Search

Mitosis is a critical criterion for meningioma grading. However, pathologists assessment of mitoses is subject to significant inter-observer variation due to challenges in locating mitosis hotspots and accurately detecting mitotic figures. To address this issue, we leverage digital pathology and propose a computational strategy to enhance pathologists mitosis assessment. The strategy has two components: (1) A depth-first search algorithm that quantifies the mathematically maximum mitotic count in 10 consecutive high-power fields, which can enhance the preciseness, especially in cases with borderline mitotic count. (2) Implementing a collaborative sphere to group a set of pathologists to detect mitoses under each high-power field, which can mitigate subjective random errors in mitosis detection originating from individual detection errors. By depth-first search algorithm (1) , we analyzed 19 meningioma slides and discovered that the proposed algorithm upgraded two borderline cases verified at consensus conferences. This improvement is attributed to the algorithms ability to quantify the mitotic count more comprehensively compared to other conventional methods of counting mitoses. In implementing a collaborative sphere (2) , we evaluated the correctness of mitosis detection from grouped pathologists and/or pathology residents, where each member of the group annotated a set of 48 high-power field images for mitotic figures independently. We report that groups with sizes of three can achieve an average precision of 0.897 and sensitivity of 0.699 in mitosis detection, which is higher than an average pathologist in this study (precision: 0.750, sensitivity: 0.667). The proposed computational strategy can be integrated with artificial intelligence workflow, which envisions the future of achieving a rapid and robust mitosis assessment by interactive assisting algorithms that can ultimately benefit patient management.

In abstract argumentation frameworks, the computation of stable extensions is an important semantic task for evaluating the acceptability of arguments. The current approaches for the computation of stable extensions are typically conducted through methodologies that are either label-based or extension-based. Label-based algorithms operate by assigning labels to each argument, thus reducing the attack relations between arguments to constraint relations among the labels. This paper analyzes the existing two-label and three-label enumeration algorithms for stable extensions through case studies. It is found that both the two-label and three-label algorithms are not precise enough in defining types of arguments. To address these issues, this paper proposes a four-label enumeration algorithm for stable extensions. This method introduces a m u s t _ i n label to pre-mark certain i n -type arguments, thereby achieving a finer classification of i n -type arguments. This enhances the labelings' propagation ability and reduces the algorithm's search space. Our proposed four-label algorithm was tested on authoritative benchmark sets of abstract argumentation framework problems: ICCMA 2019, ICCMA 2021, and ICCMA 2023. Experimental results show that the four-label algorithm significantly improves solving efficiency compared to existing two-label and three-label algorithms. Additionally, ablation experiments confirm that both the four-label transition strategy and preprocessing strategy enhance the algorithm's performance. [ABSTRACT FROM AUTHOR]

The popularization of intelligent toys enriches the lives of the general public. To provide the public with a better toy experience, we propose the intelligent toy tracking method by the mobile cloud terminal deployment and depth-first search algorithm. Firstly, we construct a toy detection model via Transformer, which realizes the positioning of toys in the image through the refined region adaptive boundary representation. Then, using these detected continuous frames, we improve the toy tracking based on a depth-first search. Long-short-term memory (LSTM) constructs the continuous frame tracking structure, and the depth-first search mechanism is embedded to realize the accurate tracking of multiple targets in continuous frames. Finally, to realize the terminal marginalization of the proposed method, this chapter proposes a lightweight model deployment method based on mobile cloud terminals to realize the maintenance of the optimal machine state of intelligent toys. The experiment proves that our proposed target method can reach the world-leading level and obtain the mAP value of 0.858. Our tracking method can also perform excellently with a MOTA value of 0.916. [ABSTRACT FROM AUTHOR]

The IOTA Tangle, a Directed Acyclic Graph (DAG)-based distributed ledger, is popular for its scalability and suitability for IoT applications, offering fee-less transactions. A critical component of IOTA’s architecture is the Cumulative Weight Calculation (CWC), essential for its tip selection mechanism. This paper introduces an optimization of the IOTA Reference Implementation (IRI) CWC process originally implemented using Breadth-First Search (BFS) by employing Depth-First Search (DFS) and Iterative Deepening Search (IDS) algorithms. We present a comparative analysis of these methods, demonstrating that DFS and IDS provide significant improvements in computational efficiency, particularly beneficial for IoT devices with limited processing capabilities. Our findings are substantiated through a series of experiments on a Tangle snapshot, highlighting the enhanced performance and reduced resource utilization of the proposed methods. This study contributes to the ongoing development of DAG-based distributed ledgers, offering insights into more efficient algorithmic solutions for large-scale, decentralized networks.

In this paper, the optimization problem of orbital transfer strategy for orbital flyby observation missions is studied. A hybrid optimization method is proposed, which is improved to make it more suitable for satellite on-board computing. This new algorithm is designed to solve the initial value sensitivity problem of the sequential quadratic programming algorithm (SQP). It is consisted of the depth-first search algorithm (DFS) and the SQP algorithm and thus has the characteristics of fast convergence, high reliability, and good robustness. With this method, the DFS with a large step size is calculated first, and then the optimal value in the calculation result is used as the initial value of the SQP algorithm for further optimization. This method can obtain the approximate optimal solution available in engineering. The numerical simulation of an orbital transfer optimization problem is set to verify the effectiveness of the new hybrid algorithm. The simulation results compared with the genetic algorithm (GA) show that the proposed hybrid algorithm can effectively reduce the on-board resource occupation when getting similar results and thus can meet the needs of satellite on-board computing. [ABSTRACT FROM AUTHOR]

The IOTA Tangle, a Directed Acyclic Graph (DAG)-based distributed ledger, is popular for its scalability and suitability for IoT applications, offering fee-less transactions. A critical component of IOTA's architecture is the Cumulative Weight Calculation (CWC), essential for its tip selection mechanism. This paper introduces an optimization of the IOTA Reference Implementation (IRI) CWC process originally implemented using Breadth-First Search (BFS) by employing Depth-First Search (DFS) and Iterative Deepening Search (IDS) algorithms. We present a comparative analysis of these methods, demonstrating that DFS and IDS provide significant improvements in computational efficiency, particularly beneficial for IoT devices with limited processing capabilities. Our findings are substantiated through a series of experiments on a Tangle snapshot, highlighting the enhanced performance and reduced resource utilization of the proposed methods. This study contributes to the ongoing development of DAG-based distributed ledgers, offering insights into more efficient algorithmic solutions for large-scale, decentralized networks. [ABSTRACT FROM AUTHOR]

This study designs a two-stage algorithm to address the bid generation problem of carriers when adding new vehicle routes in the presence of the existing vehicle routes to provide transportation service. To obtain the best auction combination and bid price of the carrier, a hybrid integer nonlinear programming model is introduced. According to the characteristics of the problem, a set of two-stage hybrid algorithms is proposed, innovatively integrating block coding within a genetic algorithm framework with a depth-first search approach. This integration effectively manages routing constraints, enhancing the algorithm's efficiency. The block coding and each route serve as decision variables in the set partition formula, enabling a comprehensive exploration of potential solutions. After a simulation-based analysis, the algorithm has been comprehensively validated analytically and empirically. The improvement of this research lies in the effectiveness of the proposed algorithm, i.e., the ability to handle a broader range of problem scales with less time in addressing complex operator bid generation in combinatorial auctions. [ABSTRACT FROM AUTHOR]

The popularization of intelligent toys enriches the lives of the general public. To provide the public with a better toy experience, we propose the intelligent toy tracking method by the mobile cloud terminal deployment and depth-first search algorithm. Firstly, we construct a toy detection model via Transformer, which realizes the positioning of toys in the image through the refined region adaptive boundary representation. Then, using these detected continuous frames, we improve the toy tracking based on a depth-first search. Long-short-term memory (LSTM) constructs the continuous frame tracking structure, and the depth-first search mechanism is embedded to realize the accurate tracking of multiple targets in continuous frames. Finally, to realize the terminal marginalization of the proposed method, this chapter proposes a lightweight model deployment method based on mobile cloud terminals to realize the maintenance of the optimal machine state of intelligent toys. The experiment proves that our proposed target method can reach the world-leading level and obtain the mAP value of 0.858. Our tracking method can also perform excellently with a MOTA value of 0.916.

Abstract Mitosis is a critical criterion for meningioma grading. However, pathologists’ assessment of mitoses is subject to significant inter-observer variation due to challenges in locating mitosis hotspots and accurately detecting mitotic figures. To address this issue, we leverage digital pathology and propose a computational strategy to enhance pathologists’ mitosis assessment. The strategy has two components: (1) A depth-first search algorithm that quantifies the mathematically maximum mitotic count in 10 consecutive high-power fields, which can enhance the preciseness, especially in cases with borderline mitotic count. (2) Implementing a collaborative sphere to group a set of pathologists to detect mitoses under each high-power field, which can mitigate subjective random errors in mitosis detection originating from individual detection errors. By depth-first search algorithm (1) , we analyzed 19 meningioma slides and discovered that the proposed algorithm upgraded two borderline cases verified at consensus conferences. This improvement is attributed to the algorithm’s ability to quantify the mitotic count more comprehensively compared to other conventional methods of counting mitoses. In implementing a collaborative sphere (2) , we evaluated the correctness of mitosis detection from grouped pathologists and/or pathology residents, where each member of the group annotated a set of 48 high-power field images for mitotic figures independently. We report that groups with sizes of three can achieve an average precision of 0.897 and sensitivity of 0.699 in mitosis detection, which is higher than an average pathologist in this study (precision: 0.750, sensitivity: 0.667). The proposed computational strategy can be integrated with artificial intelligence workflow, which envisions the future of achieving a rapid and robust mitosis assessment by interactive assisting algorithms that can ultimately benefit patient management.

It is a very important process to determine whether there is a rigid sub-chain in the type synthesis for kinematic chains. In this paper, a new method for identifying and eliminating rigid sub-chains of kinematic chains with multiple joints is proposed. Firstly, the properties of the mesh loop set and independent loop set in loop combination were introduced. Properties determine that testing the degrees of freedom (DOF) of all independent loop sets and the combined loops can verify the existence of a rigid sub-chain. Secondly, representation methods and properties of the joint-joint adjacent matrix and loop matrix were explained and the operation method of the loop matrix was explained. A method of obtaining some loop matrices, which represents an independent loop set, by joint-joint adjacency matrix is proposed. Then, the DOF of all independent loop sets and the combined loops of independent loop sets were obtained by the operation method of loop matrix to identify and eliminate the rigid sub-chain. Finally, a large number of cases have proved the feasibility and effectiveness of this method. [ABSTRACT FROM AUTHOR]

In abstract argumentation frameworks, the computation of stable extensions is an important semantic task for evaluating the acceptability of arguments. The current approaches for the computation of stable extensions are typically conducted through methodologies that are either label-based or extension-based. Label-based algorithms operate by assigning labels to each argument, thus reducing the attack relations between arguments to constraint relations among the labels. This paper analyzes the existing two-label and three-label enumeration algorithms for stable extensions through case studies. It is found that both the two-label and three-label algorithms are not precise enough in defining types of arguments. To address these issues, this paper proposes a four-label enumeration algorithm for stable extensions. This method introduces amust_in label to pre-mark certain in-type arguments, thereby achieving a finer classification of in-type arguments. This enhances the labelings’ propagation ability and reduces the algorithm’s search space. Our proposed four-label algorithm was tested on authoritative benchmark sets of abstract argumentation framework problems: ICCMA 2019, ICCMA 2021, and ICCMA 2023. Experimental results show that the four-label algorithm significantly improves solving efficiency compared to existing two-label and three-label algorithms. Additionally, ablation experiments confirm that both the four-label transition strategy and preprocessing strategy enhance the algorithm’s performance.

This article presents a novel parallel path detection algorithm for identifying suspicious fraudulent accounts in large-scale banking transaction graphs. The proposed algorithm is based on a three-step approach that involves constructing a directed graph, shrinking strongly connected components, and using a parallel depth-first search algorithm to mark potentially fraudulent accounts. The algorithm is designed to fully exploit CPU resources and handle large-scale graphs with exponential growth. The performance of the algorithm is evaluated on various datasets and compared with serial time baselines. The results demonstrate that our approach achieves high performance and scalability on multi-core processors, making it a promising solution for detecting suspicious accounts and preventing money laundering schemes in the banking industry. Overall, our work contributes to the ongoing efforts to combat financial fraud and promote financial stability in the banking sector. [ABSTRACT FROM AUTHOR]

Sistem pakar merupakan sistem yang meniru kemampuan pakar dalam menyelesaikan masalah. Penelitian terdahulu telah membangun sistem pakar berbasis web maupun mobile di berbagai bidang. Kelemahan sistem pakar berbasis web maupun mobile adalah pengembangannya memerlukan waktu dan tenaga lebih untuk membuat antarmuka. Kelemahan lainnya adalah pengguna sistem pakar harus mempelajari antarmuka kurang umum. Solusi permasalahan tersebut adalah dengan memanfaatkan chatbot aplikasi chatting. Pengembang sistem pakar tidak perlu mengembangkan antarmuka khusus, sehingga dapat menghemat waktu dan tenaga. Pengguna juga tidak perlu mempelajari antarmuka baru karena masyarakat telah terbiasa dengan antarmuka aplikasi chatting. Penelitian ini membangun sistem pakar deteksi kerusakan sepeda motor dengan metode Depth-first Search berbasis chatbot Telegram. Hasil uji coba sistem pakar pada penelitian ini menunjukkan tingkat akurasi yang mencapai 90%. Oleh karena itu, penelitian berhasil membangun sistem pakar berbasis chatbot Telegram yang mempunyai kualitas yang baik dan bermanfaat bagi pengembang dan pengguna.

In recent years, Online Social Network (OSN) has been rapidly evolving and attracted many users. In OSN, users share sensitive information; therefore, effective access control models are needed to protect information from unauthorized users. Currently, Relational Based Access Control (ReBAC) is used to protect user's private information. The authorization policy in ReBAC is based on the relationship type and depth among users; however, it is not sufficient to protect private information such as location, time, and age. In this paper, attributes are added to the social graph to establish an efficient access control in OSN, then a policy model is proposed for the new Attribute Relation Based Access Control model (A-ReBAC), and unambiguous Hybrid Logic (HL) policy language is used to formulate the access control policy model. To evaluate the proposed policy model two path-checking algorithms (depth-first search (DFS) and breadth-first search (BFS)) are applied to real datasets, and the time spent on access requests is calculated in the social graph of these datasets. The results showed DFS takes less time than BFS to do the task defined. [ABSTRACT FROM AUTHOR]

In order to improve the operational capability of automatic train supervision (ATS), reduce the software and hardware cost, a dispatch path search method is proposed to linked devices such as platforms and signals. Based on the station map data and topology theory, the proposed method acquires the connection relationship of the platform in a semi-automatic or fully automatic manner, and the connection data of the platform is stored in the data structure of the adjacent matrix. The path search rule is designed, in which both depth-first search (DFS) and breadth-first search(BFS) are supported to obtain the path group. Automatic or manual methods to meet the project requirements are also proposed. Take the platform as an application case, the method is adapted to train dispatcher and route trigger function related to paths of linked devices. The path group can be reloaded by ATS subsystem, which can improve the ATS performance effectively. The method has been applied to Changsha urban rail transit Line 3 and Line 4, Wuxi urban rail transit Line 4, to support plan editing and train scheduling, and achieves good results, such as the generation time of station map can reduce over 70%,incidental jamming of the route trigger is eliminated when there are more trains,and route files can be generated automatically within minutes (original manual generation took several days).

This study designs a two-stage algorithm to address the bid generation problem of carriers when adding new vehicle routes in the presence of the existing vehicle routes to provide transportation service. To obtain the best auction combination and bid price of the carrier, a hybrid integer nonlinear programming model is introduced. According to the characteristics of the problem, a set of two-stage hybrid algorithms is proposed, innovatively integrating block coding within a genetic algorithm framework with a depth-first search approach. This integration effectively manages routing constraints, enhancing the algorithm’s efficiency. The block coding and each route serve as decision variables in the set partition formula, enabling a comprehensive exploration of potential solutions. After a simulation-based analysis, the algorithm has been comprehensively validated analytically and empirically. The improvement of this research lies in the effectiveness of the proposed algorithm, i.e., the ability to handle a broader range of problem scales with less time in addressing complex operator bid generation in combinatorial auctions.

In this paper, a compressible and low-cost cryptography system is investigated based a new memristor system for Internet of multimedia things to acquire and transmit multiclass-image signals, such as grayscale, color and 3D images (Stereolithography format). Compression sensing is first introduced to solve the limitation of computing and storage resources faced by detectors when sampling data from different scenarios. The batch data after sampling are fused into 3D cubes, and the confusion–diffusion framework is used to encrypt them in 3D space to ensure the privacy of information during transmission. For confusion procedure, a novel technique, like 3D jigsaw puzzles, is designed to improve the efficiency of scrambling. Furthermore, a new nonlinear primitive is proposed based on three-input majority gate , and the non-sequential diffusion path generated by depth-first search is used for diffusion operation. The experimental results and extensive security analysis show that the proposed encryption technique combined with the implementation of 3D space significantly improves the security and efficiency of batch image processing. [ABSTRACT FROM AUTHOR]

Most system restoration strategies following blackouts are obtained by offline methods, which are efficient in theory. But this may not be the case. A new method is proposed to decide the generator startup sequence during the system restoration. The Monte Carlo tree search algorithm is used to search for the generator to be recovered according to the real-time situation of the power system. All paths from the restored system to the chosen generator are obtained by Depth-First Search. And the K-shortest paths are obtained by Yen’s Algorithm to decide the best restoration path. Then, key constraints are verified. Especially, the constraint on the hot start of generators is considered. Finally, the feasible scheme with the maximum power generation is selected for implementation. In addition, the cranking process of G33 to G32 in the IEEE39-bus test system is taken as an example to illustrate the efficiency of this method. Depth-First Search detects 42 restoration paths, and Yen’s Algorithm selects three most reasonable paths from them, which greatly improves the efficiency of online decision making.

The problems of recognizing series–parallel graphs, outerplanar graphs, and generalized series–parallel graphs have been studied separately in the past. Efficient algorithms have been presented. However, none of the algorithms are certifying. A certifying algorithm generates, in addition to its answer, a certificate that can be used by a checker (a separate algorithm) to verify the correctness of the answer. The certificate is positive if the answer is 'yes', and is negative if the answer is 'no'. In this paper, an O (| E | + | V |) -time certifying algorithm that simultaneously determines if a graph G = (V , E) is series–parallel, outerplanar, or generalized series–parallel is presented. The positive certificates are a construction sequence for constructing G if G is series–parallel, a generalized construction sequence for constructing G if G is generalized series–parallel but not series–parallel, and the edge set of the exterior boundary of an outerplanar embedding of G if G is outerplanar. The negative certificates are forbidden subgraphs or forbidden structures of G. All these certificates are generated by making only one pass over G after a preprocessing step decomposing G into its biconnected components. [ABSTRACT FROM AUTHOR]

When synthesizing a gene with a long DNA sequence, it is usually necessary to divide it into several fragments. Based on these fragments, a set of oligonucleotides for gene assembly is produced. Each oligonucleotide is synthesized separately by the chemical reaction, and then the obtained oligonucleotides are assembled into the full gene sequence, in a specific environment, by polymerase chain reaction (PCR) or ligase chain reaction (LCR). In this paper, an effective and efficient algorithm to divide long genes into oligonucleotide sets is presented. First, according to the length of the overlapping oligonucleotide region, the long DNA sequence to be synthesized is divided into fragments of approximately equal length. Second, the length of these fragments is iterated to dynamically optimize the length of the overlapping regions to reduce melting temperature fluctuations. Then, the improved depth-first search algorithm is used according to the design principle of pruning optimization to obtain a uniform set of oligonucleotides with very close melting temperatures. This will decrease the errors in gene assembly with PCR or LCR. Lastly, the oligonucleotides that have homologous melting temperatures needed for PCR-based synthesis and two-step assembly of the target gene are deduced and outputted.

There have been many algorithms for solving the shortest path problems through necessary nodes set, but most of them are insufficient when applied to the circumstances with extra hard constraints. To solve this problem, this paper puts forward a random search algorithm based on depth first search mechanism. The mathematical models are provided by users according to actual situation, and they are abstracted as undirected weighted graphs. Some related variables such as the beginning node, the end node, necessary nodes and extra hard constraints are defined by path planning requirements, and they are saved as adjacency matrix. In the searching process, the feasibility of the path is determined in real time to meet the requirements of extra hard constraints. Finally, the optimal shortest path solution is obtained. Simulation and actual results show that the algorithm effectively avoids the influence of intermediate path under extra hard constraints and generate a shortest path, which means that this algorithm can greatly improve problem solvability. [ABSTRACT FROM AUTHOR]

Abstract In this study, the possibility of using the Modulor in the digital era is explored by employing computational methods to solve "The Panel Exercises" that Le Corbusier presented in his well‐known book "Modulor" and confirming the explosions of their combinations. First, the mathematical characteristics of the Modulor are reconfirmed. Thereafter, we present the gradual bisection method to generate combinations of the first panel exercise and a program of the depth‐first search to enumerate all the possible solutions of the second panel exercise. It is also demonstrated that the other panel exercises that evolved from the previous two exercises can be solved with these methods.

The procedure of advance protection against water hammer of products oil pipelines is generally implemented in the case of unplanned power failure of a pump station, which has a decisive influence on the safety of pipelines. To realize safe and smooth transition of pipelines from a steady state before the accident to a better operating state after the accident, an optimal scheme for advance protection against water hammer should be established. Water hammer conditions of product oil pipelines in the case of unplanned power failure of a pump station were analyzed. An optimization model was established with the objective of maximum throughput and minimum sum of throttling pressure of all regulating valves in the next steady state after water hammer accidents and under corresponding constraint conditions, i.e. no overpressure or vaporization throughout the pipeline. Two different strategies were used for model solution: translating hydraulic gradient line method for the downstream of the accident station through optimization with the objective of minimum outlet pressure of each station and depth-first search method for the upstream of the accident station through optimization of minimum total throttling pressure so as to improve the solving speed. The optimization model was applied in practice under the condition of a product oil pipeline in the case of unplanned power failure of a pump station. The control logic of the transient process in the water hammer accident was established based on the optimization result. Its rationality was verified by simulation software SPS which indicated the rationality and reliability of the optimization model.

The outsourcing of the design and manufacturing of integrated circuits (ICs) in the current horizontal semiconductor integration flow has posed various security threats due to the presence of untrusted entities, such as overproduction of ICs, sale of out-of-specification/rejected ICs, and piracy of Intellectual Properties (IPs). Consequently, logic locking emerged as one of the prominent design for trust techniques. Unfortunately, these locking techniques are now inclined to achieve complete Boolean satisfiability (SAT) resiliency after the seminal work published in Subramanyan et al. (in: International Symposium on Hardware 907 Oriented Security and Trust, pp 137–143, 2015). In this paper, we propose a novel oracle-less attack that is based on the topological analysis of the locked netlist even though it is SAT-resilient. The attack relies on identifying and constructing unit functions with a hypothesis key to be searched in the entire netlist to find its replica. The proposed graph search algorithm efficiently finds the duplicate functions in the netlist, making it a self-referencing attack. This proposed attack is extremely efficient and can determine the secret key within a few minutes. We have also proposed a countermeasure to make the circuit resilient against this topology-guided attack to progress toward a secure logic locking technique. [ABSTRACT FROM AUTHOR]

Digital mammography acts as a unique screening technology to protect the lives of females against breast cancer for the past few decades. Mammographic breast density is a well-known biomarker and plays a substantial role in breast cancer prediction and treatments. Breast density is calculated based on the opacity of fibro-glandular tissue reflected on digital mammograms concerning the whole area of the breast. The opacity of pectoral muscle and fibro-glandular tissue is similar to each other; hence, the small presence of the pectoral muscle in the breast area can hamper the accuracy of breast density classification. Successful removal of pectoral muscle is challenging due to changes in shape, size, and texture of pectoral muscle in every MLO and LMO views of mammogram. In this article, the depth-first search (DFS) algorithm is proposed to remove artifacts and pectoral muscle from digital mammograms. In the proposed algorithm, image enhancement is performed to improve the pixel quality of the input image. The whole breast as a single connected component is identified from the background region to remove the artifacts and tags. The depth-first search method with and without the heuristic approach is used to delineate the pectoral muscle, and then final suppression is performed on it. This algorithm is tested on 2675 images of the DDSM dataset, which is further divided into four density classes as per BIRADs classification. Segmentation results are calculated individually on each BIRADs density class of the DDSM dataset. Results are validated subjectively by the expert's Radiologist's ground truth with segmentation accuracy and objectively by the Jaccard coefficient and a dice similarity coefficient. This algorithm is found robust on each density class and provides overall segmentation accuracy of 86.18%, a mean value of Jaccard index, and a Dice similarity coefficient of 0.9315 and 0.9548, respectively. The experimental results show that the proposed algorithms applied for pectoral muscle removal follow the ground truth marked by an expert radiologist. The proposed algorithm can be part of the pre-processing unit of breast density measurement and breast cancer detection system used during clinical practice. [ABSTRACT FROM AUTHOR]

Cracks are important signs of degradation in existing infrastructure systems. Automatic crack detection and segmentation plays a key role in developing smart infrastructure systems. However, this field has been challenging over the last decades due to irregular shape of the cracks and complex illumination conditions. This article proposes a novel deep-learning architecture for crack segmentation at pixel-level. In this architecture, one convolutional layer is densely connected to multiple other layers in a feed-forward fashion. Max pooling layers are used to reduce the dimensions of the features, and transposed convolution layers are used for multi-level feature fusion. A depth-first search–based algorithm is applied as post-processing tool to remove isolated pixels and improve the accuracy. The method is tested on two previously published data sets. It can reach 92.02%, 91.13%, and 91.58% for the first data set, and 92.17%, 91.61%, and 91.89% for the second data set for precision, recall, and F1 score, respectively. The performance of the proposed method outperforms other state-of-the-art methods. At the end of the article, the influence of feature fusion methods and transfer learning are also discussed. [ABSTRACT FROM AUTHOR]