECG2) obtained a maximum specificity of 86% and an accuracy of 95%. The combined bruxism classification and the sleep stages classification from the electroencephalogram channel (C4-P4) obtained a maximum specificity of 90% and an accuracy of 97%. The results show that more accurate bruxism detection is achieved by exploiting the electroencephalogram signal (C4-P4). The present work can be applied for home monitoring systems for bruxism detection. Full article

(This article belongs to the Special Issue Recent Developments in Machine Learning Techniques for Medical Image Analysis)

► Show Figures

Figure 1

13 pages, 3262 KiB

Open AccessArticle

Gradually Applying Weakly Supervised and Active Learning for Mass Detection in Breast Ultrasound Images

by JooYeol Yun, JungWoo Oh and IlDong Yun

Appl. Sci. 2020, 10(13), 4519; https://doi.org/10.3390/app10134519 - 29 Jun 2020

Cited by 3 | Viewed by 2387

Abstract

We propose a method for effectively utilizing weakly annotated image data in an object detection tasks of breast ultrasound images. Given the problem setting where a small, strongly annotated dataset and a large, weakly annotated dataset with no bounding box information are available, [...] Read more.

We propose a method for effectively utilizing weakly annotated image data in an object detection tasks of breast ultrasound images. Given the problem setting where a small, strongly annotated dataset and a large, weakly annotated dataset with no bounding box information are available, training an object detection model becomes a non-trivial problem. We suggest a controlled weight for handling the effect of weakly annotated images in a two stage object detection model. We also present a subsequent active learning scheme for safely assigning weakly annotated images a strong annotation using the trained model. Experimental results showed a 24% point increase in correct localization (CorLoc) measure, which is the ratio of correctly localized and classified images, by assigning the properly controlled weight. Performing active learning after a model is trained showed an additional increase in CorLoc. We tested the proposed method on the Stanford Dog datasets to assure that it can be applied to general cases, where strong annotations are insufficient to obtain resembling results. The presented method showed that higher performance is achievable with lesser annotation effort. Full article

(This article belongs to the Special Issue Recent Developments in Machine Learning Techniques for Medical Image Analysis)

► Show Figures

Figure 1

14 pages, 3699 KiB

Open AccessArticle

Improved U-Net: Fully Convolutional Network Model for Skin-Lesion Segmentation

by Karshiev Sanjar, Olimov Bekhzod, Jaeil Kim, Jaesoo Kim, Anand Paul and Jeonghong Kim

Appl. Sci. 2020, 10(10), 3658; https://doi.org/10.3390/app10103658 - 25 May 2020

Cited by 20 | Viewed by 4257

Abstract

The early and accurate diagnosis of skin cancer is crucial for providing patients with advanced treatment by focusing medical personnel on specific parts of the skin. Networks based on encoder–decoder architectures have been effectively implemented for numerous computer-vision applications. U-Net, one of CNN [...] Read more.

The early and accurate diagnosis of skin cancer is crucial for providing patients with advanced treatment by focusing medical personnel on specific parts of the skin. Networks based on encoder–decoder architectures have been effectively implemented for numerous computer-vision applications. U-Net, one of CNN architectures based on the encoder–decoder network, has achieved successful performance for skin-lesion segmentation. However, this network has several drawbacks caused by its upsampling method and activation function. In this paper, a fully convolutional network and its architecture are proposed with a modified U-Net, in which a bilinear interpolation method is used for upsampling with a block of convolution layers followed by parametric rectified linear-unit non-linearity. To avoid overfitting, a dropout is applied after each convolution block. The results demonstrate that our recommended technique achieves state-of-the-art performance for skin-lesion segmentation with 94% pixel accuracy and a 88% dice coefficient, respectively. Full article

(This article belongs to the Special Issue Recent Developments in Machine Learning Techniques for Medical Image Analysis)

► Show Figures

Figure 1

Journal Menu

Journal Browser

Recent Developments in Machine Learning Techniques for Medical Image Analysis

Share This Special Issue

Special Issue Editors

Special Issue Information

Keywords

Published Papers (13 papers)

Research

Further Information

Guidelines

MDPI Initiatives

Follow MDPI