Finger Image Research Articles

An Automated Clubbed Fingers Detection System Based on YOLOv8 and U-Net: A Tool for Early Prediction of Lung and Cardiovascular Diseases

Background/Objectives: Lung and cardiovascular diseases are leading causes of mortality worldwide, yet early detection remains challenging due to the subtle symptoms. Digital clubbing, characterized by the bulbous enlargement of the fingertips, serves as an early indicator of these diseases. This study aims to develop an automated system for detecting digital clubbing using deep-learning models for real-time monitoring and early intervention. Methods: The proposed system utilizes the YOLOv8 model for object detection and U-Net for image segmentation, integrated with the ESP32-CAM development board to capture and analyze finger images. The severity of digital clubbing is determined using a custom algorithm based on the Lovibond angle theory, categorizing the condition into normal, mild, moderate, and severe. The system was evaluated using 1768 images and achieved cloud-based and real-time processing capabilities. Results: The system demonstrated high accuracy (98.34%) in real-time detection with precision (98.22%), sensitivity (99.48%), and specificity (98.22%). Cloud-based processing achieved slightly lower but robust results, with an accuracy of 96.38%. The average processing time was 0.15 s per image, showcasing its real-time potential. Conclusions: This automated system provides a scalable and cost-effective solution for the early detection of digital clubbing, enabling timely intervention for lung and cardiovascular diseases. Its high accuracy and real-time capabilities make it suitable for both clinical and home-based health monitoring.

High-speed reflectance confocal microscopy using speckle modulation.

We developed a spectrally-encoded, line reflectance confocal microscope (RCM) that uses a rotating diffuser to rapidly modulate the illumination speckle pattern. The speckle modulation approach reduced speckle noise while imaging with a spatially coherent light source needed for high imaging speed and cellular resolution. The speckle-modulation RCM device achieved lateral and axial resolutions of 1.1 µm and 2.8 µm, respectively. With an imaging speed of 107 frames/sec, three-dimensional RCM imaging over 300-µm depth was completed within less than 1 second. RCM images of human fingers, forearms, and oral mucosa clearly visualized the characteristic cellular features without any noticeable speckle noise.

High-precision hemoglobin detection based on hyperspectral reconstruction of RGB images

Hemoglobin concentration is an important indicator for evaluating human health. The concentration value is an important evidence for early diagnosis of anemia and other blood diseases. However, existing optical detection methods usually suffer from low precision, strong noise and tedious operation, makes it difficult to obtain an accurate hemoglobin concentration evaluation. In this work, we propose and demonstrate a method for hemoglobin detection with the hyperspectral reconstruction of RGB images. To be specific, this overall model consists of a SHR-Net (Skin Hyperspectral Reflectance Network) to reconstruct hyperspectral images from RGB images of fingers, a feature selection process to capture full-band reflectance spectra based on the maximum blood perfusion region and an SVR (Support Vector Regression) algorithm to predict the hemoglobin concentration. The experimental results show that the coefficient of determination between the predicted and true values is 0.945. The mean square error of cross-validation is 0.49 g/dL. Therefore, this method provides a new solution for the accurate and noninvasive measurement of hemoglobin.

Efficient disentangled representation learning for multi-modal finger biometrics

Most multi-modal biometric systems use multiple devices to capture different traits and directly fuse multi-modal data while ignoring correlation information between modalities. In this paper, finger skin and finger vein images are acquired from the same region of the finger and therefore have a higher correlation. To represent data efficiently, we propose a novel Finger Disentangled Representation Learning Framework (FDRL-Net) that is based on a factorization concept, which disentangles each modality into shared and private features, thereby improving complementarity for better fusion and extracting modality-invariant features for heterogeneous recognition. Besides, to capture as much finger texture as possible, we utilize three-view finger images to reconstruct full-view multi-spectral finger traits, which increases the identity information and the robustness to finger posture variation. Finally, a Boat-Trackers-based multi-task distillation method is proposed to migrate the feature representation ability to a lightweight multi-task network. Extensive experiments on six single-view multi-spectral finger datasets and two full-view multi-spectral finger datasets demonstrate the effectiveness of our approach.

Non-invasive detection of haemoglobin, platelets, and total bilirubin using hyperspectral cameras

Hyperspectral imaging has emerged as a promising high-resolution and real-time imaging technology with potential applications in medical diagnostics and surgical guidance. In this study, we developed a high-speed hyperspectral camera by integrating a Fabry-Perot cavity filter on each CMOS pixel. We used it to non-invasively detect three blood components (haemoglobin, platelet, and total bilirubin). Specifically, we acquired transmission images of the subject's fingers, extracted spectral signals at each wavelength, and used dynamic spectroscopy to obtain non-invasive blood absorption spectra. The prediction models were established using the PLSR method and were modelled and validated based on the standard clinical-biochemical test values. The experimental results demonstrated excellent performance. The best predictions were obtained for haemoglobin, with a high related coefficient (R) of 0.85 or more in both the calibration and prediction sets and a mean absolute percentage error (MAPE) of only 5.7%. The results for total bilirubin were also ideal, with R values exceeding 0.8 in both sets and a MAPE of 10.6%. Although the prediction results for platelets were slightly less satisfactory, the error was still less than 15%, indicating that the results were also acceptable. Overall, our study highlights the potential of hyperspectral imaging technology for the development of portable and affordable devices for blood analysis, which can be used in various settings.

Practical Consideration in using Pre-trained Convolutional Neural Network (CNN) for Finger Vein Biometric

Using a pre-trained Convolutional Neural Network (CNN) model for a practical biometric authentication system requires specific procedures for training and performance evaluation. There are two criteria for a practical biometric system studied in this paper. First, the system’s ability to handle identity theft or impersonation attacks. Second, the ability of the system to generate high authentication performance with minimal enrollment period. We propose the use of the Multiple Clip Contrast Limited Adaptive Histogram Equalization (MC-CLAHE) technique to process finger images before being trained by CNN. A pre-trained CNN model called AlexNet is used to extract features as well as classify the MC-CLAHE images. The authentication performance of the pre-trained AlexNet model has increased by a maximum of 30% when using this technique. To ensure that the pre-trained AlexNet model is evaluated based on its ability to prevent impersonation attacks, a procedure to generate the Receiver Operating Characteristics (ROC) curve is proposed. An offline procedure for training the pre-trained AlexNet model is also proposed in this paper. The purpose is to minimize the user enrollment period without compromising the authentication performance. In this paper, this procedure successfully reduces the enrollment time by up to 95% compared to using on-line training.

Simulation, Analysis, and Experimentation of the Compliant Finger as a Part of Hand-Compliant Mechanism Development

Compliant mechanisms are gaining popularity in many different fields, such as in microelectromechanical systems (MEMS), medical applications and health care, opto-mechatronic technology, aerospace engineering, and semiconductor equipment. One of the areas for utilizing compliant mechanisms is building models of human hand counterparts. These models are often used as grasping and rehabilitation devices. Because of their properties, creating a human hand counterpart with compliant mechanisms is a much better choice compared with the models with traditional mechanisms; it looks more realistic, and its movements are much more natural compared with models with a traditional mechanism. A method of modeling and designing such a bio-inspired mechanism, as well as its experimental analysis with various forces applied, is presented in this paper. Two prototypes of the compliant fingers were obtained by 3D printing, and the calculation of the bending angle values was achieved by applying image processing to camera images of the compliant fingers’ prototypes. Image processing was conducted on images taken for both loaded and unloaded 3D-printed compliant finger prototype positions. Finally, these bending angle results are compared with the results obtained by Finite Element Method (FEM) analysis and experimental results acquired by a digital protractor.

Cyfrowy ślad linii papilarnych

Based on the example of activities performed in the Forensic Laboratory of the Police Headquarters in Szczecin in the case of the Central Police Investigation Bureau, the article presents the issue related to the possibility of obtaining investigation material for fingerprint examination in the form of graphic files and video files secured from smartphones with images of fingers and hands. Due to the high quality of a photograph and a single frame from a video, it was possible to positively identify the person, which made it possible to largely substantiate the person’s involvement in the criminal act. The widespread access to smartphones that enable taking and sharing high-quality graphic files, as well as a huge number of photographs and videos shared on the Internet, with increasing frequency, in combination with simple tools for processing video frames and unsophisticated procedures for rescaling traces in AFIS, make a new form of obtaining material for fingerprint identification in the form of a digital fingerprint, and may become more significant.

A Simple and Stable Method of Creating Fingerprint Features with Image Rotation

The main purpose of the classification of fingerprints is to devise a formula by which a given collection of fingerprints can be tracked and registered. To accelerate the system for classifying fingerprints, it is necessary to utilize fingerprint image characteristics and avoid the different fingerprint forms arising from fingerprint rotation. This paper presents a simple, new approach to the extraction of characteristics from fingerprint images. The proposed method demonstrates that, for a given image, the features remain constant even after being subjected to a wide range of rotations; thus, it creates an array of characteristics which can be used to identify a person from their fingerprint. To achieve this goal, a basic hit-and-miss operation with different structural components is used to detect and count various features in the fingerprint picture; these features are directly identified based on the texture of the fingerprint. The chosen features are used to index the finger image by generating a frequency of occurrences for each one, such that every fingerprint is represented as a vector of these features. The application of the proposed method shows efficient utilization of execution time and memory usage.

Hand geometry recognition: an approach for closed and separated fingers

<span lang="EN-US">Hand geometry has been a biometric trait that has attracted attention from several researchers. This stems from the fact that it is less intrusive and could be captured without contact with the acquisition device. Its application ranges from forensic examination to basic authentication use. However, restrictions in hand placement have proven to be one of its challenges. Users are either instructed to keep their fingers separate or closed during capture. Hence, this paper presents an approach to hand geometry using finger measurements that considers both closed and separate fingers. The system starts by cropping out the finger section of the hand and then resizing the cropped fingers. 20 distances were extracted from each finger in both separate and closed finger images. A comparison was made between Manhattan distance and Euclidean distance for features extraction. The support vector machine (SVM) was used for classification. The result showed a better result for Euclidean distance with a false acceptance ratio (FAR) of 0.6 and a false rejection ratio (FRR) of 1.2.</span>

Contactless Fingerprint Recognition Using Deep Learning—A Systematic Review

Contactless fingerprint identification systems have been introduced to address the deficiencies of contact-based fingerprint systems. A number of studies have been reported regarding contactless fingerprint processing, including classical image processing, the machine-learning pipeline, and a number of deep-learning-based algorithms. The deep-learning-based methods were reported to have higher accuracies than their counterparts. This study was thus motivated to present a systematic review of these successes and the reported limitations. Three methods were researched for this review: (i) the finger photo capture method and corresponding image sensors, (ii) the classical preprocessing method to prepare a finger image for a recognition task, and (iii) the deep-learning approach for contactless fingerprint recognition. Eight scientific articles were identified that matched all inclusion and exclusion criteria. Based on inferences from this review, we have discussed how deep learning methods could benefit the field of biometrics and the potential gaps that deep-learning approaches need to address for real-world biometric applications.

Modernization of gas discharge visualization for application in medical diagnostics

The object of research is the processes of the emergence and glow of a discharge around biological structures in a pulsed electric field. Such processes have found use in the method of gas discharge visualization. In medical diagnostics, the general state of human health is assessed by the characteristics of gas-discharge images of fingers. One of the most problematic areas of the correctness of medical diagnostics is the dependence of the visual components of the image on the electrical characteristics of the discharge and the physical and chemical characteristics of the surrounding environment. In the course of the study, methods of modeling the electric discharge current circuit and electrical properties of biostructures were used. The proposed solution allows taking into account: the amplitude of the impulse voltage of the discharge, the frequency of the impulses, the duration and intensity of the impulses, the polarity, which act as additional diagnostic parameters of the gas-discharge visualization process. Physical processes are considered, and a model of a chain of gas discharge around a biological object in a pulsed electric field is proposed. It is shown that the occurrence of a discharge and the characteristics of the glow depend on the amplitude, duration, frequency, and polarity of the pulse voltage. These additional parameters determine the correctness of further visual diagnostics. Their quantitative measurement and the possibility of objective comparison should be attributed to the advantages of registering the proposed parameters of gas discharge visualization. The specified properties of these parameters provide an additional opportunity to digitally describe the condition of the object under study, and subsequently to automate diagnostics. The structural diagrams of the device for conducting research using the method of gas discharge visualization, the high-voltage impulse voltage generator unit for the hardware consideration of additional gas discharge parameters and their connection with medical and biological indicators have been developed. The use of the method and means of gas discharge visualization to assess the functional state of the flight crew in the pre- and post-flight period requires the development of special equipment. The proposed technical solutions require experimental verification. Comparative studies of diagnostic conclusions by the method of gas-discharge visualization with traditional medical diagnostics are necessary.

Optical Scanning System for Imaging of Heterogeneity in Biological Tissues

A non-invasive hand-held optical probe for imaging of biological tissues consists of nine units. Each unit is equipped with one red LED and three photodetectors placed at distances 7, 12, and 17 mm from the source along the x-axis. For testing, three phantoms of paraffin wax are embedded with solid inhomogeneity placed at various depths. Human biological tissues consisting of right index finger, four fingers of the same hand, and a part of the forearm below elbow joint are used. The data on optical backscattering are collected by moving the probe on the surface of phantom/tissue and after processing the respective images are obtained. Three images of the index finger show the tissue structural variation in various layers, which are in agreement with anatomical details and radiograph of the finger. The tissue details of four fingers are in agreement with that of individual finger. Similarly, the three images of the forearm show the structural variation. These observations suggest that the present technique may be used to detect tissue compositional changes below the skin.

HOS-FingerCode: Bispectral invariants based contactless multi-finger recognition system using ridge orientation and feature fusion

Fingerprints are a commonly used biometric for civil and forensic applications and the majority of them are captured by contact sensors. More recently, there is a trend toward the use of high resolution digital and video cameras to acquire contactless images. However, the processing of such images is challenging because of intra-class variations such as pose, rotation, translation and scaling due to hand orientation and deformation. They can, however, facilitate the use of full finger or multiple fingers or hand information rather than the fingertip alone. Although some recent research has investigated representation of full finger and hand information under relaxed conditions, current state-of-the-art methods are unable to address transformation issues without manual intervention. To overcome the issues of relaxed constrained nature, geometrical variations and invariant feature encoding regardless of finger orientation, we have used higher order spectral features (HOS-FingerCode) from ridge patterns. We employ an automated method to detect key points and constructs a graph of key-lines. Ridge orientation profiles along with the selected key-lines provide a 1D signal from which bispectral invariant features are extracted. In addition, two fusion schemes are proposed where fusion from multiple key lines combinations and feature fusion from multiple key line configurations are investigated to extract ridge pattern profile for a better performance. The features are pooled from multiple fingers by concatenation or compact bilinear pooling to yield the HOS-FingerCode for multi-finger biometrics. For a single-finger system using 3369 high resolution index and middle finger images, the algorithm better performs when different key lines configurations are fused. For geometrical variations of the images, the system is most impacted by pose than rotation and scale changes of the images, and is tolerant to such variations. For the multi-finger system using 3111 images from both fingers, the algorithm has achieved 93.8% TPR and 97.95% classification accuracy at a setting of 2% FPR. For the feature fusion scheme proposed for multi-finger system using concatenation and compact bilinear pooling, concatenation of features performs better than the compact bilinear pooling. Since multi-finger recognition system performed better than the single finger biometrics system, it can help to diminish the high traffic scenario in busy premises and will facilitate the soft identity verification.

Understanding and modeling finger vascular pattern imaging

In this paper, new insights in the near infrared imaging process used in finger-vein recognition by developing a physical model are presented. A realistic phantom finger that mimics the living human finger and also includes veins has been developed to validate this model. NIR phantom finger images show that the phantom can emulate the optical properties of a living human finger and can provide ground truth for the locations of the veins. Through physical modeling, it is particularly learned that—besides blood and soft tissue—bone also plays an important role in generating reliable NIR finger-vein images.

Continuous spectral zooming for in vivo live 4D-OCT with MHz A-scan rates and long coherence.

We present continuous three-dimensional spectral zooming in live 4D-OCT using a home-built FDML based OCT system with 3.28 MHz A-scan rate. Improved coherence characteristics of the FDML laser allow for imaging ranges up to 10 cm. For the axial spectral zoom feature, we switch between high resolution and long imaging range by adjusting the sweep range of our laser. We present a new imaging setup allowing for synchronized adjustments of the imaging range and lateral field of view during live OCT imaging. For this, a novel inline recalibration algorithm was implemented that enables numerical k-linearization of the raw OCT fringes for every frame instead of every volume. This is realized by acquiring recalibration data within the dead time of the raster scan at the turning points of the fast axis scanner. We demonstrate in vivo OCT images of fingers and hands at different resolution modes and show real three-dimensional zooming during live 4D-OCT. A three-dimensional spectral zooming feature for live 4D-OCT is expected to be a useful tool for a wide range of biomedical, scientific and research applications, especially in OCT guided surgery.

Vasodilatation Measurement Using Finger Vascular Images by Near-Infrared Light and Comparison with RH-PAT

Examination of vascular endothelial function can help infer atherosclerosis progression. This study investigated whether vascular visualization by near-infrared (NIR) light can detect vasodilatation after cuff pressure release of the upper arm and what the correlation is between the brightness decrease ratio (R1) corresponding to vasodilation and the reactive hyperemia index (RHI). We obtained finger vascular images of 53 male subjects by photographing NIR light (wavelength 850 nm) transmitted through the middle phalanx of the middle finger with a charge-coupled device camera. The upper arm was compressed for 10 min using a cuff (200 mmHg), and vascular images before and after cuff compression release were obtained. We analyzed the finger vascular images by NIR light and digital pulse volume using endothelial peripheral arterial tonometry (Endo-PAT). We also calculated the average brightness of each vascular image. Using only the data of the ischemic finger, R1 was defined using the average brightness just before cuff release and the minimum average brightness after cuff release. The brightness of vascular images of the ischemic finger decreased after cuff release probably because of vasodilation. We found significant correlation between R1 and the RHI (r = 0.52; P < 0.001). R1 in the lowest RHI quartile was significantly smaller compared to the highest and second-highest RHI quartiles (P < 0.05). Vascular visualization by NIR light can detect vasodilation after cuff release. This is significantly correlated with the RHI on Endo-PAT.

Prediction of Rheumatoid Arthritis disease using improved segmentation algorithm

Image processing and segmentation algorithms have improved to a great extent in diagnosing numerous complications by analyzing Electronic Medical Records (EMR) like X-Rays, CT scan and MRI Images. However the accuracy of predicting the outcome is still a major challenge, which mainly depends on the quality of algorithms used, the testing dataset, and the segmentation process. Therefore, it is essential to identify novel algorithms for image processing to improve the accuracy of prediction. One of the important medical complication that is considered for our research is Rheumatoid Arthritis (RA). It is a persistent immune disorder that causes reduced bone density, cartilage damage and inflammations in joints. Traditional methods, though cost effective like blood test, X- Ray analysis and cell diagnosis, they have their own error factors. On the other hand techniques like advanced thermal imaging are accurate but with extremely high cost. Therefore, the proposed idea is to comparatively assess the cost, accuracy and early detection of RA based on scan images of hands and fingers. A Novel Gaussian filtering and segmentation algorithms is proposed for removing unwanted disturbances and extracting the clear images of the joints, that segments only the affected part in the image to provide a better accuracy. Gamma correction luminosity is used for the sample images which is then processed for identifying the shadows of the image. Major implication of the proposed novel algorithms helps to detect the present stages of RA for affected persons and also predict the possibility of RA at early stages.

Performance analysis of local binary pattern and k-nearest neighbor on image classification of fingers leaves

The K-Nearest Neighbor (KNN) method is often used by researchers for the classification process because it has a relatively great level of accuracy, however it also has a weakness which is sensitive of the noises. This research is aims to introduce an object recognition (identification) system of fingers leaves by classified using the KNN method. To resolves the weaknesses of the KNN method, the researcher has used the Local Binary Pattern (LBP) method to extract features of the leaves. For the comparison in feature extraction, the researcher has used the Gray Level Co-Occurrence Matrix (GLCM) method. The data that were used on this research are papaya leaves and chaya leaves (with the labels such as good and damage forms). In this research, an experimental design has been carried out that was differentiated by according to the comparison (of ratio) between training data and testing data (NI/Np), there were 90 training data and 45 testing data, where the feature extraction method used the 10 of features. Experimentally, it was shown that by using the ratio NI/Np = 67%:33%, the performance or system performance for classifying the images of fingers leaves by using the LBP extraction method showed that training data was obtained the results close to 95% and testing data was obtained the results close to 76%, while by using the GLCM extraction showed that training data was obtained the results close to 83% and testing data was obtained the results close to 58%.

An Application of Machine Learning Algorithms on the Finger Image Prediction

In recent years, using machine learning (ML) algorithm to analyze a picture, obtain its features, and finally identify what the picture is about is getting more and more important. This is because with the popularity of the electronic equipment and high-performance computing equipment, people began to pursue a more convenient and automatic life. The science of the image recognition frees people’s hands to a certain extent through the training of algorithms, thus making people’s life more convenient. This paper presents a comparison of two ML algorithms: Multi-layer Perceptron (MLP), and Convolutional Neural Network (CNN) with three different optimization methods on the data-set by measuring their test accuracy and their running time. The said data-set consists of a training-set of 1080 pictures (64 by 64 pixels) of signs representing numbers from 0 to 5 (180 pictures per number) and a test set of 120 pictures (64 by 64 pixels) of signs representing numbers from 0 to 5 (20 pictures per number). For the implementation of the ML algorithms, the data-set was partitioned in the following fashion: 90% for training phase, and 10% for testing phase. The hyper-parameters used for all the classifiers were manually assigned. Results show that most of the presented ML algorithms performed not bad with a test accuracy over 80%, and the CNN algorithm performed best among all the implemented algorithms with a test accuracy about 91.04%.