Selection of pre-trained weights for transfer learning in automated cytomegalovirus retinitis classification.

Cytomegalovirus retinitis (CMVR) is a significant cause of vision loss. Regular screening is crucial but challenging in resource-limited settings. A convolutional neural network is a state-of-the-art deep learning technique to generate automatic diagnoses from retinal images. However, there are limited numbers of CMVR images to train the model properly. Transfer learning (TL) is a strategy to train a model with a scarce dataset. This study explores the efficacy of TL with different pre-trained weights for automated CMVR classification using retinal images. We utilised a dataset of 955 retinal images (524 CMVR and 431 normal) from Siriraj Hospital, Mahidol University, collected between 2005 and 2015. Images were processed using Kowa VX-10i or VX-20 fundus cameras and augmented for training. We employed DenseNet121 as a backbone model, comparing the performance of TL with weights pre-trained on ImageNet, APTOS2019, and CheXNet datasets. The models were evaluated based on accuracy, loss, and other performance metrics, with the depth of fine-tuning varied across different pre-trained weights. The study found that TL significantly enhances model performance in CMVR classification. The best results were achieved with weights sequentially transferred from ImageNet to APTOS2019 dataset before application to our CMVR dataset. This approach yielded the highest mean accuracy (0.99) and lowest mean loss (0.04), outperforming other methods. The class activation heatmaps provided insights into the model's decision-making process. The model with APTOS2019 pre-trained weights offered the best explanation and highlighted the pathologic lesions resembling human interpretation. Our findings demonstrate the potential of sequential TL in improving the accuracy and efficiency of CMVR diagnosis, particularly in settings with limited data availability. They highlight the importance of domain-specific pre-training in medical image classification. This approach streamlines the diagnostic process and paves the way for broader applications in automated medical image analysis, offering a scalable solution for early disease detection.

Deep Upscale U-Net for automatic tongue segmentation.

In a treatment or diagnosis related to oral health conditions such as oral cancer and oropharyngeal cancer, an investigation of tongue's movements is a major part. In an automatic measurement of such movement, it must first start with a task of tongue segmentation. This paper proposes a solution of tongue segmentation based on a decoder-encoder CNN-based structure i.e., U-Net. However, it could suffer from a problem of feature loss in deep layers. This paper proposes a Deep Upscale U-Net (DU-UNET). An additional up-sampling of the feature map from a contracting path is concatenated to an upper layer of an expansive path, based on an original U-Net structure. The segmentation model is constructed by training DU-UNET on the two publicly available datasets, and transferred to the self-collected dataset of tongue images with five tongue postures which were recorded at a far distance from a camera under a real-world scenario. The proposed DU-UNET outperforms the other existing methods in our literature reviews, with accuracy of 99.2%, mean IoU of 97.8%, Dice score of 96.8%, and Jaccard score of 96.8%.

Factors Influencing Telehealth Service Use and Health Outcomes in Patients Undergoing Continuous Ambulatory Peritoneal Dialysis: Cross-Sectional Study.

BACKGROUND: Several studies have demonstrated the efficacy and user acceptance of telehealth in managing patients with chronic conditions, including continuous ambulatory peritoneal dialysis (CAPD). However, the rates of telehealth service use in various patient groups have been low and have declined over time, which may affect important health outcomes. Telehealth service use in patients undergoing CAPD has been recognized as a key challenge that needs to be examined further. OBJECTIVE: This study aimed to explore the rates of telehealth service use over 4 months, identify factors influencing its use, and examine the relationship between telehealth service use and health outcomes in Thai people undergoing CAPD. METHODS: This cross-sectional study, which was a part of a pragmatic randomized controlled trial study, was conducted at a dialysis center in Bangkok, Thailand. The study included patients who were undergoing CAPD. These patients were randomly enrolled in the intervention group to receive telehealth service and additional standard care for 4 months. Data were collected using self-reported questionnaires, including a demographic form, Functional, Communicative, and Critical Health Literacy Scale, Perceived Usefulness Questionnaire, Brief Illness Perception Questionnaire, Patient-Doctor Relationship Questionnaire, and Kidney Disease Quality of Life 36 Questionnaire. Additionally, Google Analytics was used to obtain data on the actual use of the telehealth service. These data were analyzed using descriptive statistics, repeated-measures ANOVA, and regression analyses. RESULTS: A total of 159 patients were included in this study. The mean rate of telehealth service use throughout the period of 4 months was 62.06 (SD 49.71) times. The rate of telehealth service use was the highest in the first month (mean 23.48, SD 16.28 times) and the lowest in the third month (mean 11.09, SD 11.48 times). Independent variables explained 27.6% of the sample variances in telehealth service use. Older age (ß=.221; P=.002), higher perceived usefulness (ß=.414; P<.001), unemployment (ß=-.155; P=.03), and positive illness perception (ß=-.205; P=.004) were associated with a significantly higher rate of telehealth service use. Regarding the relationship between telehealth service use and health outcomes, higher rates of telehealth service use were linked to better quality of life (ß=.241; P=.002) and lower peritonitis (odds ratio 0.980, 95% CI 0.962-0.997; P=.03). CONCLUSIONS: This study provides valuable insights into factors impacting telehealth service use, which in turn affect health outcomes in patients undergoing CAPD.

Utilization of images and three-dimensional custom-made nostril retainer fabricate for patients with cleft lip and cleft lip nose deformities at Siriraj Hospital: preliminary phase.

A prospective study utilizing image analysis to assess nostril openings in post-operative patients with cleft lip and cleft lip nose deformities. This preliminary study seeks to employ two-dimensional (2D) images to fabricate a custom-made nostril retainer. This study was performed at Faculty of Medicine, Siriraj Hospital, Mahidol University, Thailand. This study included 30 healthy volunteers and 15 patients with cleft lip and cleft lip nose deformities. The nostril opening width and height for all participants were measured, and photographs were taken. An image analysis application was used to fabricate a three-dimensional (3D) custom-made nostril retainer. The mean differences between the direct measurements of the nostril aperture and the measurements obtained through the program did not exceed 2 mm in terms of nostril height, width, or columella. Two-dimensional photographs can be used to create a custom-made, three-dimensional nostril retainer. This retainer allows post-operative patients to maintain their nares without needing to visit the hospital, thereby reducing the cost of care.

Encoder-decoder network with RMP for tongue segmentation.

Tongue and its movements can be used for several medical-related tasks, such as identifying a disease and tracking a rehabilitation. To be able to focus on a tongue region, the tongue segmentation is needed to compute a region of interest for a further analysis. This paper proposes an encoder-decoder CNN-based architecture for segmenting a tongue in an image. The encoder module is mainly used for the tongue feature extraction, while the decoder module is used to reconstruct a segmented tongue from the extracted features based on training images. In addition, the residual multi-kernel pooling (RMP) is also applied into the proposed network to help in encoding multiple scales of the features. The proposed method is evaluated on two publicly available datasets under a scenario of front view and one tongue posture. It is then tested on a newly collected dataset of five tongue postures. The reported performances show that the proposed method outperforms existing methods in the literature. In addition, the re-training process could improve applying the trained model on unseen dataset, which would be a necessary step of applying the trained model on the real-world scenario.

A retrospective study of 3D deep learning approach incorporating coordinate information to improve the segmentation of pre- and post-operative abdominal aortic aneurysm.

Abdominal aortic aneurysm (AAA) is one of the most common diseases worldwide. 3D segmentation of AAA provides useful information for surgical decisions and follow-up treatment. However, existing segmentation methods are time consuming and not practical in routine use. In this article, the segmentation task will be addressed automatically using a deep learning based approach which has been proved to successfully solve several medical imaging problems with excellent performances. This article therefore proposes a new solution of AAA segmentation using deep learning in a type of 3D convolutional neural network (CNN) architecture that also incorporates coordinate information. The tested CNNs are UNet, AG-DSV-UNet, VNet, ResNetMed and DenseVoxNet. The 3D-CNNs are trained with a dataset of high resolution (256 × 256) non-contrast and post-contrast CT images containing 64 slices from each of 200 patients. The dataset consists of contiguous CT slices without augmentation and no post-processing step. The experiments show that incorporation of coordinate information improves the segmentation results. The best accuracies on non-contrast and contrast-enhanced images have average dice scores of 97.13% and 96.74%, respectively. Transfer learning from a pre-trained network of a pre-operative dataset to post-operative endovascular aneurysm repair (EVAR) was also performed. The segmentation accuracy of post-operative EVAR using transfer learning on non-contrast and contrast-enhanced CT datasets achieved the best dice scores of 94.90% and 95.66%, respectively.

Multi-level pooling encoder-decoder convolution neural network for MRI reconstruction.

MRI reconstruction is one of the critical processes of MRI machines, along with the acquisition. Due to a slow processing time of signal acquiring, parallel imaging and reconstruction techniques are applied for acceleration. To accelerate the acquisition process, fewer raw data are sampled simultaneously with all RF coils acquisition. Then, the reconstruction uses under-sampled data from all RF coils to restore the final MR image that resembles the fully sampled MR image. These processes have been a traditional procedure inside the MRI system since the invention of the multi-coils MRI machine. This paper proposes the deep learning technique with a lightweight network. The deep neural network is capable of generating the high-quality reconstructed MR image with a high peak signal-to-noise ratio (PSNR). This also opens a high acceleration factor for MR data acquisition. The lightweight network is called Multi-Level Pooling Encoder-Decoder Net (MLPED Net). The proposed network outperforms the traditional encoder-decoder networks on 4-fold acceleration with a significant margin on every evaluation metric. The network can be trained end-to-end, and it is a lightweight structure that can reduce training time significantly. Experimental results are based on a publicly available MRI Knee dataset from the fastMRI competition.

Gait recognition using a few gait frames.

Gait has been deemed as an alternative biometric in video-based surveillance applications, since it can be used to recognize individuals from a far distance without their interaction and cooperation. Recently, many gait recognition methods have been proposed, aiming at reducing the influence caused by exterior factors. However, most of these methods are developed based on sufficient input gait frames, and their recognition performance will sharply decrease if the frame number drops. In the real-world scenario, it is impossible to always obtain a sufficient number of gait frames for each subject due to many reasons, e.g., occlusion and illumination. Therefore, it is necessary to improve the gait recognition performance when the available gait frames are limited. This paper starts with three different strategies, aiming at producing more input frames and eliminating the generalization error cause by insufficient input data. Meanwhile, a two-branch network is also proposed in this paper to formulate robust gait representations from the original and new generated input gait frames. According to our experiments, under the limited gait frames being used, it was verified that the proposed method can achieve a reliable performance for gait recognition.

COVID-19 detection and heatmap generation in chest x-ray images.

Purpose: The outbreak of COVID-19 or coronavirus was first reported in 2019. It has widely and rapidly spread around the world. The detection of COVID-19 cases is one of the important factors to stop the epidemic, because the infected individuals must be quarantined. One reliable way to detect COVID-19 cases is using chest x-ray images, where signals of the infection are located in lung areas. We propose a solution to automatically classify COVID-19 cases in chest x-ray images. Approach: The ResNet-101 architecture is adopted as the main network with more than 44 millions parameters. The whole net is trained using the large size of 1500 × 1500 x-ray images. The heatmap under the region of interest of segmented lung is constructed to visualize and emphasize signals of COVID-19 in each input x-ray image. Lungs are segmented using the pretrained U-Net. The confidence score of being COVID-19 is also calculated for each classification result. Results: The proposed solution is evaluated based on COVID-19 and normal cases. It is also tested on unseen classes to validate a regularization of the constructed model. They include other normal cases where chest x-ray images are normal without any disease but with some small remarks, and other abnormal cases where chest x-ray images are abnormal with some other diseases containing remarks similar to COVID-19. The proposed method can achieve the sensitivity, specificity, and accuracy of 97%, 98%, and 98%, respectively. Conclusions: It can be concluded that the proposed solution can detect COVID-19 in a chest x-ray image. The heatmap and confidence score of the detection are also demonstrated, such that users or human experts can use them for a final diagnosis in practical usages.

A 3D deep learning approach to epicardial fat segmentation in non-contrast and post-contrast cardiac CT images.

Epicardial fat (ECF) is localized fat surrounding the heart muscle or myocardium and enclosed by the thin-layer pericardium membrane. Segmenting the ECF is one of the most difficult medical image segmentation tasks. Since the epicardial fat is infiltrated into the groove between cardiac chambers and is contiguous with cardiac muscle, segmentation requires location and voxel intensity. Recently, deep learning methods have been effectively used to solve medical image segmentation problems in several domains with state-of-the-art performance. This paper presents a novel approach to 3D segmentation of ECF by integrating attention gates and deep supervision into the 3D U-Net deep learning architecture. The proposed method shows significant improvement of the segmentation performance, when compared with standard 3D U-Net. The experiments show excellent performance on non-contrast CT datasets with average Dice scores of 90.06%. Transfer learning from a pre-trained model of a non-contrast CT to contrast-enhanced CT dataset was also performed. The segmentation accuracy on the contrast-enhanced CT dataset achieved a Dice score of 88.16%.

Pyramid graph cut: Integrating intensity and gradient information for grayscale medical image segmentation.

Segmentation of grayscale medical images is challenging because of the similarity of pixel intensities and poor gradient strength between adjacent regions. The existing image segmentation approaches based on either intensity or gradient information alone often fail to produce accurate segmentation results. Previous approaches in the literature have approached the problem by embedded or sequential integration of different information types to improve the performance of the image segmentation on specific tasks. However, an effective combination or integration of such information is difficult to implement and not sufficiently generic for closely related tasks. Integration of the two information sources in a single graph structure is a potentially more effective way to solve the problem. In this paper we introduce a novel technique for grayscale medical image segmentation called pyramid graph cut, which combines intensity and gradient sources of information in a pyramid-shaped graph structure using a single source node and multiple sink nodes. The source node, which is the top of the pyramid graph, embeds intensity information into its linked edges. The sink nodes, which are the base of the pyramid graph, embed gradient information into their linked edges. The min-cut uses intensity information and gradient information, depending on which one is more useful or has a higher influence in each cutting location of each iteration. The experimental results demonstrate the effectiveness of the proposed method over intensity-based segmentation alone (i.e. Gaussian mixture model) and gradient-based segmentation alone (i.e. distance regularized level set evolution) on grayscale medical image datasets, including the public 3DIRCADb-01 dataset. The proposed method archives excellent segmentation results on the sample CT of abdominal aortic aneurysm, MRI of liver tumor and US of liver tumor, with dice scores of 90.49±5.23%, 88.86±11.77%, 90.68±2.45%, respectively.

A 'what-if' scenario: Nipah virus attacks pig trade chains in Thailand.

BACKGROUND: Nipah virus (NiV) is a fatal zoonotic agent that was first identified amongst pig farmers in Malaysia in 1998, in an outbreak that resulted in 105 fatal human cases. That epidemic arose from a chain of infection, initiating from bats to pigs, and which then spilled over from pigs to humans. In Thailand, bat-pig-human communities can be observed across the country, particularly in the central plain. The present study therefore aimed to identify high-risk areas for potential NiV outbreaks and to model how the virus is likely to spread. Multi-criteria decision analysis (MCDA) and weighted linear combination (WLC) were employed to produce the NiV risk map. The map was then overlaid with the nationwide pig movement network to identify the index subdistricts in which NiV may emerge. Subsequently, susceptible-exposed-infectious-removed (SEIR) modeling was used to simulate NiV spread within each subdistrict, and network modeling was used to illustrate how the virus disperses across subdistricts. RESULTS: Based on the MCDA and pig movement data, 14 index subdistricts with a high-risk of NiV emergence were identified. We found in our infectious network modeling that the infected subdistricts clustered in, or close to the central plain, within a range of 171 km from the source subdistricts. However, the virus may travel as far as 528.5 km (R0 = 5). CONCLUSIONS: In conclusion, the risk of NiV dissemination through pig movement networks in Thailand is low but not negligible. The risk areas identified in our study can help the veterinary authority to allocate financial and human resources to where preventive strategies, such as pig farm regionalization, are required and to contain outbreaks in a timely fashion once they occur.

3D segmentation of exterior wall surface of abdominal aortic aneurysm from CT images using variable neighborhood search.

A 3D model of abdominal aortic aneurysm (AAA) can provide useful anatomical information for clinical management and simulation. Thin-slice contiguous computed tomographic (CT) angiography is the best source of medical images for construction of 3D models, which requires segmentation of AAA in the images. Existing methods for segmentation of AAA rely on either manual process or 2D segmentation in each 2D CT slide. However, a traditional manual segmentation is a time consuming process which is not practical for routine use. The construction of a 3D model from 2D segmentation of each CT slice is not a fully satisfactory solution due to rough contours that can occur because of lack of constraints among segmented slices, as well as missed segmentation slices. To overcome such challenges, this paper proposes the 3D segmentation of AAA using the concept of variable neighborhood search by iteratively alternating between two different segmentation techniques in the two different 3D search spaces of voxel intensity and voxel gradient. The segmentation output of each method is used as the initial contour to the other method in each iteration. By alternating between search spaces, the technique can escape local minima that naturally occur in each search space. Also, the 3D search spaces provide more constraints across CT slices, when compared with the 2D search spaces in individual CT slices. The proposed method is evaluated with 10 easy and 10 difficult cases of AAA. The results show that the proposed 3D segmentation technique achieves the outstanding segmentation accuracy with an average dice similarity value (DSC) of 91.88%, when compared to the other methods using the same dataset, which are the 2D proposed method, classical graph cut, distance regularized level set evolution, and registration based geometric active contour with the DSCs of 87.57 ± 4.52%, 72.47 ± 8.11%, 58.50 ± 8.86% and 76.21 ± 10.49%, respectively.

Hard exudates segmentation based on learned initial seeds and iterative graph cut.

(Background and Objective): The occurrence of hard exudates is one of the early signs of diabetic retinopathy which is one of the leading causes of the blindness. Many patients with diabetic retinopathy lose their vision because of the late detection of the disease. Thus, this paper is to propose a novel method of hard exudates segmentation in retinal images in an automatic way. (Methods): The existing methods are based on either supervised or unsupervised learning techniques. In addition, the learned segmentation models may often cause miss-detection and/or fault-detection of hard exudates, due to the lack of rich characteristics, the intra-variations, and the similarity with other components in the retinal image. Thus, in this paper, the supervised learning based on the multilayer perceptron (MLP) is only used to identify initial seeds with high confidences to be hard exudates. Then, the segmentation is finalized by unsupervised learning based on the iterative graph cut (GC) using clusters of initial seeds. Also, in order to reduce color intra-variations of hard exudates in different retinal images, the color transfer (CT) is applied to normalize their color information, in the pre-processing step. (Results): The experiments and comparisons with the other existing methods are based on the two well-known datasets, e_ophtha EX and DIARETDB1. It can be seen that the proposed method outperforms the other existing methods in the literature, with the sensitivity in the pixel-level of 0.891 for the DIARETDB1 dataset and 0.564 for the e_ophtha EX dataset. The cross datasets validation where the training process is performed on one dataset and the testing process is performed on another dataset is also evaluated in this paper, in order to illustrate the robustness of the proposed method. (Conclusions): This newly proposed method integrates the supervised learning and unsupervised learning based techniques. It achieves the improved performance, when compared with the existing methods in the literature. The robustness of the proposed method for the scenario of cross datasets could enhance its practical usage. That is, the trained model could be more practical for unseen data in the real-world situation, especially when the capturing environments of training and testing images are not the same.

Outer Wall Segmentation of Abdominal Aortic Aneurysm by Variable Neighborhood Search Through Intensity and Gradient Spaces.

Aortic aneurysm segmentation remains a challenge. Manual segmentation is a time-consuming process which is not practical for routine use. To address this limitation, several automated segmentation techniques for aortic aneurysm have been developed, such as edge detection-based methods, partial differential equation methods, and graph partitioning methods. However, automatic segmentation of aortic aneurysm is difficult due to high pixel similarity to adjacent tissue and a lack of color information in the medical image, preventing previous work from being applicable to difficult cases. This paper uses uses a variable neighborhood search that alternates between intensity-based and gradient-based segmentation techniques. By alternating between intensity and gradient spaces, the search can escape from local optima of each space. The experimental results demonstrate that the proposed method outperforms the other existing segmentation methods in the literature, based on measurements of dice similarity coefficient and jaccard similarity coefficient at the pixel level. In addition, it is shown to perform well for cases that are difficult to segment.

Recognizing Gaits Across Views Through Correlated Motion Co-Clustering.

Human gait is an important biometric feature, which can be used to identify a person remotely. However, view change can cause significant difficulties for gait recognition because it will alter available visual features for matching substantially. Moreover, it is observed that different parts of gait will be affected differently by view change. By exploring relations between two gaits from two different views, it is also observed that a part of gait in one view is more related to a typical part than any other parts of gait in another view. A new method proposed in this paper considers such variance of correlations between gaits across views that is not explicitly analyzed in the other existing methods. In our method, a novel motion co-clustering is carried out to partition the most related parts of gaits from different views into the same group. In this way, relationships between gaits from different views will be more precisely described based on multiple groups of the motion co-clustering instead of a single correlation descriptor. Inside each group, a linear correlation between gait information across views is further maximized through canonical correlation analysis (CCA). Consequently, gait information in one view can be projected onto another view through a linear approximation under the trained CCA subspaces. In the end, a similarity between gaits originally recorded from different views can be measured under the approximately same view. Comprehensive experiments based on widely adopted gait databases have shown that our method outperforms the state-of-the-art.

Gait recognition across various walking speeds using higher order shape configuration based on a differential composition model.

Gait has been known as an effective biometric feature to identify a person at a distance. However, variation of walking speeds may lead to significant changes to human walking patterns. It causes many difficulties for gait recognition. A comprehensive analysis has been carried out in this paper to identify such effects. Based on the analysis, Procrustes shape analysis is adopted for gait signature description and relevant similarity measurement. To tackle the challenges raised by speed change, this paper proposes a higher order shape configuration for gait shape description, which deliberately conserves discriminative information in the gait signatures and is still able to tolerate the varying walking speed. Instead of simply measuring the similarity between two gaits by treating them as two unified objects, a differential composition model (DCM) is constructed. The DCM differentiates the different effects caused by walking speed changes on various human body parts. In the meantime, it also balances well the different discriminabilities of each body part on the overall gait similarity measurements. In this model, the Fisher discriminant ratio is adopted to calculate weights for each body part. Comprehensive experiments based on widely adopted gait databases demonstrate that our proposed method is efficient for cross-speed gait recognition and outperforms other state-of-the-art methods.