Graphics Processing Units Research Articles

Artificial intelligence changing the future of healthcare diagnostics

Artificial intelligence (AI) is a computer’s capacity to carry out operations like speech and image recognition and decision-making that ordinarily require human intelligence. Healthcare is using AI to automate tasks such as medical image analysis and diagnosis that require high precision and accuracy. The healthcare industry is significantly impacted by the rapid development of machine learning algorithms, which are frequently implemented using deep learning, as well as the growth of digital data and computing power supported by improvements in hardware technologies. Significant progress has been made in the field of artificial intelligence in recent years and is now widely used in healthcare to automate a variety of tasks, which require a high degree of accuracy and precision. The creation of machine learning algorithms, which can learn from data and make predictions based on that learning, has made it possible to use AI in healthcare. Neural networks are used in deep learning, a subfield of machine learning, to simulate how the human brain functions. Crucial advances have been made in clinical decision support, drug discovery, and medical imaging. Furthermore, the rapid development of hardware technologies, such as graphics processing units, has allowed AI systems to process enormous amounts of data quickly and accurately. Due to this, AI-based tools and platforms can help healthcare professionals with tasks such as patient monitoring, disease diagnosis, and treatment planning.

Optimizing MRI Data Processing by exploiting GPU Acceleration for Efficient Image Analysis and Reconstruction

Magnetic Resonance Imaging (MRI) is a non-invasive medical imaging modality, offering detailed anatomical insights without ionizing radiation The advent of Graphics Processing Units (GPUs) has stimulated a paradigm shift, propelling MRI techniques to new frontiers. Through parallel processing capabilities, GPUs have expedited real-time imaging, complex image reconstruction, noise reduction, and intricate data analysis. This paper provides an extensive survey on the GPUs based MRI techniques and its synergistic impact on core methodologies such as Diffusion Tensor Imaging (DTI) and Functional MRI (fMRI). Through parallel processing and frameworks like CUDA and OpenCL, GPUs have overcome computational hurdles in MRI data processing. Challenges like memory constraints and data transfer bottlenecks are addressed through hybrid CPU-GPU strategies and algorithmic enhancements. The integration of GPUs yields faster scans, enhanced image quality, and real-time insights, benefiting patient care and accelerating medical research. Considering the ethical issues regarding patient data privacy and algorithmic fairness, GPUs' potential in MRI research and development is evident. This paper concludes by looking ahead about the future of GPU based MRI, urging further exploration to uncover new possibilities and shape a transformative path for the future of medical imaging.

GPU versus FPGA implementation of a digital predistortion linearizer for wideband radiofrequency power amplifiers

This paper presents and compares the implementation of a digital predistortion (DPD) linearizer for radiofrequency power amplifiers (PAs) considering two different hardware acceleration platforms. Both graphics processing unit (GPU)-based and field programmable gate arrays (FPGA)-based implementations of the DPD are capable to meet the high throughput requirements for linearizing the PA with current 5G new radio (NR) wideband signals. A comparison in terms of hardware resources usage, precision, throughput and linearization performance is provided for both GPU and FPGA implementations of a DPD based on the generalized memory polynomial (GMP) behavioral model.

BPMF: A Backprojection and Matched-Filtering Workflow for Automated Earthquake Detection and Location

Abstract We introduce BPMF (backprojection and matched filtering)—a complete and fully automated workflow designed for earthquake detection and location, and distributed in a Python package. This workflow enables the creation of comprehensive earthquake catalogs with low magnitudes of completeness using no or little prior knowledge of the study region. BPMF uses the seismic wavefield backprojection method to construct an initial earthquake catalog that is then densified with matched filtering. BPMF integrates recent machine learning tools to complement physics-based techniques, and improve the detection and location of earthquakes. In particular, BPMF offers a flexible framework in which machine learning detectors and backprojection can be harmoniously combined, effectively transforming single-station detectors into multistation detectors. The modularity of BPMF grants users the ability to control the contribution of machine learning tools within the workflow. The computation-intensive tasks (backprojection and matched filtering) are executed with C and CUDA-C routines wrapped in Python code. This leveraging of low-level, fast programming languages and graphic processing unit acceleration enables BPMF to efficiently handle large datasets. Here, we first summarize the methodology and describe the application programming interface. We then illustrate BPMF’s capabilities to characterize microseismicity with a 10 yr long application in the Ridgecrest, California area. Finally, we discuss the workflow’s runtime scaling with numerical resources and its versatility across various tectonic environments and different problems.

Document Classification Using Lightweight Neural Network

<p>In recent years, OCR data has been used for learning and analyzing document classification. In addition, some neural networks have used image recognition for training, such as the network published by the ImageNet Large Scale Visual Recognition Challenge for document image training, AlexNet, GoogleNet, and MobileNet. Document image classification is important in data extraction processes and often requires significant computing power. Furthermore, it is difficult to implement image classification using general computers without a graphics processing unit (GPU). Therefore, this study proposes a lightweight neural network application that can perform document image classification on general computers or the Internet of Things (IoT) without a GPU. Plustek Inc. provided 3065 receipts belonging to 58 categories. Three datasets were considered as test samples while the remaining were considered as training samples to train the network to obtain a classifier. After the experiments, the classifier achieved 98.26% accuracy, and only 3 out of 174 samples showed errors.</p> <p> </p>

Accelerating Static Timing Analysis Using CPU–GPU Heterogeneous Parallelism

Static timing analysis (STA) is an essential yet time-consuming task during the circuit design flow to ensure the correctness and performance of the design. Thanks to the advancement of general-purpose computing on graphics processing units (GPUs), new possibilities and challenges have arisen for boosting the performance of STA. In this work, we present an efficient and holistic GPU-accelerated STA engine. We accelerate major STA tasks, including levelization, delay computation, graph propagation, and multi-corner analysis, by developing high-performance GPU kernels and data structures. By dividing the STA workloads into CPU-GPU concurrent tasks with managed dependencies, our acceleration framework supports versatile incremental updates. Furthermore, we have extended our approach to multi-corner analysis by exploring a large amount of corner-level data parallelism using GPU computing. Our implementation based on the open-source STA engine OpenTimer has achieved up to 4.07× speed-up on single corner analysis, and up to 25.67× speed-up on multi-corner analysis on TAU 2015 contest designs and a 14nm technology.

SOKS: Automatic Searching of the Optimal Kernel Shapes for Stripe-Wise Network Pruning.

In spite of the remarkable performance, deep convolutional neural networks (CNNs) are typically over-parameterized and computationally expensive. Network pruning has become a popular approach to reducing the storage and calculations of CNN models, which commonly prunes filters in a structured way or discards single weights without structural constraints. However, the redundancy in convolution kernels and the influence of kernel shapes on the performance of CNN models have attracted little attention. In this article, we develop a framework, termed searching of the optimal kernel shape (SOKS), to automatically search for the optimal kernel shapes and perform stripe-wise pruning (SWP). To be specific, we introduce coefficient matrices regularized by a variety of regularization terms to locate important kernel positions. The optimal kernel shapes not only provide appropriate receptive fields for each convolution layer, but also remove redundant parameters in convolution kernels. SWP is also achieved by utilizing these irregular kernels and actual inference speedups on the graphics processing unit (GPU) are obtained. Comprehensive experimental results demonstrate that SOKS searches high-efficiency kernel shapes and achieves superior performance in terms of both compression ratio and inference latency. Embedding the searched kernels into VGG-16 increases the accuracy from 93.53% to 94.26% on CIFAR-10, while pruning 59.27% model parameters and reducing 27.07% inference latency.

ADEROS: Artificial Intelligence-Based Detection System of Critical Events for Road Security

The deployment of artificial intelligence (AI) in Intelligent Transportation Systems (ITS), especially in the field of Intelligent Transportation Cyber-Physical Systems (ITCPS) has a strong potential to achieve higher efficiency, reliability, and increased safety in both transportation and traffic. This work focuses on the real-world implementation of ITCPS, in which structure and elements in combination with advanced image processing methods increase safety and fluidity of road traffic at crossroads and railway crossings. In this work, we present a novel system called Artificial Intelligence-based Detection System for Road Security (ADEROS), which combines elements of CPS systems, object detection, and classification, computer vision (CV) which analyzes vehicle trajectory tracking, vehicle and pedestrian presence, light signaling systems, railway barriers at railway crossings, and railway warnings. The presented system is based on a camera module that is suitably positioned to capture the entire scene. The module uses graphics processing units (GPU) for accelerated image processing techniques and the YOLOv4 deep neural network model to detect traffic participants and then dangerous situations in various crossroads and railway crossings. Our improved unique detector can distinguish between individual types of road users and the status of several safety devices at crossroads and railway crossings (for example, the state of traffic lights (TL) or rail barriers). Furthermore, we present experimental implementation details of the ADEROS system, which includes a central server web interface for live traffic situation monitoring, various communication channels for the camera module, and a central server based on.NET core, Cassandra DB, and different security protocols. All data from risky situations are evaluated and transferred to the central server securely without human intervention. The central server aggregates and archives all risky situational data from connected cameras. Finally, we present our experimental results from a real-world pilot project that consists of a camera module prototype deployed in a real crossroad and an operational central web server.

GPU-accelerated polyenergetic forward projection for 9 MeV industrial CT system

Polyenergetic forward projection has great significance in inspecting hazardous materials, establishing optimal radiographic variables and investigating beam hardening effects. However, it is computationally intensive to perform polyenergetic forward-projection calculations for high-resolution phantoms. To address this issue, a rapid polyenergetic forward-projection algorithm is proposed for a 9 MeV industrial computed tomography (CT) system. The FLUktuierende KAskade (FLUKA) software package is used to generate the 9 MeV X-ray spectrum data. Two voxelised phantoms are used to model scanned objects, one being a multi-material cylinder and the other a single-material turbine blade. An incremental version of Siddon's algorithm is adopted to calculate the intersection lengths between the X-rays and the auxiliary phantoms. Three strategies are utilised to accelerate the calculation, in which: the intersection lengths do not vary with the energy bins and can be used repeatedly until all the energy bins are counted; a graphics processing unit (GPU) is used to accelerate the ray tracing algorithm by utilising a parallel computing technique; and faster memory access is achieved by binding the auxiliary phantoms to texture objects. The simulation results in this paper show that the GPU-based approach not only maintains the image precision but also gains significant speed-ups over the conventional central processing unit (CPU)-based Siddon method. Furthermore, beam hardening artefacts can clearly be seen from the profile curves of the reconstructed slices, indicating that this method is effective.

Improving graphics processing unit performance based on neural network direct memory access controller

In this paper proposes the design and implementation of the back propagation algorithm based neural network DMA (Direct Memory Access) Controller for use of multimedia applications. The proposed DMA controller work with the back propagation-training algorithm. The advantages of the back propagation algorithm it will be work on the gradient loss w.r.t the network weights. So this back propagation algorithm is used as training algorithm for the DMA controller. The proposed method is test with the different workload characteristics like heavy workload, medium workload and normal workload. The performance parameters are considered here is like accuracy, precision, recall and F1 score etc. The proposed method is compared with existing methods like CNN (Convolutional Neural Network), RNN (Recurrent Neural Network), LSTM (Long Sort term Memory) and GRU (Gated Recurrent Unit) etc. Finally, the proposed design will give the better performance than existing methods.

Exponential distribution functions for positivity-preserving lattice Boltzmann schemes: Application to 2D compressible flow simulations

A family of positivity-preserving lattice Boltzmann methods (LBMs) is proposed for compressible flow simulations in the continuum regime. It relies on the efficient collide-and-stream algorithm with a collision step based on exponential distribution functions. The latter serves as a generalization of Grad's post-collision distribution functions for which here (1) the linearized non-equilibrium contributions are replaced by their exponential forms and (2) the number of these contributions can be chosen arbitrary. In practice, post-collision moments of our exponential formulation are enforced through an iterative moment-matching approach to recover any macroscopic physics of interest, with or without external forces. This methodology directly flows from the extended framework on numerical equilibria [J. Latt et al., Philos. Trans. R. Soc. A 378, 20190559 (2020)] and goes one step further by allowing for the independent relaxation of hydrodynamic and high-order modes in a given moment space, notably, making the Prandtl number freely adjustable. The model is supplemented by a shock-capturing technique, based on the deviation of non-equilibrium moments from their equilibrium counterparts, to ensure good numerical properties of the model in inviscid and under-resolved conditions. A second exponential distribution accounts for extra degrees of freedom of molecules and allows for the simulation of polyatomic gases. To validate this novel approach and to quantify the accuracy of different lattices and moment closures, several 2D benchmark tests of increasing complexity are considered: double shear layer, linear wave decay, Poiseuille flow, Riemann problem, compressible Blasius flow over a flat plate, and supersonic flow past an airfoil. Corresponding results confirm the accuracy and stability properties of our approach for the simulation of compressible flows with LBMs. Eventually, the performance analysis further highlights its efficiency on general purpose graphical processing units.

Martian global current systems and related solar wind energy transfer: hybrid simulation under nominal conditions

ABSTRACT The magnetized solar wind drives a current system around Mars that maintains its induced magnetosphere. The solar wind also transfers its energy to the atmospheric ions, causing continuous atmospheric erosion, which has a profound impact on the planet’s evolution history. Here, we use Amitis, a Graphics Processing Unit (GPU)-based hybrid plasma model to first reproduce the global pattern of the net electric current and ion currents under an interplanetary magnetic field perpendicular to the solar wind flow direction. The resultant current distribution matches the observations and reveals more details. Using the electric field distribution characterized earlier with the same model, we calculate for the first time the spatial distribution of energy transfer rate to the plasmas in general and to different ion species at Mars. We find out that (1) the solar wind kinetic energy is the dominant energy source that drives Martian induced magnetosphere, (2) the energy flux of the shocked solar wind flows from the magnetic equatorial plane towards the plasma sheet in the induced magnetotail, (3) both the bow shock and the induced magnetospheric boundary are dynamos where plasma energy is transferred to the electromagnetic field, and (4) the planetary ions act as loads and gain energy from the electromagnetic field. The most intense load region is the planetary ion plume. The general pattern of the energy transfer rate revealed in this study is common for induced magnetospheres. Its variabilities with the upstream conditions can provide physical insight into the observed ion escape variabilities.

Effect of the Configuration and Shape of External Ribs of Sealed Enclosures of Electronic Devices on Heat Removal Efficiency

Introduction. Modern computing and electronic devices are constructed on the basis of radio-electronic components, such as processors, graphics processing units, etc. During operation, these components emit tens of watts of thermal energy. Therefore, effective excess heat removal from both semiconductor electronic devices and electronic systems as a whole through the use of passive or active cooling systems represents an important research problem. Aim. To study the influence of the configuration and shape of external ribs of sealed enclosures of electronic devices, which use solely passive cooling systems inside, on the efficiency of heat removal from the processor for eachenclosure design under consideration based on their comparative analysis. Materials and methods. Simulation experiments were carried out using 3D parametric models of various device types, which were developed in the SolidWorks Flow Simulation software environment. These models differed in terms of configuration of thermal channels formed by the external enclosure ribs. Results. The conducted simulation experiments allowed the authors to study the cooling process of processors installed in modern electronic devices. The influence of the configuration and shape of the enclosure ribs on excess heat removal from the processor was studied in a passive cooling mode and when blowing the devices with air moving from above (perpendicular to the cover) or laterally (parallel to the cover) with a gradual increase in a processor power from 10 to 25 W. A ribbed enclosure with passive cooling was shown to ensure a more effective heat removal from a 10 W processor compared to a non-ribbed enclosure (the temperature drop is 4.1 °C). For a 25 W processor, this value comprises 11.01 °C. When blowing the device, the direction (perpendicular or parallel) of air movement significantly affects the cooling efficiency of the heated surface (with a processor power of 45 W, the difference is more than 10 °C). Conclusion. The developed 3D models effectively simulate the cooling system of heat-loaded high-power radio-electronic components located in sealed enclosures, due to the implementation of their external ribbing.

Implementation and evaluation of two parallel computational models for the simulation of a long‐haul DWDM system limited by FWM

AbstractFour‐Wave Mixing (FWM) is one of the non‐linear phenomena affecting long‐reach communication systems and high bandwidth. Research communities use simulation tools for parameter optimization. Unfortunately, such a simulation is time‐consuming and requires more time as the number of channels increases. This paper proposes two fast implementations of Dense Wavelength Division Multiplexing (DWDM) system, limited by FWM and the intrinsic Amplified Spontaneous Emission (ASE) noise of optical amplifiers employed in each segment. Additionally, this work compares the efficiency and speed improvement of the proposed parallelization model versus an earlier sequential model. We present the computational complexity analysis of sequential and parallel models. The paper considers two different parallel implementations: a multicore processor using Open MultiProcessing (OpenMP) and Compute Unified Device Architecture (CUDA), which is based on the use of a Graphics Processing Unit (GPU). Results show that parallelism using CUDA improves by up to 70 times the simulation performance compared to the sequential model. Parallelism with CUDA is up to 15 times compared to OpenMP using 12 logical processors. It is possible to simulate an increased number of channels within our parallel execution, which was impractical in the sequential simulation.

MRT-lattice Boltzmann simulation of MHD natural convection of Bingham nanofluid in a C-shaped enclosure with response surface analysis

This paper focuses on the natural convection of heat transfer using magnetohydrodynamic (MHD) Bingham nanofluid. Utilizing the multiple-relaxation-time (MRT) lattice Boltzmann method (LBM) within a C-shaped enclosure and the NVIDIA graphics processing unit (GPU)-based compute unified architecture (CUDA) C/C++ platform, the simulation is carried out numerically. Inside the cavity, the base fluid is water and the nanofluid is Al2O3. Boundary conditions are presented in accordance with the heated, cold, and adiabatic conditions present in the cavity's various walls. Several parameters including Bingham number (Bn=0,0.5,1,1.5,2), Rayleigh number (Ra=104,105,106), Hartmann number (Ha=0,10,20,30), and nanoparticle volume fraction (ϕ=0,0.01,0.02,0.03,0.04). The results of the numerical simulation are shown using streamlines and isotherms, velocity-temperature, Local Nusselt number, and average Nusselt number. From the obtained results it is found that for the variation of Bn, Ha, and ϕ with different Ra the rate of heat transfer decreases along the bottom wall and increases for the left and top walls. The average Nusselt number decreases while Bn, Ha rises. On the other contrary, the average Nu increases as ϕ increases. Response surface methodology (RSM) is added here to have better understanding of the effects of parameter used. RSM includes statistical table for the combination of data set and their graphs to understand the accuracy. Moreover, regression analysis shows how average Nu increases or decreases with the variation of different parameters. The C-shaped geometry provides an excellent option in heat exchanging or electronic cooling equipment the chip designing technology. This study is only for two-dimensional laminar flow.

Bringing uncertainty quantification to the extreme-edge with memristor-based Bayesian neural networks

Safety-critical sensory applications, like medical diagnosis, demand accurate decisions from limited, noisy data. Bayesian neural networks excel at such tasks, offering predictive uncertainty assessment. However, because of their probabilistic nature, they are computationally intensive. An innovative solution utilizes memristors’ inherent probabilistic nature to implement Bayesian neural networks. However, when using memristors, statistical effects follow the laws of device physics, whereas in Bayesian neural networks, those effects can take arbitrary shapes. This work overcome this difficulty by adopting a variational inference training augmented by a “technological loss”, incorporating memristor physics. This technique enabled programming a Bayesian neural network on 75 crossbar arrays of 1,024 memristors, incorporating CMOS periphery for in-memory computing. The experimental neural network classified heartbeats with high accuracy, and estimated the certainty of its predictions. The results reveal orders-of-magnitude improvement in inference energy efficiency compared to a microcontroller or an embedded graphics processing unit performing the same task.

GPU-Accelerated Signal Processing for Passive Bistatic Radar

Passive bistatic radar is a novel radar technology that passively detects targets without actively emitting signals. Since passive bistatic radar entails larger data volumes and computations compared to traditional active radiation radar, the development of hardware and software platforms capable of efficiently processing signals from passive bistatic radar has emerged as a research focus in this field. This research investigates the signal processing flow of passive bistatic radar based on its characteristics and devises a parallel signal processing scheme under graphic processing unit (GPU) architecture for computation-intensive tasks. The proposed scheme utilizes high-computing-power GPU as the hardware platform and compute unified device architecture (CUDA) as the software platform and optimizes the extensive cancellation algorithm batches (ECA-B), range Doppler and constant false alarm detection algorithms. The detection and tracking of a single target are realized on the passive bistatic radar dataset of natural scenarios, and experiments show that the design of this algorithm can achieve a maximum acceleration ratio of 113.13. Comparative experiments conducted with varying data volumes revealed that this method significantly enhances the signal processing rate for passive bistatic radar.

WFA-GPU: gap-affine pairwise read-alignment using GPUs.

Advances in genomics and sequencing technologies demand faster and more scalable analysis methods that can process longer sequences with higher accuracy. However, classical pairwise alignment methods, based on dynamic programming (DP), impose impractical computational requirements to align long and noisy sequences like those produced by PacBio and Nanopore technologies. The recently proposed wavefront alignment (WFA) algorithm paves the way for more efficient alignment tools, improving time and memory complexity over previous methods. However, high-performance computing (HPC) platforms require efficient parallel algorithms and tools to exploit the computing resources available on modern accelerator-based architectures. This paper presents WFA-GPU, a GPU (graphics processing unit)-accelerated tool to compute exact gap-affine alignments based on the WFA algorithm. We present the algorithmic adaptations and performance optimizations that allow exploiting the massively parallel capabilities of modern GPU devices to accelerate the alignment computations. In particular, we propose a CPU-GPU co-design capable of performing inter-sequence and intra-sequence parallel sequence alignment, combining a succinct WFA-data representation with an efficient GPU implementation. As a result, we demonstrate that our implementation outperforms the original multi-threaded WFA implementation by up to 4.3× and up to 18.2× when using heuristic methods on long and noisy sequences. Compared to other state-of-the-art tools and libraries, the WFA-GPU is up to 29× faster than other GPU implementations and up to four orders of magnitude faster than other CPU implementations. Furthermore, WFA-GPU is the only GPU solution capable of correctly aligning long reads using a commodity GPU. WFA-GPU code and documentation are publicly available at https://github.com/quim0/WFA-GPU.

Soft- and Hard-Kill Hybrid GPU-based Bi-Directional Evolutionary Structural Optimization

Abstract Bi-directional evolutionary structural optimization (BESO) is a well-recognized method for generating optimal topologies of structures. Its soft-kill variant has a high computational cost, especially for large-scale structures, whereas the hard-kill variant often faces convergence issues. Addressing these issues, this paper proposes a hybrid BESO model tailored for graphics processing units (GPUs) by combining the soft-kill and hard-kill approaches for large-scale structures. A GPU-based algorithm is presented for dynamically isolating the solid/hard elements from the void/soft elements in the finite element analysis (FEA) stage. The hard-kill approach is used in the FEA stage with an assembly-free solver to facilitate the use of high-resolution meshes without exceeding the GPU memory limits, whereas for the rest of the optimization procedure, the soft-kill approach with a material interpolation scheme is implemented. Furthermore, the entire BESO method pipeline is accelerated for both the proposed hybrid and the standard soft-kill BESO. The comparison of the hybrid BESO with the GPU-accelerated soft-kill BESO using four benchmark problems with more than two million degrees of freedom reveals three key benefits of the proposed hybrid model: reduced execution time, decreased memory consumption, and improved FEA convergence, all of which mitigate the major computational issues associated with BESO.

Artificial Intelligence-Based Methods for Integrating Local and Global Features for Brain Cancer Imaging: Scoping Review.

Transformer-based models are gaining popularity in medical imaging and cancer imaging applications. Many recent studies have demonstrated the use of transformer-based models for brain cancer imaging applications such as diagnosis and tumor segmentation. This study aims to review how different vision transformers (ViTs) contributed to advancing brain cancer diagnosis and tumor segmentation using brain image data. This study examines the different architectures developed for enhancing the task of brain tumor segmentation. Furthermore, it explores how the ViT-based models augmented the performance of convolutional neural networks for brain cancer imaging. This review performed the study search and study selection following the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) guidelines. The search comprised 4 popular scientific databases: PubMed, Scopus, IEEE Xplore, and Google Scholar. The search terms were formulated to cover the interventions (ie, ViTs) and the target application (ie, brain cancer imaging). The title and abstract for study selection were performed by 2 reviewers independently and validated by a third reviewer. Data extraction was performed by 2 reviewers and validated by a third reviewer. Finally, the data were synthesized using a narrative approach. Of the 736 retrieved studies, 22 (3%) were included in this review. These studies were published in 2021 and 2022. The most commonly addressed task in these studies was tumor segmentation using ViTs. No study reported early detection of brain cancer. Among the different ViT architectures, Shifted Window transformer-based architectures have recently become the most popular choice of the research community. Among the included architectures, UNet transformer and TransUNet had the highest number of parameters and thus needed a cluster of as many as 8 graphics processing units for model training. The brain tumor segmentation challenge data set was the most popular data set used in the included studies. ViT was used in different combinations with convolutional neural networks to capture both the global and local context of the input brain imaging data. It can be argued that the computational complexity of transformer architectures is a bottleneck in advancing the field and enabling clinical transformations. This review provides the current state of knowledge on the topic, and the findings of this review will be helpful for researchers in the field of medical artificial intelligence and its applications in brain cancer.