Deep Process Research Articles

17O Destruction Rate in Stars

In recent years, several laboratory studies of CNO cycle-related nuclear reactions have been carried out. Nevertheless, extant models of stellar nucleosynthesis still adopt CNO reaction rates reported in old compilations, such as NACRE or CF88. In order to update these rates, we performed new calculations based on a Monte Carlo R-Matrix analysis. In more detail, a method was developed that is based on the collection of all the available data, including recent low-energy measurements obtained by the LUNA collaboration in the reduced background environment of the INFN-LNGS underground laboratory, on R-Matrix cross-section calculations with the AZURE2 code and on uncertainty evaluations with a Monte Carlo analysis. As a first scientific benchmark case, the reactions 17O(p,γ)18F and 17O(p,α)14N were investigated. Among the different stellar scenarios they can influence, the 16O/17O abundance ratio in RGB and AGB stars is the one that can be directly confirmed from spectroscopic measurements. The aim is to reduce the nuclear physics uncertainties, thus providing a useful tool to constrain deep mixing processes eventually taking place in these stars. In this work, we present the procedure we followed to calculate the 17O(p,γ)18F and the 17O(p,α)14N reaction stellar rates and preliminary comparisons with similar rates reported in widely used nuclear physics libraries are discussed.

Control of slab tears and slab flat wedging on volcanism in the Alaska subduction zone

Multistage plate subduction plays a crucial role in magmatism; however, the mechanisms by which deep geodynamic processes govern volcanism in the Alaska subduction zone remain controversial. Using numerous travel-time data from several seismic arrays, we constructed high-resolution tomographic models to investigate the velocity structure of the Pacific Plate and Yakutat slab. Our tomographic results revealed high-velocity anomalies in the Pacific Plate and Yakutat slab, while the low-velocity areas within the Pacific Plate were identified as slab tears. We suggest that the Pacific Plate transitioned from oblique subduction along the Aleutian volcano chain to lower-angle subduction beneath the Pacific-Yakutat Plate interaction zone, forming two slab tears that enhance hot asthenosphere materials upwelling. The partial melting of the mantle wedge induced by Pacific slab dehydration and the concurrent upwelling of mantle materials jointly drove volcanism in the transition zone. Conversely, the flat subduction of the Yakutat slab into the mantle wedge overlying the Pacific slab effectively hindered the upwelling of hot hybrid materials, cooling the Pacific mantle wedge. These results offer a new perspective on the influence of slab dynamics on volcanic and magmatic processes in the region and represent an advancement in our understanding compared to previous studies, which did not resolve the tears within the slab or their geodynamic implications at this level of detail.

Anatomical landmark detection on bi-planar radiographs for predicting spinopelvic parameters.

Accurate landmark detection is essential for precise analysis of anatomical structures, supporting diagnosis, treatment planning, and monitoring in patients with spinal deformities. Conventional methods rely on laborious landmark identification by medical experts, which motivates automation. The proposed deep learning pipeline processes bi-planar radiographs to determine spinopelvic parameters and Cobb angles without manual supervision. The dataset used for training and evaluation consisted of 555 bi-planar radiographs from un-instrumented patients, which were manually annotated by medical professionals. The pipeline performed a pre-processing step to determine regions of interest, including the cervical spine, thoracolumbar spine, sacrum, and pelvis. For each ROI, a segmentation network was trained to identify vertebral bodies and pelvic landmarks. The U-Net architecture was trained on 455 bi-planar radiographs using binary cross-entropy loss. The post-processing algorithm determined spinal alignment and angular parameters based on the segmentation output. We evaluated the pipeline on a test set of 100 previously unseen bi-planar radiographs, using the mean absolute difference between annotated and predicted landmarks as the performance metric. The spinopelvic parameter predictions of the pipeline were compared to the measurements of two experienced medical professionals using intraclass correlation coefficient(ICC) and mean absolute deviation (MAD). The pipeline was able to successfully predict the Cobb angles in 61% of all test cases and achieved mean absolute differences of 3.3° (3.6°) and averaged ICC of 0.88. For thoracic kyphosis, lumbar lordosis, sagittal vertical axis, sacral slope, pelvic tilt, and pelvic incidence, the pipeline produced reasonable outputs in 69%, 58%, 86%, 85%, 84%, and 84% of the cases. The MAD was 5.6° (7.8°), 4.7° (4.3°), 2.8mm (3.0mm), 4.5° (7.2°), 1.8° (1.8°), and 5.3° (7.7°), while the ICC was measured at 0.69, 0.82, 0.99, 0.61, 0.96, and 0.70, respectively. Despite limitations in patients with severe pathologies and high BMI, the pipeline automatically predicted coronal and sagittal spinopelvic parameters, which has the potential to simplify clinical routines and large-scale retrospective data analysis.

Assessment of grain starch content and responses to CMS-S and CMS-C in high-starch maize hybrids

Background. Increasing the production of native and modified starch from maize requires raw materials with high starch content in grain. Materials and methods. An experimental panel of 780 simple high-starch maize hybrids produced with CMS-S and CMS-C lines underwent two-year testing. Starch content in the grain of the lines and their hybrids was assessed with IR spectrometry. Native starch content in the grain of hybrids with highest yields was measured at the All-Russian Research Institute of Starch and Starch-Containing Raw Materials Processing using the method proposed by L. P. Nosovskaya with coauthors. Responses to CMS were scored according to G. S. Galeev’s scale. Results. Grain starch content was found to vary from 58% to 72% DMB throughout the tested panel. IR spectrometry helped to identify 22 hybrids with high (72.03–72.67%) starch content, and 5 hybrids promising for deep grain processing, combining high protein (10.3–13.53%) and oil (3.77–5.03%) levels with high starch content (69.02–70.4%) in their grain. Native starch extraction using L. P. Nosovskaya’s method showed that grain starch content in the best 68 hybrids ranged from 70.03 to 71.95% DMB. The collection was ranked according to the main heterotic groups: 57 lines of Iowa Dent, 26 lines of Stiff Stalk Synthetic, and 28 lines of Lancaster. For CMS-S and CMS-C types, 33 and 6 maintainers, and 9 and 8 restorers were selected, respectively. The hybrids were distributed across the following FAO maturity groups for maize: FAO 200–299 (14 hybrids), FAO 300–399 (7), FAO 400–449 (21), and FAO 450–500 (29). Conclusion. Assessing agronomic and breeding prospects of the best 68 hybrids between high-starch maize lines and sterile testers proved their potential for producing native starch to at least 70–72% DMB. Five hybrids were identified as promising for yielding native starch (69.02–70.4% DMB), as well as protein (10.3–13.5% DMB) and oil (3.77–5.03% DMB) by-products during deep grain processing.

CO2-assisted Controllable Synthesis of PdNi Nanoalloys for Highly Selective Hydrogenation of Biomass-derived 5-Hydroxymethylfurfural.

The selective hydrogenation of 5-hydroxymethylfurfural (HMF) to 2,5-bishydroxymethyltetrahydrofuran (BHMTHF), a vital fuel precursor and solvent, is crucial for biomass refining. Herein, we report highly selective and stable PdNi nanoalloy catalysts for this deep hydrogenation process. A CO2-assisted green method was developed for the controllable synthesis of various bimetallic and monometallic catalysts. The PdNi/SBA-15 catalysts with various Pd/Ni ratios exhibited a volcano-like trend between BHMTHF yield and Pd/Ni ratio. Among all catalysts tested, Pd2Ni1/SBA-15 achieved the best performance, converting 99.0% of HMF to BHMTHF with 96.0% selectivity, surpassing previously reported catalysts. Additionally, the Pd2Ni1/SBA-15 catalyst maintained excellent stability even after five recycling runs. Catalyst characterizations (e.g., HAADF-STEM) and DFT calculations confirmed the successful formation of the alloy structure with electron transfer between Ni and Pd, which accounts for the remarkable performance and stability of the catalyst. This work paves the way for developing highly selective and stable alloy catalysts for biomass valorization.

AI-Driven Particulate Matter Estimation Using Urban CCTV: A Comparative Analysis Under Various Experimental Conditions

Despite the high public interest in particulate matter (PM), a key determinant for indoor and outdoor activities, the current PM information provided by monitoring stations (e.g., data per administrative district) is insufficient. This study employed the closed-circuit television (CCTV) cameras densely installed within a city to explore the spatial expansion of PM information. It conducted a comparative analysis of PM estimation effects under diverse experimental conditions based on AI image recognition. It also fills a gap by providing an optimal analysis framework that comprehensively considers the combination of variables, including the sun’s position, day and night settings, and the PM distribution per class. In the deep learning model structure and process comparison experiment, the hybrid DL-ML model using ResNet152 and XGBoost showed the highest predictive power. The classification model was better than the ResNet regression model, and the hybrid DL-ML model with the post-processed XGBoost was better than the single ResNet152 model regarding AI prediction of PM. All four experiments that excluded the nighttime, added the solar incidence angle variable, applied the distribution of PM per class, and removed the outlier removal algorithm showed high predictive power. In particular, the final experiment that satisfied all conditions, including the exclusion of nighttime, addition of solar incidence angle variable, and application of outlier removal algorithm, derived predictive values that are expected to be commercialized.

Efficient labeling of french mammogram reports with MammoBERT

Recent advances in deep learning and natural language processing (NLP) have broadened opportunities for automatic text processing in the medical field. However, the development of models for low-resource languages like French is challenged by limited datasets, often due to legal restrictions. Large-scale training of medical imaging models often requires extracting labels from radiology text reports. Current methods for report labeling primarily rely on sophisticated feature engineering based on medical domain knowledge or manual annotations by radiologists. These methods can be labor-intensive. In this work, we introduce a BERT-based approach for the efficient labeling of French mammogram image reports. Our method leverages both the expansive scale of existing rule-based systems and the precision of radiologist annotations. Our experimental results showcase the superiority of the proposed approach. It was initially fine-tuned on a limited dataset of radiologist annotations. Then, it underwent training on annotations generated by a rule-based labeler. Our findings reveal that our final model, MammoBERT, significantly outperforms the rule-based labeler while simultaneously reducing the necessity for radiologist annotations during training. This research not only advances the state of the art in medical image report labeling but also offers an efficient and effective solution for large-scale medical imaging model development.

Calibration of discrete meta-parameters of bamboo flour based on magnitude analysis and BP neural network.

In the research and development of technology and equipment for bamboo products deep processing, such as filling, drying, and medicinal use of bamboo flour (BF), the poor compaction and fluidity of BF materials entails the need for accurate discrete element model (DEM) and BF parameters to provide a reference for the simulation of BF processing operationsand the development of related equipment. The average particle size of the 5 types of BFs ranges from 0.136 mm to 0.293 mm, and the small particle size of BF particles causes to the number of BF particles in bamboo processing equipment to reach tens of millions or even billions. When conventional methods are used for simulation, ordinary computers cannot provide the required computing power. To address the aforementioned challenges, this paper proposes a calibration method for the discrete element contact parameters of BFs based on dimensional analysis and a back propagation (BP) neural network. Using particle scaling theory and dimensional analysis methods, the average particle size of the BF was increased to 1 mm, and the main discrete element contact parameters of the five types of BF to be tested were used as input layers. The injection method and sidewall collapse method were used to obtain the angle of repose (AR) as the output layer. Fifty groups were randomly selected using MATLAB for EDEM simulation, and the simulation results were trained using the BP neural network algorithm; an ideal neural network model was obtained, the discrete element parameters of different BFs were predicted, and physical experiments were performed to verify two types of AR and mold hole compression under calibrated parameters. The relative error between the simulated AR obtained through calibration parameters and the physical experimental values is less than 2.3%. Through BF parameter validity verification, the simulated maximum compression displacement and compression ratio after stabilization were 34.81 mm and 0.477, which were close to the actual experimental results of 34.77 mm and 0.461, respectively, verifying the accuracy of the neural network prediction model. The research results provide a reference for the simulation of BF processing operations and the development of related equipment.

Specific features of production and use of transportation biofuels in selected developing countries (part II)

Aim. To identify specific features of production and use of transportation biofuels in selected developing countries, as well as to assess the potential of bioethanol industry development in the Russian Federation (RF).Objectives. To assess the economic efficiency of bioethanol production in Russia; to reveal the potential of applying foreign experience in the development of bioethanol industry in our country.Methods. In the process of the research the methods of system analysis, comparative analysis, method of expert evaluations, mathematical, statistical methods were applied.Results. The actual estimation of economic efficiency of bioethanol production in Russia was obtained. The cost of a unit of bioethanol, which is equivalent to a liter of gasoline in terms of energy content, is 5-36% higher than the cost of a liter of gasoline (depending on the production technology). At the same time, state support for the production and use of bioethanol in our country is reasonable. This will allow to create additional demand for grain crops and reduce carbon dioxide emissions from motor transport. The production and use of bioethanol will form an additional supply in the gasoline market, will give an opportunity to support projects on deep processing of grain and wood chemistry.Conclusions. The experience of Brazil, China and Indonesia in the development of bioethanol industry is relevant for Russia. At the same time, in order to minimize possible risks in the field of food security at the initial stage of bioethanol industry development, low-quality grain stocks should be used as feedstock and the production of second-generation bioethanol should be stimulated.

Detection of Hidden Low-Frequency Earthquakes in Southern Vancouver Island with Deep Learning

Low-frequency earthquakes (LFEs) are small-magnitude earthquakes that are depleted in high-frequency content relative to traditional earthquakes of the same magnitude. These events occur in conjunction with slow slip events (SSEs) and can be used to infer the space and time evolution of SSEs. However, because LFEs have weak signals, and the methods used to identify them are computationally expensive, LFEs are not routinely cataloged in most places. Here, we develop a deep-learning model that learns from an existing LFE catalog to detect LFEs in 14 years of continuous waveform data in southern Vancouver Island. The result shows significant increases in detection rates at individual stations. We associate the detections and locate them using a grid search approach in a 3D regional velocity model, resulting in over 1 million LFEs during the performing period. Our resulting catalog is consistent with a widely used tremor catalog during periods of large-magnitude SSEs. However, there are time periods where it registers far more LFEs than the tremor catalog. We highlight a 16-day period in May 2010, when our model detects nearly 3,000 LFEs, whereas the tremor catalog contains only one tremor detection in the same region. This suggests the possibility of hidden small-magnitude SSEs that are undetected by current approaches. Our approach improves the temporal and spatial resolution of the LFE activities and provides new opportunities to understand deep subduction zone processes in this region.

Review in deep processing of whey

Review in deep processing of whey

PPDNN-CRP: privacy-preserving deep neural network processing for credit risk prediction in cloud: a homomorphic encryption-based approach

This study proposes a Privacy-Preserving Deep Neural Network for Credit Risk Prediction (PPDNN-CRP) framework that leverages homomorphic encryption (HE) to ensure data privacy throughout the credit risk prediction process. The PPDNN-CRP framework employs the Paillier homomorphic encryption scheme to secure sensitive loan application data during both the training and inference phases. Implemented using TensorFlow for deep neural network operations and TenSEAL for HE, the framework uses the Kaggle loan dataset to evaluate its performance. The results show that PPDNN-CRP achieved an accuracy of 80.48%, demonstrating competitive performance compared to Privacy-Preserving Logistic Regression (PPLR) at 77.23% and a slight decrease from the non-private DNN-CRP model at 86.18%. The model exhibited strong metrics with a precision of 0.84, recall of 0.91, F1-score of 0.87, and an AUC of 0.74. Security analysis confirms that PPDNN-CRP effectively defends against various privacy attacks, including poisoning, evasion, membership inference, model inversion, and model extraction, through robust encryption techniques and privacy measures. This framework offers a promising approach for achieving high-quality credit risk prediction while maintaining data privacy and complying with legal and ethical standards.

Lithospheric deformation and corresponding deep geodynamic process of the SE Tibetan Plateau

Lithospheric deformation and corresponding deep geodynamic process of the SE Tibetan Plateau

Deep Learning Methods to Mitigate Human-Factor-Related Accidents in Maritime Transport

Artificial intelligence aims to be the solution to multiple engineering problems by trying to emulate the human learning process. In this sense, maritime transport standards have clearly evolved, which are based on two principal pillars: the International Convention for the Safety of Life at Sea Convention (SOLAS) and the International Convention for the Prevention of Pollution from Ships (MARPOL). Based on a formal safety assessment research process, these pillars try to solve most of the maritime transport accidents, which, in their final steps, are associated with human factors. In this research, an original methodology employing a deep learning process for image recognition during mooring line operation, a dangerous process on ships, is developed. The main results indicate that the proposed method is an excellent tool for advising ship officers on watch and, consequently, provides a new way to prevent human factors onboard from causing accidents, which in the future must be considered in international standards.

AI-Generated Spam Review Detection Framework with Deep Learning Algorithms and Natural Language Processing

Spam reviews pose a significant challenge to the integrity of online platforms, misleading consumers and undermining the credibility of genuine feedback. This paper introduces an innovative AI-generated spam review detection framework that leverages Deep Learning algorithms and Natural Language Processing (NLP) techniques to identify and mitigate spam reviews effectively. Our framework utilizes multiple Deep Learning models, including Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, Gated Recurrent Unit (GRU), and Bidirectional LSTM (BiLSTM), to capture intricate patterns in textual data. The system processes and analyzes large volumes of review content to detect deceptive patterns by utilizing advanced NLP and text embedding techniques such as One-Hot Encoding, Word2Vec, and Term Frequency-Inverse Document Frequency (TF-IDF). By combining three embedding techniques with four Deep Learning algorithms, a total of twelve exhaustive experiments were conducted to detect AI-generated spam reviews. The experimental results demonstrate that our approach outperforms the traditional machine learning models, offering a robust solution for ensuring the authenticity of online reviews. Among the models evaluated, those employing Word2Vec embeddings, particularly the BiLSTM_Word2Vec model, exhibited the strongest performance. The BiLSTM model with Word2Vec achieved the highest performance, with an exceptional accuracy of 98.46%, a precision of 0.98, a recall of 0.97, and an F1-score of 0.98, reflecting a near-perfect balance between precision and recall. Its high F2-score (0.9810) and F0.5-score (0.9857) further highlight its effectiveness in accurately detecting AI-generated spam while minimizing false positives, making it the most reliable option for this task. Similarly, the Word2Vec-based LSTM model also performed exceptionally well, with an accuracy of 97.58%, a precision of 0.97, a recall of 0.96, and an F1-score of 0.97. The CNN model with Word2Vec similarly delivered strong results, achieving an accuracy of 97.61%, a precision of 0.97, a recall of 0.96, and an F1-score of 0.97. This study is unique in its focus on detecting spam reviews specifically generated by AI-based tools rather than solely detecting spam reviews or AI-generated text. This research contributes to the field of spam detection by offering a scalable, efficient, and accurate framework that can be integrated into various online platforms, enhancing user trust and the decision-making processes.

Does deep processing protect against mind wandering and other lapses of attention during learning?

ABSTRACT Three experiments examined if deep processing would aid in reducing off-task thoughts during learning. In each experiment, participants learned words or pairs of words under deep or shallow conditions. During learning, participants were periodically presented with thought-probes to examine if they were experiencing off-task thoughts (mind wandering, external distraction, mind blanking). Levels of processing was manipulated both within (Experiment 1) and between subjects (Experiments 2 and 3) using structural, rhyme, and semantic judgements and testing memory with cued-recall (Experiments 1 and 2) or recognition (Experiment 3). All three experiments demonstrated a levels of processing effect on memory with better performance in deep compared to shallow conditions. Importantly, in all three experiments rates of off-task thinking (and mind wandering more specifically) were the same across conditions and Bayes factors suggested moderate evidence in favour of the null hypothesis. The results suggest that deep processing does not necessarily protect against mind wandering and other lapses of attention.

Image-Based Auto-Focus Microscope System with Visual Servo Control for Micro-Stereolithography.

Micro-stereolithography (μSL) is an advanced additive manufacturing technique that enables the fabrication of highly precise microstructures with fine feature resolution. One of the primary challenges in μSL is achieving and maintaining precise focus throughout the fabrication process. For the successful application of μSL, it is essential to maintain the sample surface within a focal depth of several microns. Despite the growing interest in auto-focus devices, limited attention has been directed towards auto-focus systems in image-based auto-focus microscope systems for precision μSL. To address this challenge, we propose an image-based auto-focus microscope system incorporating visual servo control. In the optical design, a transflective beam splitter is employed, allowing the laser beam to pass through for fabrication while reflecting the focused beam on the sample surface to the microscope and camera. Utilizing captured spot images and the Foucault knife-edge test, a deep learning-based laser spot image processing algorithm is developed to determine the focus position based on spot size and the number of spot pixels on both sides. Experimental results demonstrate that the proposed auto-focus system effectively determines the relative position of the focal point using the laser spot image and achieves auto-focusing through visual servo control.

Depositional and diagenetic controls on the reservoir quality of marginal marine sandstones: An example from the Early Devonian subbat member, jauf formation, northwest Saudi Arabia

The reservoir quality of marginal marine sandstones can vary significantly across different depositional lithofacies. This variation is primarily due to differences in depositional processes and subsequent diagenetic alterations occurring during various stages of burial. To address this issue, this study focuses on the Early Devonian Subbat sandstones of the Jauf Formation, which serve as an important analogue for subsurface gas reservoirs and have potential for CO2 sequestration. An integrated approach was employed, including field observations, petrography, geochemical analyses, and petrophysical analyses, to investigate the impact of depositional and diagenetic processes on reservoir quality. The findings indicate that the sandstones range from poorly to well sorted, very fine to medium-grained arkosic and quartz arenites that were deposited in a range of environments including shoreface-to-offshore transition zones, wave-dominated deltas, fluvial and tidal estuarine channels. The sandstones experienced both shallow and deep burial diagenesis, while key diagenetic events affecting reservoir quality include compaction, carbonate cementation, the formation of authigenic clays (kaolinite, illite, and chlorite), quartz overgrowth, and the dissolution of unstable feldspar and mica grains. Authigenic pore-filling cements and grain-coating clays significantly influence reservoir quality. The effect of quartz cementation was minimal where illite and/or chlorite coatings are well-developed. Mechanical compaction is more pronounced in the shoreface-to-offshore transition compared to the estuarine channels likely due to the higher detrital matrix content (21.7–35%, avg. 26.7%). The well-sorted fluvial to estuarine channel sandstones exhibit the best reservoir quality, with porosity ranging from 15.9% to 25.5% (average 20.4%) and permeability between 433.8 mD and 1261.2 mD (average 720 mD). This high reservoir quality is attributed to the low detrital matrix content (0–1%, average 0.2%) and limited pore-filling cements. In comparison, the moderately sorted, wave-dominated delta sandstones have lower porosity (5.3–11.2%, average 8.2%) and permeability (17.8–80 mD, average 42.8 mD), despite a relatively higher detrital matrix content (0–11.5%, average 2.8%). The poorly sorted shoreface-to-offshore transition sandstones exhibit the lowest reservoir quality, with porosity ranging from 1% to 2% (average 1.5%) and permeability between 0.3 and 1.1 mD (average 0.7 mD). These variations in reservoir quality among different depositional facies are primarily linked to depositional factors (sorting and detrital clay matrix) and shallow to deep burial diagenetic processes (compaction and cementation). A comprehensive understanding of these processes can enhance reservoir quality prediction in the Jauf Formation and similar reservoirs elsewhere.

Nonlinear optical cyanurate crystals

The exploration of ultraviolet (UV) nonlinear optical (NLO) crystals is of great significance in various fields, such as optical communication, medical diagnosis, deep space exploration and optical digital signal processing. Accordingly, cyanurates attracted widespread attention due to their strong second-harmonic-generation (SHG) effects, large birefringence for phase-matching, short UV cutoffs, high solar blind transmittances and easy crystal preparations. This article reviews the optical properties of so far reported cyanurate NLO crystals, describes the preparation approaches of cyanurates, compares their structures to further analyze the source of their large SHG effect, and provides some insights for designing and synthesizing new UV cyanurates in the future.

An EfficientNet integrated ResNet deep network and explainable AI for breast lesion classification from ultrasound images

AbstractBreast cancer is one of the major causes of deaths in women. However, the early diagnosis is important for screening and control the mortality rate. Thus for the diagnosis of breast cancer at the early stage, a computer‐aided diagnosis system is highly required. Ultrasound is an important examination technique for breast cancer diagnosis due to its low cost. Recently, many learning‐based techniques have been introduced to classify breast cancer using breast ultrasound imaging dataset (BUSI) datasets; however, the manual handling is not an easy process and time consuming. The authors propose an EfficientNet‐integrated ResNet deep network and XAI‐based framework for accurately classifying breast cancer (malignant and benign). In the initial step, data augmentation is performed to increase the number of training samples. For this purpose, three‐pixel flip mathematical equations are introduced: horizontal, vertical, and 90°. Later, two pre‐trained deep learning models were employed, skipped some layers, and fine‐tuned. Both fine‐tuned models are later trained using a deep transfer learning process and extracted features from the deeper layer. Explainable artificial intelligence‐based analysed the performance of trained models. After that, a new feature selection technique is proposed based on the cuckoo search algorithm called cuckoo search controlled standard error mean. This technique selects the best features and fuses using a new parallel zero‐padding maximum correlated coefficient features. In the end, the selection algorithm is applied again to the fused feature vector and classified using machine learning algorithms. The experimental process of the proposed framework is conducted on a publicly available BUSI and obtained 98.4% and 98% accuracy in two different experiments. Comparing the proposed framework is also conducted with recent techniques and shows improved accuracy. In addition, the proposed framework was executed less than the original deep learning models.