Convexity Defects Research Articles

Abstract 230: Deep learning embedding-based segmentation for morphological analysis in organoids

Abstract Background: The organoid model is a useful tool for modeling the cellular microenvironment of the organ from which they are derived. Organoids recapitulate the self-organization of heterogenous cell types and the microenvironment. Quantifying the morphological features of organoids can provide valuable insights into cellular organizational defects and growth characteristics, which can facilitate drug discovery. Measurement of these characteristics in live organoids can be performed quickly and easily directly in culture using widefield microscopy. The current state-of-the-art method to detect and identify the shape of individual organoids in brightfield images uses the U-Net Convolutional Neural Network (CNN) with a watershed transform to label pixels. However, this method yields jagged shape proposals and cannot detect overlapping regions. We propose the use of an embedding-based segmentation network based on the Branched ERF-Net to solve these issues. Methods: Organoids were established from the proximal colon of BrafV600E heterozygous mice, and 50 phase-contrast images of the organoids were acquired at 2.5x and 4x magnification. Organoids present in acquired images were labeled manually. The network was trained on this collection of manually labeled images and validated on a separate validation dataset. Results: The network accurately labels organoids. Overlapping organoids are segmented correctly and shape proposals are smoother. Detection of overlapping organoids with the Branched ERF-Net architecture yields an accurate organoid count, verified against manual counting. Smoother shape proposals also enable the use of convexity defects to measure organoid budding. Conclusions: Deep-learning analysis of widefield images enables the rapid assessment of morphological characteristics, such as size, count, and budding, which are key to understanding proliferation and differentiation changes. The modified Branched ERF-Net architecture we propose for organoid segmentation is a robust and versatile method to quantify organoid morphology. Citation Format: Akash Sureshkumar, Shilpa Bisht, Hariharan Easwaran. Deep learning embedding-based segmentation for morphological analysis in organoids [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 230.

Karyotyping of human chromosomes in metaphase images using faster R‐CNN and inception models

AbstractKaryotyping is the process of pairing and ordering human chromosomes from metaphase chromosomal images depending on their size, centromere position, and banding patterns. It is used to analyze human chromosomes for various genetic disorders especially during prenatal screenings. Since manual karyotyping is a labor‐intensive and a time‐consuming task, developing an automatic or semi‐automatic computer‐assisted karyotyping system is the need of the hour. The proposed karyotyping system aims to detect human chromosomes (22 pairs of autosomes and one pair of sex chromosome) from the microscopic metaphase images which is followed by separation of overlapped chromosomes. Extracted chromosomes are first classified into seven Denver groups (A to G) followed by classifying individual chromosomes within their Denver group. Then the classified chromosome pairs are used to create a karyotype with the assistance of cytologists or by using estimated chromosome length as cytogenetic parameter. A total of 234 chromosomal images from two different public datasets are used for experimentation. Human chromosomes in the microscopic metaphase images are detected using faster regions with convolutional neural networks (Faster R‐CNN) combined with Inception v2. Then convexity defect algorithm is preferred for separation of overlapped chromosomes. The detected chromosomes are classified using the proposed two‐step approach in which Inception v3 model is used. Then the classified chromosome pairs are used to create a karyotype with the assistance of cytologists or by using estimated chromosome length as cytogenetic parameter. Faster R‐CNN model gives a detection accuracy of 98.53%. Denver group classification with the two‐step approach provides a better accuracy of 84.59% when network is trained for 64 902 epochs. Faster R‐CNN outperforms in detection and works faster as it searches chromosomes within the regional proposals. The two‐step classification approach gives better classification accuracy. The proposed approach also works very well for overlapping chromosomes due to use of convexity defects algorithm. The classified chromosome pairs are useful tools for the cytologists to create a karyotype with minimal efforts.

Comparative analysis on real-time hand gesture and sign language recognition using convexity defects and YOLOv3

Comparative analysis on real-time hand gesture and sign language recognition using convexity defects and YOLOv3

Sign Language Recognition Using Convexity Defects and Hough Line

Sign Language Recognition Using Convexity Defects and Hough Line

Employment of baby sign language in interactive learning for Hearing-impaired people in Malaysia

<span lang="EN-GB">Communication is very important for people including adults and babies to exchange their thoughts, views and information. They can express their thoughts verbally and non-verbally. Baby sign language is non-verbal type that consists of a set of standard and organised form of hand gestures motion to represent different word. This language allows the children who have hearing impairment problem to communicate with others. In addition, it can also be taught to normal babies to convey their need before they have the ability to speak. It is very challenging for adults to understand babies and children with hearing impaired problem. Thus, a real-time hand gesture recognition system has been developed for adult to interactively learn the baby sign language. Firstly, the face region was detected in order to use the skin colour model of the face to detect the hand. After that, the hand shape was determined by finding the palm centre and convex hull. Convexity defects which representing gaps between the fingers were used to determine the number of fingertips whereas the movement of the hand was detected based on motion analysis. By setting the background colour to blue and adding extra light in front of the laptop, the system successfully obtained an accuracy higher than 85% rate for sign “mom”, “eat” and “milk” when tested on Malay, Chinese and Indian races in Malaysia.</span>

Nondestructive Visual Inspection Method of Double-Yolked Duck Egg

Duck eggs are rich in good protein and deeply favored by consumers. Among them, double-yolked duck eggs have high commercial value but are not easy to incubate, so they are often picked out by inspection before processing, circulation or being incubated in egg products industry. This paper used the computer vision technology on single- and double-yolked duck eggs with similar shape and size that were hard to distinguish merely by their external characteristics, studied the effective image processing algorithm and built a cognition model. Then, we collected duck eggs’ transmission light color image, extracted B channel image, carried out iterative morphological open reconstruction and threshold segmentation to acquire yolk area, used convexity defects to identify double-yolked eggs and finally utilized watershed and ellipse fitting to separate yolk. Through inspection, 150 single- and double-yolked eggs were distinguished by using this method, with correct recognition rate reaching to 99% and 100%, respectively. The experimental results indicate that this method has a high accuracy in recognizing double-yolked duck eggs and can provide technical support in nondestructive identification of double-yolked duck egg.

Analisis Penerapan Metode Convex Hull Dan Convexity Defects Untuk Pengenalan Isyarat Tangan

Research of Human Computer Interaction (HCI) and Computer Vision (CV) is increasingly focused on advanced interface for interacting with humans and creating system models for various purposes. Especially for input device problem to interact with computer. Humans are accustomed to communicate with fellow human beings using voice communication and accompanied by body pose and hand gesture. The main purpose of this research is to applying the Convex Hull and Convexity Defects methods for Hand Gesture Recognition system.
 In this research, the Hand Gesture Recognition system designed with the OpenCV library and then receives input from the user's hand gesture using an integrated webcam on the computer and system generates a language output from the hand-recognizable gestures.
 Testing involves several variables which affect success in recognizing user's hand gestures, such as hand distance with webcam, corner of the finger, light conditions and background conditions. As a result, the user's hand gestures can be recognized with a stable and accurate when at a distance of 50cm-70cm, corner of the finger 25o–70o, light conditions 150lux-460lux and plain background conditions.

Icing Condition Assessment of In-Service Glass Insulators Based on Graphical Shed Spacing and Graphical Shed Overhang

Icing on transmission lines might lead to ice flashovers of insulators, collapse of towers, tripping faults of transmission lines, and other accidents. Shed spacing and shed overhang of insulators are clues for evaluating the probability of ice flashover. This paper researches image-processing methods for the natural icing of in-service glass insulators. Calculation methods of graphical shed spacing and graphical shed overhang are proposed via recognizing the convexity defects of the contours of an icing insulator string based on the GrabCut segmentation algorithm. The experiments are carried out with image data from our climatic chamber and the China Southern Power Grid Disaster (Icing) Warning System of Transmission Lines. The results show that the graphical shed overhang of insulators show evident change due to icing. This method can recognize the most serious icing conditions where the insulator sheds are completely bridged. Also, it can detect bridging positions including the left side, right side, or both sides of the insulator strings in the images.

Recent Trends and Technologies in Hand Gesture Recognition

Hand gesture recognition is a growing field of research. Gestures are an important aspect of human interaction, both interpersonal and in the context of human computer interface. The interaction between computers and humans are provided by the hand gestures. Gestures provide a way for computers to understand human body language. Hand gesture recognition deals with the goal of interpreting hand gestures via mathematical algorithms. The techniques discussed are based on detection of some shape based features like centroid, orientation of hand, presence of finger, thumb in image and contour based features like convex hull and convexity defects. Hand gesture recognition system provides a user friendly way of interaction with the computer which is more familiar to the human beings. Hand gesture recognition process includes capturing input image which will be processed and segmented so as to extract features to identify different gestures. Different methods of hand gesture recognition techniques are evaluated and proposed a novel method for recognition. This method explains how the users hand is detected and the corresponding gesture is identified in an accurate manner. This paper also explains the various applications that can take advantage of this type of interaction mechanism.

Real Time Vision Based Method for Finger Counting Through Shape Analysis with Convex Hull and PCA Techniques

This paper presents a real time vision-based finger counting method combining convex-hull detection and PCA techniques. The method starts by segmenting an input image to detect the area corresponding to the observed hand. For that purpose, a skin color detection method is used to differentiate the foreground containing the hand from the image background. Lighting variation can affect the accuracy of the segmentation which has an impact on the functioning of the proposed method. In order to deal with the lighting variation problem, HLS color space is used to represent the colors. Hand contour is then calculated and fingertips are detected through the detection of the convex-hull and convexity defects of the hand shape. The use of convex-hull algorithm is simple and gives accurate results when more than one finger is observed in the input image but the accuracy decreases when it comes to deal with the case where only one finger is observed. To overcome this problem, principal component analysis technique (PCA) is used to analyze the hand shape and detect the case where only one finger is observed with better accuracy. The proposed method could be utilized in Human Computer Interaction System (HCI) where the machine reacts to each detected number. Both real and synthetic images are used to test and demonstrate the potential of our method.

Hand gesture recognition using average background and logical heuristic equations

This paper introduces a hand gesture recognition algorithm for Human Computer Interaction using real-time video streaming .The background subtraction technique is used to extract the ROI (Region Of Interest) of the hand. Fingertip is detected using logical heuristics equations that are applied on hand contour , convex hull and convexity defects points. A combination between background subtraction and logical heuristic technique that leads to more accurate results is introduced. Experimental results prove that the proposed algorithm improve the finger's tips detection by 52 % compared to the reference model.

Gravity optimised particle filter for hand tracking

This paper presents a gravity optimised particle filter (GOPF) where the magnitude of the gravitational force for every particle is proportional to its weight. GOPF attracts nearby particles and replicates new particles as if moving the particles towards the peak of the likelihood distribution, improving the sampling efficiency. GOPF is incorporated into a technique for hand features tracking. A fast approach to hand features detection and labelling using convexity defects is also presented. Experimental results show that GOPF outperforms the standard particle filter and its variants, as well as state-of-the-art CamShift guided particle filter using a significantly reduced number of particles.

Optimizing convexity defect in a tile industry using fuzzy goal programming

In most designed experiments, the main focus is to find the factor settings that optimize a quality response regardless of engineer’s preferences about factor settings. Further, in tiles industry convexity defects result in huge quality costs as well as production losses. This research, therefore, aims at optimizing convexity defect while considering process engineers’ preferences using fuzzy goal programming (FGP). Three two-level key process factors are considered, including below-rollers temperature, above-rollers temperature, direct blow air. Experiments are conducted with two repetitions; in each the convexity is measured on four tiles. Two optimization techniques are employed to determine the combination of optimal factor settings, including the Taguchi method and latter technique. The Taguchi approach and FGP approach provide relative improvements of 61.2% and 41.2%, respectively. Although the former technique reduces convexity larger than latter approach, it failed to satisfy the preferences on the settings of process factors. In contrast, the optimal factor settings obtained using FGP completely satisfy engineers’ preferences. In conclusion, FGP successfully optimizes process performance and completely satisfies process engineers’ preferences in tiles industry.

Human action recognition using hull convexity defect features with multi-modality setups

We consider developing a taxonomic shape driven algorithm to solve the problem of human action recognition and develop a new feature extraction technique using hull convexity defects. To test and validate this approach, we use silhouettes of subjects performing ten actions from a commonly used video database by action recognition researchers. A morphological algorithm is used to filter noise from the silhouette. A convex hull is then created around the silhouette frame, from which convex defects will be used as the features for analysis. A complete feature consists of thirty individual values which represent the five largest convex hull defects areas. A consecutive sequence of these features form a complete action. Action frame sequences are preprocessed to separate the data into two sets based on perspective planes and bilateral symmetry. Features are then normalized to create a final set of action sequences. We then formulate and investigate three methods to classify ten actions from the database. Testing and training of the nine test subjects is performed using a leave one out methodology. Classification utilizes both PCA and minimally encoded neural networks. Performance evaluation results show that the Hull Convexity Defect Algorithm provides comparable results with less computational complexity. This research can lead to a real time performance application that can be incorporated to include distinguishing more complex actions and multiple person interaction.

Managing the firing process of ceramics tiles to reduce convexity defect: design of experiment approach

Ceramics industry is an integral part of any country’s economic growth. Because of its competitive nature and wide range availability of ceramics, for both industrial and domestic consumers, it is important to have consistency in the quality of ceramics tiles. There are many defects identified in ceramics such as flatness, cracks, orthogonality, wrinkles etc. In this paper, many factors were identified during the firing process of production, which results in the convexity defect in tiles. A full factorial experiment that requires a total of 23 experimental runs (treatment combinations) for a complete replicate was carried out to determine the factors that affect the convexity occurrence. Minitab software was used to carry out regression analysis, analysis of variance, and residual analysis. The results show that the main factors below roll and above roll temperatures along with their respective interactions with the direct air blowing significantly affect the convexity of the tiles. Model implementation shows 41% reduction in convexity defects.

An Interactive System for Sensory and Gustatory Impaired People Based On Hand Gesture Recognition

Sign Language is an interactive system for sensory and gustatory impaired people for communicating information among them. Hand Gesture Recognition (HGR) is a system for sign language detection which uses computers for enabling communication with sensory impaired people. In this research work, an efficient and effective technique is introduced for the identification of the number of fingers opened or closed in a gesture representing an alphabet of the American Sign Language (ASL). Finger Detection is accomplished with the use of Contour Hull and Convexity Defects. Contour Hull helps to form the boundary of the hand gestures whereas the Convexity Defects helps to find tip of the fingers captured by the web camera. This proposed system applies image processing techniques to store the pre-processed data, Machine learning technique is used in supervised clustering and computer vision techniques is used to recognize the hand gestures. Hence this intelligent system does not require the hand to be perfectly aligned to the camera for detecting the signs. In addition to this, to detect the signs, this system does not need any colored markers or gloves to be wearied on the hands.

Application of machine vision for automated cell injection

This paper presents image processing algorithms recognising cell structures for an automated cell injection system. For suspended zebrafish embryos, the convex hull and convexity defects of the cytoplasm contour are used. For adherent endothelial cells, the surface and shadow information of the nucleoli is used. Overall, 550 zebrafish embryos were 100% correctly recognised in online injection experiments. Compared to manual injection, the automated injection speed and reproducibility increase by up to 1.5-fold and 67% respectively. In the endothelial cell case, 436 nucleoli were 92% correctly recognised, paving the way for an automated adherent cell injection system to be developed.