Towards real-time photorealistic 3D holography with deep neural networks.

The ability to present three-dimensional (3D) scenes with continuous depth sensation has a profound impact on virtual and augmented reality, human-computer interaction, education and training. Computer-generated holography (CGH) enables high-spatio-angular-resolution 3D projection via numerical simulation of diffraction and interference1. Yet, existing physically based methods fail to produce holograms with both per-pixel focal control and accurate occlusion2,3. The computationally taxing Fresnel diffraction simulation further places an explicit trade-off between image quality and runtime, making dynamic holography impractical4. Here we demonstrate a deep-learning-based CGH pipeline capable of synthesizing a photorealistic colour 3D hologram from a single RGB-depth image in real time. Our convolutional neural network (CNN) is extremely memory efficient (below 620 kilobytes) and runs at 60 hertz for a resolution of 1,920 × 1,080 pixels on a single consumer-grade graphics processing unit. Leveraging low-power on-device artificial intelligence acceleration chips, our CNN also runs interactively on mobile (iPhone 11 Pro at 1.1 hertz) and edge (Google Edge TPU at 2.0 hertz) devices, promising real-time performance in future-generation virtual and augmented-reality mobile headsets. We enable this pipeline by introducing a large-scale CGH dataset (MIT-CGH-4K) with 4,000 pairs of RGB-depth images and corresponding 3D holograms. Our CNN is trained with differentiable wave-based loss functions5 and physically approximates Fresnel diffraction. With an anti-aliasing phase-only encoding method, we experimentally demonstrate speckle-free, natural-looking, high-resolution 3D holograms. Our learning-based approach and the Fresnel hologram dataset will help to unlock the full potential of holography and enable applications in metasurface design6,7, optical and acoustic tweezer-based microscopic manipulation8-10, holographic microscopy11 and single-exposure volumetric 3D printing12,13.

Time-resolved cryo-EM using a combination of droplet microfluidics with on-demand jetting.

Single-particle cryogenic electron microscopy (cryo-EM) allows reconstruction of high-resolution structures of proteins in different conformations. Protein function often involves transient functional conformations, which can be resolved using time-resolved cryo-EM (trEM). In trEM, reactions are arrested after a defined delay time by rapid vitrification of protein solution on the EM grid. Despite the increasing interest in trEM among the cryo-EM community, making trEM samples with a time resolution below 100 ms remains challenging. Here we report the design and the realization of a time-resolved cryo-plunger that combines a droplet-based microfluidic mixer with a laser-induced generator of microjets that allows rapid reaction initiation and plunge-freezing of cryo-EM grids. Using this approach, a time resolution of 5 ms was achieved and the protein density map was reconstructed to a resolution of 2.1 Å. trEM experiments on GroEL:GroES chaperonin complex resolved the kinetics of the complex formation and visualized putative short-lived conformations of GroEL-ATP complex.

All-optical spiking neurosynaptic networks with self-learning capabilities.

Software implementations of brain-inspired computing underlie many important computational tasks, from image processing to speech recognition, artificial intelligence and deep learning applications. Yet, unlike real neural tissue, traditional computing architectures physically separate the core computing functions of memory and processing, making fast, efficient and low-energy computing difficult to achieve. To overcome such limitations, an attractive alternative is to design hardware that mimics neurons and synapses. Such hardware, when connected in networks or neuromorphic systems, processes information in a way more analogous to brains. Here we present an all-optical version of such a neurosynaptic system, capable of supervised and unsupervised learning. We exploit wavelength division multiplexing techniques to implement a scalable circuit architecture for photonic neural networks, successfully demonstrating pattern recognition directly in the optical domain. Such photonic neurosynaptic networks promise access to the high speed and high bandwidth inherent to optical systems, thus enabling the direct processing of optical telecommunication and visual data.

Advancing convergence research: Renewable energy solutions for off-grid communities.

Millions of people across the world live off-grid not by choice but because they live in rural areas, have low income, and have no political clout. Delivering sustainable energy solutions to such a substantial amount of the world's population requires more than a technological fix; it requires leveraging the knowledge of underserved populations working together with a transdisciplinary team to find holistically derived solutions. Our original research has resulted in an innovative Convergence Framework integrating the fields of engineering, social sciences, and communication, and is based on working together with communities and other stakeholders to address the challenges posed by delivering clean energy solutions. In this paper, we discuss the evolution of this Framework and illustrate how this Framework is being operationalized in our on-going research project, cocreating hybrid renewable energy systems for off-grid communities in the Brazilian Amazon. The research shows how this Framework can address clean energy transitions, strengthen emerging industries at local level, and foster Global North-South scholarly collaborations. We do so by the integration of social science and engineering and by focusing on community engagement, energy justice, and governance for underserved communities. Further, this solution-driven Framework leads to the emergence of unique approaches that advance scientific knowledge, while at the same time addressing community needs.

Assessment and validation of glottic motion using cone-beam CT and real-time cine MRI.

PURPOSE: This study aimed to assess the margin for the planning target volume (PTV) using the Van Herk formula. We then validated the proposed margin by real-time magnetic resonance imaging (MRI). METHODS: An analysis of cone-beam computed tomography (CBCT) data from early glottic cancer patients was performed to evaluate organ motion. Deformed clinical target volumes (CTV) after rigid registration were acquired using the Velocity program (Varian Medical Systems, Palo Alto, CA, USA). Systematic (Σ) and random errors (σ) were evaluated. The margin for the PTV was defined as 2.5 Σ + 0.7 σ according to the Van Herk formula. To validate this margin, we accrued healthy volunteers. Sagittal real-time cine MRI was conducted using the ViewRay system (ViewRay Inc., Oakwood Village, OH, USA). Within the obtained sagittal images, the vocal cord was delineated. The movement of the vocal cord was summed up and considered as the internal target volume (ITV). We then assessed the degree of overlap between the ITV and the PTV (vocal cord plus margins) by calculating the volume overlap ratio, represented as (ITV∩PTV)/ITV. RESULTS: CBCTs of 17 early glottic patients were analyzed. Σ and σ were 0.55 and 0.57 for left-right (LR), 0.70 and 0.60 for anterior-posterior (AP), and 1.84 and 1.04 for superior-inferior (SI), respectively. The calculated margin was 1.8 mm (LR), 2.2 mm (AP), and 5.3 mm (SI). Four healthy volunteers participated for validation. A margin of 3 mm (AP) and 5 mm (SI) was applied to the vocal cord as the PTV. The average volume overlap ratio between ITV and PTV was 0.92 (range 0.85-0.99) without swallowing and 0.77 (range 0.70-0.88) with swallowing. CONCLUSION: By evaluating organ motion by using CBCT, the margin was 1.8 (LR), 2.2 (AP), and 5.3 mm (SI). The margin acquired using CBCT fitted well in real-time cine MRI. Given that swallowing during radiotherapy can result in a substantial displacement, it is crucial to consider strategies aimed at minimizing swallowing and related motion.

Towards a genomics-informed, real-time, global pathogen surveillance system.

The recent Ebola and Zika epidemics demonstrate the need for the continuous surveillance, rapid diagnosis and real-time tracking of emerging infectious diseases. Fast, affordable sequencing of pathogen genomes - now a staple of the public health microbiology laboratory in well-resourced settings - can affect each of these areas. Coupling genomic diagnostics and epidemiology to innovative digital disease detection platforms raises the possibility of an open, global, digital pathogen surveillance system. When informed by a One Health approach, in which human, animal and environmental health are considered together, such a genomics-based system has profound potential to improve public health in settings lacking robust laboratory capacity.

Informed consent for artificial intelligence in emergency medicine: A practical guide.

As artificial intelligence (AI) expands its presence in healthcare, particularly within emergency medicine (EM), there is growing urgency to explore the ethical and practical considerations surrounding its adoption. AI holds the potential to revolutionize how emergency physicians (EPs) make clinical decisions, but AI's complexity often surpasses EPs' capacity to provide patients with informed consent regarding its use. This article underscores the crucial need to address the ethical pitfalls of AI in EM. Patient autonomy necessitates that EPs engage in conversations with patients about whether to use AI in their evaluation and treatment. As clinical AI integration expands, this discussion should become an integral part of the informed consent process, aligning with ethical and legal requirements. The rapid availability of AI programs, fueled by vast electronic health record (EHR) datasets, has led to increased pressure on hospitals and clinicians to embrace clinical AI without comprehensive system evaluation. However, the evolving landscape of AI technology outpaces our ability to anticipate its impact on medical practice and patient care. The central question arises: Are EPs equipped with the necessary knowledge to offer well-informed consent regarding clinical AI? Collaborative efforts between EPs, bioethicists, AI researchers, and healthcare administrators are essential for the development and implementation of optimal AI practices in EM. To facilitate informed consent about AI, EPs should understand at least seven key areas: (1) how AI systems operate; (2) whether AI systems are understandable and trustworthy; (3) the limitations of and errors AI systems make; (4) how disagreements between the EP and AI are resolved; (5) whether the patient's personally identifiable information (PII) and the AI computer systems will be secure; (6) if the AI system functions reliably (has been validated); and (7) if the AI program exhibits bias. This article addresses each of these critical issues, aiming to empower EPs with the knowledge required to navigate the intersection of AI and informed consent in EM.

Finite difference-embedded UNet for solving transcranial ultrasound frequency-domain wavefield.

Transcranial ultrasound imaging assumes a growing significance in the detection and monitoring of intracranial lesions and cerebral blood flow. Accurate solution of partial differential equation (PDE) is one of the prerequisites for obtaining transcranial ultrasound wavefields. Grid-based numerical solvers such as finite difference (FD) and finite element methods have limitations including high computational costs and discretization errors. Purely data-driven methods have relatively high demands on training datasets. The fact that physics-informed neural network can only target the same model limits its application. In addition, compared to time-domain approaches, frequency-domain solutions offer advantages of reducing computational complexity and enabling stable and accurate inversions. Therefore, we introduce a framework called FD-embedded UNet (FEUNet) for solving frequency-domain transcranial ultrasound wavefields. The PDE error is calculated using the optimal 9-point FD operator, and it is integrated with the data-driven error to jointly guide the network iterations. We showcase the effectiveness of this approach through experiments involving idealized skull and brain models. FEUNet demonstrates versatility in handling various input scenarios and excels in enhancing prediction accuracy, especially with limited datasets and noisy information. Finally, we provide an overview of the advantages, limitations, and potential avenues for future research in this study.

Conv3D-Based Video Violence Detection Network Using Optical Flow and RGB Data.

Detecting violent behavior in videos to ensure public safety and security poses a significant challenge. Precisely identifying and categorizing instances of violence in real-life closed-circuit television, which vary across specifications and locations, requires comprehensive understanding and processing of the sequential information embedded in these videos. This study aims to introduce a model that adeptly grasps the spatiotemporal context of videos within diverse settings and specifications of violent scenarios. We propose a method to accurately capture spatiotemporal features linked to violent behaviors using optical flow and RGB data. The approach leverages a Conv3D-based ResNet-3D model as the foundational network, capable of handling high-dimensional video data. The efficiency and accuracy of violence detection are enhanced by integrating an attention mechanism, which assigns greater weight to the most crucial frames within the RGB and optical-flow sequences during instances of violence. Our model was evaluated on the UBI-Fight, Hockey, Crowd, and Movie-Fights datasets; the proposed method outperformed existing state-of-the-art techniques, achieving area under the curve scores of 95.4, 98.1, 94.5, and 100.0 on the respective datasets. Moreover, this research not only has the potential to be applied in real-time surveillance systems but also promises to contribute to a broader spectrum of research in video analysis and understanding.

Effect of Light-Emitting Grid Panel on Indoor Aquaculture for Measuring Fish Growth.

This study is related to Smart Aqua Farm, which combines artificial intelligence (AI) and Internet of things (IoT) technology. This study aimed to monitor fish growth in indoor aquaculture while automatically measuring the average size and area in real time. Automatic fish size measurement technology is one of the essential elements for unmanned aquaculture. Under the condition of labor shortage, operators have much fatigue because they use a primitive method that samples the size and weight of fish just before fish shipment and measures them directly by humans. When this kind of process is automated, the operator's fatigue can be significantly reduced. Above all, after measuring the fish growth, predicting the final fish shipment date is possible by estimating how much feed and time are required until the fish becomes the desired size. In this study, a video camera and a developed light-emitting grid panel were installed in indoor aquaculture to acquire images of fish, and the size measurement of a mock-up fish was implemented using the proposed method.

YOLO-IHD: Improved Real-Time Human Detection System for Indoor Drones.

In the field of unmanned systems, the combination of artificial intelligence with self-operating functionalities is becoming increasingly important. This study introduces a new method for autonomously detecting humans in indoor environments using unmanned aerial vehicles, utilizing the advanced techniques of a deep learning framework commonly known as "You Only Look Once" (YOLO). The key contribution of this research is the development of a new model (YOLO-IHD), specifically designed for human detection in indoor using drones. This model is created using a unique dataset gathered from aerial vehicle footage in various indoor environments. It significantly improves the accuracy of detecting people in these complex environments. The model achieves a notable advancement in autonomous monitoring and search-and-rescue operations, highlighting its importance for tasks that require precise human detection. The improved performance of the new model is due to its optimized convolutional layers and an attention mechanism that process complex visual data from indoor environments. This results in more dependable operation in critical situations like disaster response and indoor rescue missions. Moreover, when combined with an accelerating processing library, the model shows enhanced real-time detection capabilities and operates effectively in a real-world environment with a custom designed indoor drone. This research lays the groundwork for future enhancements designed to significantly increase the model's accuracy and the reliability of indoor human detection in real-time drone applications.

Prediction of diabetes disease using an ensemble of machine learning multi-classifier models.

BACKGROUND AND OBJECTIVE: Diabetes is a life-threatening chronic disease with a growing global prevalence, necessitating early diagnosis and treatment to prevent severe complications. Machine learning has emerged as a promising approach for diabetes diagnosis, but challenges such as limited labeled data, frequent missing values, and dataset imbalance hinder the development of accurate prediction models. Therefore, a novel framework is required to address these challenges and improve performance. METHODS: In this study, we propose an innovative pipeline-based multi-classification framework to predict diabetes in three classes: diabetic, non-diabetic, and prediabetes, using the imbalanced Iraqi Patient Dataset of Diabetes. Our framework incorporates various pre-processing techniques, including duplicate sample removal, attribute conversion, missing value imputation, data normalization and standardization, feature selection, and k-fold cross-validation. Furthermore, we implement multiple machine learning models, such as k-NN, SVM, DT, RF, AdaBoost, and GNB, and introduce a weighted ensemble approach based on the Area Under the Receiver Operating Characteristic Curve (AUC) to address dataset imbalance. Performance optimization is achieved through grid search and Bayesian optimization for hyper-parameter tuning. RESULTS: Our proposed model outperforms other machine learning models, including k-NN, SVM, DT, RF, AdaBoost, and GNB, in predicting diabetes. The model achieves high average accuracy, precision, recall, F1-score, and AUC values of 0.9887, 0.9861, 0.9792, 0.9851, and 0.999, respectively. CONCLUSION: Our pipeline-based multi-classification framework demonstrates promising results in accurately predicting diabetes using an imbalanced dataset of Iraqi diabetic patients. The proposed framework addresses the challenges associated with limited labeled data, missing values, and dataset imbalance, leading to improved prediction performance. This study highlights the potential of machine learning techniques in diabetes diagnosis and management, and the proposed framework can serve as a valuable tool for accurate prediction and improved patient care. Further research can build upon our work to refine and optimize the framework and explore its applicability in diverse datasets and populations.

Quantitative Prediction of Inorganic Nanomaterial Cellular Toxicity via Machine Learning.

Organic chemistry has seen colossal progress due to machine learning (ML). However, the translation of artificial intelligence (AI) into materials science is challenging, where biological behavior prediction becomes even more complicated. Nanotoxicity is a critical parameter that describes their interaction with the living organisms screened in every bio-related research. To prevent excessive experiments, such properties have to be pre-evaluated. Several existing ML models partially fulfill the gap by predicting whether a nanomaterial is toxic or not. Yet, this binary categorization neglects the concentration dependencies crucial for experimental scientists. Here, an ML-based approach is proposed to the quantitative prediction of inorganic nanomaterial cytotoxicity achieving the precision expressed by 10-fold cross-validation (CV) Q2  = 0.86 with the root mean squared error (RMSE) of 12.2% obtained by the correlation-based feature selection and grid search-based model hyperparameters optimization. To provide further model flexibility, quantitative atom property-based nanomaterial descriptors are introduced allowing the model to extrapolate on unseen samples. Feature importance is calculated to find an interpretable model with optimal decision-making. These findings allow experimental scientists to perform primary in silico candidate screening and minimize the number of excessive, labor-intensive experiments enabling the rapid development of nanomaterials for medicinal purposes.

A combination predicting methodology based on T-LSTNet_Markov for short-term wind power prediction.

Wind power has been valued by countries for its renewability and cleanness and has become most of the focus of energy development in all countries. However, due to the uncertainty and volatility of wind power generation, making the grid-connected wind power system presents some serious challenges. Improving the accuracy of wind power prediction has become the focus of current research. Therefore, this paper proposes a combined short-term wind power prediction model based on T-LSTNet_markov to improve prediction accuracy. First, perform data cleaning and data preprocessing operations on the original data. Second, forecast using T-LSTNet model in original wind power data. Finally, calculate the error between the forecast value and the actual value. The k-means++ method and Weighted Markov process are used to correct errors and to get the result of the final prediction. The data that are collected from a wind farm in Inner Mongolia Autonomous Region, China, are selected as a case study to demonstrate the effectiveness of the proposed combined models. The empirical results show that the prediction accuracy is further improved after correcting errors.

Usability Evaluation of a Knowledge Graph-Based Dementia Care Intelligent Recommender System: Mixed Methods Study.

BACKGROUND: Knowledge graph-based recommender systems offer the possibility of meeting the personalized needs of people with dementia and their caregivers. However, the usability of such a recommender system remains unknown. OBJECTIVE: This study aimed to evaluate the usability of a knowledge graph-based dementia care intelligent recommender system (DCIRS). METHODS: We used a convergent mixed methods design to conduct the usability evaluation, including the collection of quantitative and qualitative data. Participants were recruited through social media advertisements. After 2 weeks of DCIRS use, feedback was collected with the Computer System Usability Questionnaire and semistructured interviews. Descriptive statistics were used to describe sociodemographic characteristics and questionnaire scores. Qualitative data were analyzed systematically using inductive thematic analysis. RESULTS: A total of 56 caregivers were recruited. Quantitative data suggested that the DCIRS was easy for caregivers to use, and the mean questionnaire score was 2.14. Qualitative data showed that caregivers generally believed that the content of the DCIRS was professional, easy to understand, and instructive, and could meet users' personalized needs; they were willing to continue to use it. However, the DCIRS also had some shortcomings. Functions that enable interactions between professionals and caregivers and that provide caregiver support and resource recommendations might be added to improve the system's usability. CONCLUSIONS: The recommender system provides a solution to meet the personalized needs of people with dementia and their caregivers and has the potential to substantially improve health outcomes. The next step will be to optimize and update the recommender system based on caregivers' suggestions and evaluate the effect of the application.

Investigating phoneme-dependencies of spherical voice directivity patterns II: Various groups of phonemes.

The substantial variation between articulated phonemes is a fundamental feature of human voice production. However, while the spectral and temporal aspects of the phonemes have been extensively studied, few have investigated the spatial aspects and analyzed phoneme-dependent differences in voice directivity. This paper extends our previous research focusing on the directivity patterns of selected vowels and fricatives [Pörschmann and Arend, J. Acoust. Soc. Am. 149(6), 4553-4564 (2021)] and examines different groups of phonemes, such as plosives, nasals, voiced alveolars, and additional fricatives. For this purpose, full-spherical voice directivity measurements were performed for 13 persons while they articulated the respective phonemes. The sound radiation was recorded simultaneously using a surrounding spherical microphone array with 32 microphones and then spatially upsampled to a dense sampling grid. Based on these upsampled datasets, the spherical voice directivity was studied, and phoneme-dependent variations were analyzed. The results show significant differences between the groups of phonemes. However, within three groups (plosives, nasals, and voiced alveolars), the differences are small, and the variations in the directivity index were statistically insignificant.

Anisotropic minimum dissipation subgrid-scale model in hybrid aeroacoustic simulations of human phonation.

This article deals with large-eddy simulations of three-dimensional incompressible laryngeal flow followed by acoustic simulations of human phonation of five cardinal English vowels, /ɑ, æ, i, o, u/. The flow and aeroacoustic simulations were performed in OpenFOAM and in-house code openCFS, respectively. Given the large variety of scales in the flow and acoustics, the simulation is separated into two steps: (1) computing the flow in the larynx using the finite volume method on a fine moving grid with 2.2 million elements, followed by (2) computing the sound sources separately and wave propagation to the radiation zone around the mouth using the finite element method on a coarse static grid with 33 000 elements. The numerical results showed that the anisotropic minimum dissipation model, which is not well known since it is not available in common CFD software, predicted stronger sound pressure levels at higher harmonics, and especially at first two formants, than the wall-adapting local eddy-viscosity model. The model on turbulent flow in the larynx was employed and a positive impact on the quality of simulated vowels was found.

A Moving Magnetic Grid-Type Long-Range Linear Absolute Displacement Sensor.

In order to overcome the shortcomings of the traditional magnetic absolute linear displacement sensors in which cables affect the flexibility and measurement range in linear motor transmission systems, this paper proposes a novel cable-free moving magnetic grid-type long-range absolute displacement sensor. The sensor consists of a magnetic grid and a signal acquisition board. The magnetic grid is a moving component that contains two rows of permanent magnet arrays, one for relative displacement measurement and the other for the displacement interval code. The signal acquisition board is a fixed component that uses n groups of two-row Hall sensor arrays for continuous absolute displacement measurement. The principle of the sensor using the 2D magnetic field signal for the relative displacement measurement is analyzed, and a measurement method based on Hall sensor arrays for coding and absolute displacement detection over n cycles is proposed. Finally, a sensor prototype is fabricated and the experiments are performed. The experimental results show that the measurement resolution of the sensor is 5 µm, and the measurement accuracy is ±14.8 µm within the measurement range of 0-98.3 mm. The proposed sensor can realize continuous absolute displacement measurement over multiple cycles without cable binding.

Applying Enhanced Real-Time Monitoring and Counting Method for Effective Traffic Management in Tashkent.

This study describes an applied and enhanced real-time vehicle-counting system that is an integral part of intelligent transportation systems. The primary objective of this study was to develop an accurate and reliable real-time system for vehicle counting to mitigate traffic congestion in a designated area. The proposed system can identify and track objects inside the region of interest and count detected vehicles. To enhance the accuracy of the system, we used the You Only Look Once version 5 (YOLOv5) model for vehicle identification owing to its high performance and short computing time. Vehicle tracking and the number of vehicles acquired used the DeepSort algorithm with the Kalman filter and Mahalanobis distance as the main components of the algorithm and the proposed simulated loop technique, respectively. Empirical results were obtained using video images taken from a closed-circuit television (CCTV) camera on Tashkent roads and show that the counting system can produce 98.1% accuracy in 0.2408 s.

DEHA-Net: A Dual-Encoder-Based Hard Attention Network with an Adaptive ROI Mechanism for Lung Nodule Segmentation.

Measuring pulmonary nodules accurately can help the early diagnosis of lung cancer, which can increase the survival rate among patients. Numerous techniques for lung nodule segmentation have been developed; however, most of them either rely on the 3D volumetric region of interest (VOI) input by radiologists or use the 2D fixed region of interest (ROI) for all the slices of computed tomography (CT) scan. These methods only consider the presence of nodules within the given VOI, which limits the networks' ability to detect nodules outside the VOI and can also encompass unnecessary structures in the VOI, leading to potentially inaccurate segmentation. In this work, we propose a novel approach for 3D lung nodule segmentation that utilizes the 2D region of interest (ROI) inputted from a radiologist or computer-aided detection (CADe) system. Concretely, we developed a two-stage lung nodule segmentation technique. Firstly, we designed a dual-encoder-based hard attention network (DEHA-Net) in which the full axial slice of thoracic computed tomography (CT) scan, along with an ROI mask, were considered as input to segment the lung nodule in the given slice. The output of DEHA-Net, the segmentation mask of the lung nodule, was inputted to the adaptive region of interest (A-ROI) algorithm to automatically generate the ROI masks for the surrounding slices, which eliminated the need for any further inputs from radiologists. After extracting the segmentation along the axial axis, at the second stage, we further investigated the lung nodule along sagittal and coronal views by employing DEHA-Net. All the estimated masks were inputted into the consensus module to obtain the final volumetric segmentation of the nodule. The proposed scheme was rigorously evaluated on the lung image database consortium and image database resource initiative (LIDC/IDRI) dataset, and an extensive analysis of the results was performed. The quantitative analysis showed that the proposed method not only improved the existing state-of-the-art methods in terms of dice score but also showed significant robustness against different types, shapes, and dimensions of the lung nodules. The proposed framework achieved the average dice score, sensitivity, and positive predictive value of 87.91%, 90.84%, and 89.56%, respectively.