Showing papers on "Standard test image published in 2022"

Hand Detection and Segmentation in First Person Image Using Mask R-CNN

Abstract: In this work, we propose a technique to automatically detect and segment hands on first-person images of patientsin upper limb rehabilitation exercises. The aim is to automate the assessment of the patient's recovery processthrough rehabilitation exercises. The proposed technique includes the following steps: 1) setting up a wearablecamera system and collecting upper extremity rehabilitation exercise data. The data is filtered, selected andannotated with the left and right hand as well as segmented the image area of the patient's hand. The datasetconsists of 3700 images with the name RehabHand. This dataset is used to train hand detection and segmentationmodels on first-person images. 2) conducted a survey of automatic hand detection and segmentation models usingMask-RCNN network architecture with different backbones. From the experimental architectures, the Mask -RCNN architecture with the Res2Net backbone was selected for all three tasks: hand detection; left - right handidentification; and hand segmentation. The proposed model has achieved the highest performance in the tests. Toovercome the limitation on the amount of training data, we propose to use the transfer learning method alongwith data enhancement techniques to improve the accuracy of the model. The results of the detection of objects onthe test dataset for the left hand is AP = 92.3%, the right hand AP = 91.1%. The segmentation result on the test dataset forleft hand is AP = 88.8%, right hand being AP = 87%. These results suggest that it is possible to automatically quantifythe patient's ability to use their hands during upper extremity rehabilitation.

DLTTA: Dynamic Learning Rate for Test-Time Adaptation on Cross-Domain Medical Images

Abstract: Test-time adaptation (TTA) has increasingly been an important topic to efficiently tackle the cross-domain distribution shift at test time for medical images from different institutions. Previous TTA methods have a common limitation of using a fixed learning rate for all the test samples. Such a practice would be sub-optimal for TTA, because test data may arrive sequentially therefore the scale of distribution shift would change frequently. To address this problem, we propose a novel dynamic learning rate adjustment method for test-time adaptation, called DLTTA, which dynamically modulates the amount of weights update for each test image to account for the differences in their distribution shift. Specifically, our DLTTA is equipped with a memory bank based estimation scheme to effectively measure the discrepancy of a given test sample. Based on this estimated discrepancy, a dynamic learning rate adjustment strategy is then developed to achieve a suitable degree of adaptation for each test sample. The effectiveness and general applicability of our DLTTA is extensively demonstrated on three tasks including retinal optical coherence tomography (OCT) segmentation, histopathological image classification, and prostate 3D MRI segmentation. Our method achieves effective and fast test-time adaptation with consistent performance improvement over current state-of-the-art test-time adaptation methods. Code is available at https://github.com/med-air/DLTTA .

72‐2: A Proposal for Image‐Compression Algorithm for Display Test Images

Abstract: A generative image compression algorithm is applied to display test images to reduce the need for storage spaces. The preliminary result showed better preservation of image details at roughly equal compression rate.

Joint approach based quality assessment scheme for compressed and distorted images

Abstract: The Digital era is improving day by day. We can easily send and receive multimedia data, but it is challenging to know that the data is of actual quality or degraded and compressed. A similar problem arises in the medical imaging domain; it is too tedious to determine whether the image has a particular quality level or not to modify it further. So here we represent one specific method that is termed as “Joint”- Image Quality Estimation Approach as it is a combination of reference-based and no-reference-based Image Quality assessment methods; due to this fact, we termed it “Joint” approach. In some cases, the reference-based image quality assessment methods cannot predict the exact values because we don't know that the reference image that is considered to find the quality of a test image is an actual one or previously compressed. So, this will create a situation where we get the wrong IQA value for the test image. The method proposed by us can overcome this problem. First, we decide the quality of the reference image by using No-reference-based models. Then, we check the final IQA value for a test image with the reference-based models. We created a database of 72 chest images of COVID-19 infected patients and its four-level compressed images for the experiment. Results that are shown in this work are very effective and elaborated with proper justifications.

An image-based concrete crack detection method using convolutional neural networks

Abstract: This paper proposes a CNN-based crack detection method that can recognize and extract cracks from photos of concrete structures. The algorithm consists of two subsequent procedures, classification, and segmentation, achieved by two convolutional neural networks respectively. First, full images are divided into patches and classified as positive and negative. Then, those sub-images classified as positive are further processed by the image segmentation procedure to obtain the pixel level geometry of the cracks. For the classification part, the performance of transfer learning models based on pre-trained VGG16, Inception V3, MobileNet and DenseNet169 is compared with different classifier. Finally, the CNN based on MobileNet was trained with 30,000 training images and reached 97% testing accuracy and 0.96 F1 score on testing image. For the segmentation part, different neural networks based on the elegant U-net architecture are built and tested. The models are trained with 3840 crack images and annotated ground truth and compared quantitatively and qualitatively. The model with the best performance reached 88% sensitivity on test data set. The combination of the classification and segmentation neural networks achieves an image-based crack detection method with high efficiency and accuracy. The algorithm can process any full image size as input. Compared with most machine learning based crack detection algorithms using sub-image classification, a relatively larger patch size is used in this paper and in this way the classification is more robust and accurate. On the other hand, the negative areas in the full image will not be concerned in the segmentation procedure and this fact not only saves a lot of computational power but also significantly increases the accuracy compared to the segmentation performed on full images.

Test-time image-to-image translation ensembling improves out-of-distribution generalization in histopathology

Abstract: Histopathology whole slide images (WSIs) can reveal significant inter-hospital variability such as illumination, color or optical artifacts. These variations, caused by the use of different scanning protocols across medical centers (staining, scanner), can strongly harm algorithms generalization on unseen protocols. This motivates development of new methods to limit such drop of performances. In this paper, to enhance robustness on unseen target protocols, we propose a new test-time data augmentation based on multi domain image-to-image translation. It allows to project images from unseen protocol into each source domain before classifying them and ensembling the predictions. This test-time augmentation method results in a significant boost of performances for domain generalization. To demonstrate its effectiveness, our method has been evaluated on 2 different histopathology tasks where it outperforms conventional domain generalization, standard H&E specific color augmentation/normalization and standard test-time augmentation techniques. Our code is publicly available at https://gitlab.com/vitadx/articles/test-time-i2i-translation-ensembling.

The Application of The Manhattan Method to Human Face Recognition

Abstract: In face recognition, the input image used will be converted into a simple image, which will then be analyzed. The analysis was carried out by calculating the distance of data similarity. In the process of measuring data similarity distances, they often experience problems implementing complex algorithm formulas. This research will solve this problem by implementing the Manhattan method as a method of measuring data similarity distances. In this study, it is hoped that the Manhattan method can be used properly in the process of matching test images and training images by calculating the proximity distance between the two variables. The distance sought is the shortest distance; the smaller the distance obtained, the higher the level of data compatibility. The image used in this study was converted into grayscale to facilitate the facial recognition process by thresholding, namely the process of converting a grayscale image into a binary image. The binary image of the test data is compared with the binary image of the training data. The image used in this study is in the Joint Photographic Experts Group (JPEG) format. Testing was carried out with 20 respondents, with each having two training images and two test images. The research was conducted by conducting experiments as many as 20 times. Facial recognition research using the Manhattan method obtains an accuracy of 70%. The image lighting used as the dataset influenced the accuracy results obtained in this study. Based on the results of this study, it can be concluded that the Manhattan method is not good for use in facial recognition research with poor lighting.

DETECTION AND LOCALIZATION OF RETINAL BREAKS IN ULTRAWIDEFIELD FUNDUS PHOTOGRAPHY USING a YOLO v3 ARCHITECTURE-BASED DEEP LEARNING MODEL

Abstract: We aimed to develop a deep learning model for detecting and localizing retinal breaks in ultrawidefield fundus (UWF) images.We retrospectively enrolled treatment-naive patients diagnosed with retinal break or rhegmatogenous retinal detachment and who had UWF images. The YOLO v3 architecture backbone was used to develop the model, using transfer learning. The performance of the model was evaluated using per-image classification and per-object detection.Overall, 4,505 UWF images from 940 patients were used in the current study. Among them, 306 UWF images from 84 patients were included in the test set. In per-object detection, the average precision for the object detection model considering every retinal break was 0.840. With the best threshold, the overall precision, recall, and F1 score were 0.6800, 0.9189, and 0.7816, respectively. In the per-image classification, the model showed an area under the receiver operating characteristic curve of 0.957 within the test set. The overall accuracy, sensitivity, and specificity in the test data set were 0.9085, 0.8966, and 0.9158, respectively.The UWF image-based deep learning model evaluated in the current study performed well in diagnosing and locating retinal breaks.