Region-Aware Fashion Contrastive Learning for Unified Attribute Recognition and Composed Retrieval

Clothing attribute recognition has become an essential technology,which enables users to automatically identify the characteristics of clothes and search for clothing images with similar attributes.However,existing methods cannot recognize newly added attributes and may fail to capture region-level visual features.To address the aforementioned issues,a region-aware fashion contrastive language-image pre-training(RaF-CLIP)model was proposed.This model aligned cropped and segmented images with category and multiple fine-grained attribute texts,achieving the matching of fashion region and corresponding texts through contrastive learning.Clothing retrieval found suitable clothing based on the user-specified clothing categories and attributes,and to further improve the accuracy of retrieval,an attribute-guided composed network(AGCN)as an additional component on RaF-CLIP was introduced,specifically designed for composed image retrieval.This task aimed to modify the reference image based on textual expressions to retrieve the expected target.By adopting a transformer-based bidirectional attention and gating mechanism,it realized the fusion and selection of image features and attribute text features.Experimental results show that the proposed model achieves a mean precision of 0.6633 for attribute recognition tasks and a recall@10(recall@k is defined as the percentage of correct samples appearing in the top k retrieval results)of 39.18 for composed image retrieval task,satisfying user needs for freely searching for clothing through images and texts.

Differentiating consistency of meningioma based on grey-level histogram analysis of T2WI

Objective To observe the value of grey-level histogram analysis based on T2WI for differentiating consistency of meningioma.Methods Data of 109 patients with meningioma were retrospectively analyzed.The patients were divided into hard group(n=71)and soft group(n=38)according to the consistency of tumors.Tumor ROI was outlined on axial T2WI showing the largest tumor section,gray levels were extracted and histogram analysis was performed.The value of each histogram parameter were compared between groups.Then receiver operating characteristic curve was drawn,and the area under the curve(AUC)was calculated to evaluate the efficiency for differentiating soft and hard meningioma.Results P 1,P 10,P 50,P 90,P 99 and the mean grey levels on T2WI in soft group were all higher than those in hard group(all P<0.05),while the variance,the kurtosis and the skewness were not significantly different between groups(all P>0.05).The differentiating efficiency of P 1,P 10,P 50,P 90,P 99 and the mean grey levels on T2WI were all fine,with AUC of 0.774 to 0.833,and no significant difference was found(all P>0.05).Conclusion Parameters of grey-level histogram analysis such as P 1,P 10,P 50,P 90,P 99 and the mean values based on T2WI were all valuable for differentiating soft and hard meningioma.

Research progresses of artificial intelligence in MRI of lumbar degenerative diseases

The incidence of lumbar degenerative diseases is increasing year by year,and MRI is often used in clinical diagnosis.In recent years,artificial intelligence(AI)has rapidly developed in medical field and can be used for image segmentation and auxiliary diagnosis of lumbar degenerative diseases.The research progresses of AI in MRI of lumbar degenerative diseases were reviewed in this article.

Long wavelength infrared metalens fabricated by photolithography

Metasurfaces in the long wave infrared(LWIR)spectrum hold great potential for applications in ther-mal imaging,atmospheric remote sensing,and target identification,among others.In this study,we designed and experimentally demonstrated a 4 mm size,all-silicon metasurface metalens with large depth of focus opera-tional across a broadband range from 9µm to 11.5µm.The experimental results confirm effective focusing and imaging capabilities of the metalens in LWIR region,thus paving the way for practical LWIR applications of met-alens technology.

Advancing automated pupillometry:a practical deep learning model utilizing infrared pupil images

AIM:To establish pupil diameter measurement algorithms based on infrared images that can be used in real-world clinical settings.METHODS:A total of 188 patients from outpatient clinic at He Eye Specialist Shenyang Hospital from Spetember to December 2022 were included,and 13470 infrared pupil images were collected for the study.All infrared images for pupil segmentation were labeled using the Labelme software.The computation of pupil diameter is divided into four steps:image pre-processing,pupil identification and localization,pupil segmentation,and diameter calculation.Two major models are used in the computation process:the modified YoloV3 and Deeplabv 3+models,which must be trained beforehand.RESULTS:The test dataset included 1348 infrared pupil images.On the test dataset,the modified YoloV3 model had a detection rate of 99.98% and an average precision(AP)of 0.80 for pupils.The DeeplabV3+model achieved a background intersection over union(IOU)of 99.23%,a pupil IOU of 93.81%,and a mean IOU of 96.52%.The pupil diameters in the test dataset ranged from 20 to 56 pixels,with a mean of 36.06±6.85 pixels.The absolute error in pupil diameters between predicted and actual values ranged from 0 to 7 pixels,with a mean absolute error(MAE)of 1.06±0.96 pixels.CONCLUSION:This study successfully demonstrates a robust infrared image-based pupil diameter measurement algorithm,proven to be highly accurate and reliable for clinical application.

Correlations of transcranial sonography of substantia nigra characteristics with MRI iron deposition on substantia nigra in patients with Parkinson disease

Objective To explore the correlations of transcranial sonography of substantia nigra(SN-TCS)characteristics with MRI iron deposition on substantia nigra in patients with Parkinson disease(PD).Methods Data of SN-TCS and craniocerebral MRI in 120 PD patients were retrospectively analyzed.The patients were divided into iron deposition positive group(positive group,n=46)and iron deposition negative group(negative group,n=74)according to quantitative susceptibility mapping(QSM)value.Then parameters of SN-TCS and MRI were compared between groups(both P<0.05),and correlation analysis were also performed.Results The proportion of high echo positive,strong echo area and QSM value of substantia nigra,as well as of hyper-substantia nigra area/midbrain area(S/M)in positive group were all higher than those in negative group(all P<0.001).No significant difference of midbrain area was found between groups(P>0.05).Strong echo area of substantia nigra and S/M based on SN-TCS were both low-medium positively correlated with substantia nigra QSM value showed on MRI(r=0.497,0.529,both P<0.001).Conclusion SN-TCS characteristics of PD patients were correlated with MRI iron deposition on substantia nigra,among which strong echo area and S/M were valuable for evaluating iron deposition on substantia nigra.

E-Cervix imaging displaying cervical elasticity characteristics of healthy adult nulliparous women at different age groups and different menstrual cycle stages

Objective To observe the cervical elasticity of healthy adult nulliparous women at different age groups and different stages of menstrual cycle with E-Cervix imaging technology.Methods A total of 218 healthy adult nulliparous women who underwent transvaginal ultrasound examination for routine physical examination were retrospectively enrolled,including 103 in follicular phase,78 in ovulation phase and 37 in luteal phase.Cervical canal length(CL)and E-Cervix elasticity parameters were compared among different age groups and different stages of menstrual cycle,including elasticity contrast index(ECI),hardness ratio(HR),cervical internal and external orifice strain values(IOS and EOS)and IOS/EOS ratio.Results No significant difference of CL nor cervical elasticity parameters was detected among healthy adult nulliparous women at different age groups(all P>0.05).There were significant differences of ECI,HR and IOS among different menstrual cycle stages(all P<0.05),among which women in follicular phase had higher ECI and IOS but lower HR than those in luteal phase(all P<0.05).Conclusion No significant difference of cervical elasticity existed among healthy adult nulliparous women at different age groups.Meanwhile,cervical elasticity of healthy adult nulliparous women changed during menstrual cycle,in follicular phase had higher ECI and IOS but lower HR than in luteal phase.

Preliminary clinical application of total free-breathing cardiac MR examination

Objective To observe the clinical application value of total free-breathing cardiac MR(CMR)examination preliminarily.Methods Two patients who underwent CMR scanning under free-breathing state,including cine,motion correction T1 and T2 mapping,blood flow imaging,and late gadolinium enhancement scanning were retrospectively enrolled,and the qualities of the above images were evaluated and compared with that of conventional CMR images under breath-holding state.Results No significant difference of imaging quality was found between total free-breathing and conventional breath-holding CMR.The differences of left ventricular ejection fraction,cardiac output,left ventricular end-diastolic volume index and left ventricular mass measured based on CMR images under different breath conditions were limited.Conclusion Total free-breathing CMR was feasible in clinical practice,which could provide"one-stop"evaluation of cardiac structure,function and myocardial histological characteristics,hence having promising clinical prospects.

In-situ thermal Raman mapping and stress analysis of CNT/CF/epoxy interfaces

A study of the interfacial behavior and internal thermal stress distribution in fiber-reinforced composites is essential to assess their performance and reliability.CNT/carbon fiber(CF)hybrid fibers were constructed using electrophoretic deposition.The interfacial properties of CF/epoxy and CNT/CF/epoxy composites were statistically investigated and compared using in-situ thermal Raman mapping by dispersing CNTs as a Raman sensing medium(CNT_(R))in a resin.The associated local thermal stress changes can be simulated by capturing the G'band position distribution of CNT_(R) in the epoxy at different temperatures.It was found that the G'band shifted to lower positions with increasing temperature,reaching a maximum difference of 2.43 cm^(−1) at 100℃.The interfacial bonding between CNT/CF and the matrix and the stress distribution and changes during heat treatment(20-100℃)were investig-ated in detail.This work is important for studying thermal stress in fiber-reinforced composites by in-situ thermal Raman mapping technology.

Intravoxel incoherent motion and dynamic contrast-enhanced MRI for assessing abnormalities of brucellosis spondylitis without conventional MRI changes

Objective To observe the value of intravoxel incoherent motion(IVIM)and dynamic contrast-enhanced MRI(DCE-MRI)for assessing abnormalities of brucellosis spondylitis(BS)without conventional MRI changes.Methods Data of 36 brucellosis patients with definite spinal lesions displayed on conventional MRI(BS 1 group),14 cases without brucellosis infection nor abnormal spinal signals on MRI(control group)and 36 brucellosis patients without definite spinal lesions on conventional MRI(BS 2 group)were retrospectively analyzed.The values of IVIM parameters,including perfusion fraction(f),pure water diffusion coefficient(D)and pseudo-diffusion coefficient(D*),also of DCE-MRI parameters,including time-intensity curve(TIC)type,volume transport constant(K trans),the rate constant(K ep)and volume fraction of extravascular extracellular space per unit tissue volume(V e)were compared among groups.Univariate and multivariate logistic regression were used to screen independent factors for discriminating BS 1 and BS 2.Receiver operating characteristic curves were drawn,and the areas under the curve(AUC)were calculated to evaluate the efficiency of the above parameters for discriminating BS 1 and BS 2.Results Among IVIM parameters,compared with control group,D*values decreased but D values increased in BS 1 group,while D*values increased in BS 2 group(all adjusted P<0.05).Compared with BS 2 group,BS 1 group had higher values of f and D and lower D*(all adjusted P<0.05).In BS 1 group,the TIC types were predominantly typeⅠ(23/36,63.89%),which were wholly or predominantly typeⅢin BS 2 group and control group,and of the former was significantly different with latter 2(both adjusted P<0.05).Compared with control group,K trans increased progressively in both BS 1 and BS 2 groups(both adjusted P<0.05).BS 1 group had lower K ep and higher V e than BS 2 and control groups(all adjusted P<0.05).Among univariate logistic regression models,the model including only f had lower capability for discriminating BS 1 and BS 2(AUC=0.759)than those including D,K trans and V e(AUC=0.951,0.833,0.894,all P<0.05).No significant different was found among multivariate logistic regression model including f and D,model including K trans and V e nor model including all above parameters(all P>0.05).Conclusion Both IVIM and DCE-MRI could be used to evaluate BS abnormality without conventional MRI changes.

Deep learning-based recognition of stained tongue coating images

Objective To build a dataset encompassing a large number of stained tongue coating images and process it using deep learning to automatically recognize stained tongue coating images.Methods A total of 1001 images of stained tongue coating from healthy students at Hunan University of Chinese Medicine and 1007 images of pathological(non-stained)tongue coat-ing from hospitalized patients at The First Hospital of Hunan University of Chinese Medicine withlungcancer;diabetes;andhypertensionwerecollected.Thetongueimageswererandomi-zed into the training;validation;and testing datasets in a 7:2:1 ratio.A deep learning model was constructed using the ResNet50 for recognizing stained tongue coating in the training and validation datasets.The training period was 90 epochs.The model’s performance was evaluated by its accuracy;loss curve;recall;F1 score;confusion matrix;receiver operating characteristic(ROC)curve;and precision-recall(PR)curve in the tasks of predicting stained tongue coating images in the testing dataset.The accuracy of the deep learning model was compared with that of attending physicians of traditional Chinese medicine(TCM).Results The training results showed that after 90 epochs;the model presented an excellent classification performance.The loss curve and accuracy were stable;showing no signs of overfitting.The model achieved an accuracy;recall;and F1 score of 92%;91%;and 92%;re-spectively.The confusion matrix revealed an accuracy of 92%for the model and 69%for TCM practitioners.The areas under the ROC and PR curves were 0.97 and 0.95;respectively.Conclusion The deep learning model constructed using ResNet50 can effectively recognize stained coating images with greater accuracy than visual inspection of TCM practitioners.This model has the potential to assist doctors in identifying false tongue coating and prevent-ing misdiagnosis.

Long TR three-dimensional inversion recovery sequence with real reconstruction(3D real IR)for quantifying inner ear endolymphatic hydrops

Objective To observe the value of long TR three-dimensional inversion recovery sequence with real reconstruction(3D real IR)for quantifying inner ear endolymphatic hydrops(EH).Methods Totally 46 Ménière's disease(MD)patients and 21 healthy volunteers were prospectively enrolled.MR scanning for inner ear based on 3D real IR and 3D fluid attenuated inversion recovery(3D FLAIR)sequence 4—6 h after administration of contrast agents were performed.The imaging qualities were scored and compared between groups.The endolymphatic space area and the membranous labyrinth area of cochlea and vestibule,as well as endolymph/membranous labyrinth area percentage were calculated,the present or not of EH and the grade of EH were evaluated.EH inner ears of MD patients were enrolled in EH group,while inner ears of healthy volunteers were taken as controls(control group).The endolymphatic space area,membranous labyrinth area and endolymph/membranous labyrinth area percentage of cochlea and vestibule were compared between groups.The receiver operating characteristic(ROC)curve was drawn to calculate the diagnostic efficacy of the above indexes.Results Cochlear and/or vestibular EH were detected in 56 ears,including cochlear EH in 52 ears and vestibular EH in 45 ears among 46 MD patients(EH group),but not in 42 ears in control group.The subjective quality scores of 3D real IR images were higher than those of 3D-FLAIR(both P<0.05).Quantitative analysis based on 3D real IR images revealed that,compared with control group,significantly larger endolymph areas and endolymph/membranous labyrinth area percentages in both cochlea and vestibule were found in EH group(all P<0.001).The area under the curve(AUC)of cochlear or vestibular endolymph/membranous labyrinth area percentage for identifying inner ear EH was 0.999 and 0.985,respectively.Taken 13.64%and 24.13%as the critical value of cochlear or vestibular endolymph,the specificity was 100%and 92.86%,respectively,and the sensitivity was 96.43%and 96.43%,respectively.Conclusion MR long TR 3D real IR was helpful to quantifying inner ear EH.

Research progresses of resting-state functional MRI in mild cognitive impairment and Alzheimer's disease

Alzheimer's disease(AD)is a prevalent neurodegenerative disease characterized by cognitive decline in the early stage.Mild cognitive impairment(MCI)is considered as an intermediate stage between normal aging and AD.In recent years,studies related to resting-state functional MRI(rs-fMRI)indicated that the occurrence and development process of MCI and AD might be closely linked to spontaneous brain activity and alterations in functional connectivity among brain regions,and rs-fMRI could provide important reference for specific diagnosis and early treatment of MCI and AD.The research progresses of rs-fMRI for MCI and AD were reviewed in this article.

Research on the LA-UMamba Model for Asymmetric Modules with Added Auxiliary Information

Deep learning techniques are revolutionizing the developmentof medical image segmentation.With the advancement of Transformer models,especially ViT and Swin-Transformer,which enhances the remote-dependent modeling capability of the model through the self-attention mechanism,better segmentation performance can be achieve.Moreover,the high computational cost of Transformer has motivated researchers to explore more efficient models,such as the Mamba model based on state-space modeling(SSM),and for the field of medical segmentation,reducing the number of model parameters is also necessary.In this study,a novel asymmetric model called LA-UMamba was proposed,which integrates visual Mamba module to efficiently capture complex visual features and remote dependencies.The classical design of U-Net was adopted in the upsampling phase to help reduce the number of references and recover more details.To mitigate the information loss problem,an auxiliary U-Net downsampling layer was designed to focus on sizing without extracting features,thus enhancing the protection of input information while maintaining the efficiency of the model.The experiments were conducted on the ACDC MRI cardiac segmentation dataset,and the results showed that the proposed LA-UMamba achieves proved performance compared to the baseline model in several evaluation metrics,such as IoU,Accuracy,Precision,HD and ASD,which improved that the model is successful in optimizing the detail processing and reducing the complexity of the model,providing a new perspective for further optimization of medical image segmentation techniques.

Isotropic volumetric MRI for displaying cranial perineural spread of cranial nerve in nasopharyngeal carcinoma

Objective To observe the value of isotropic volumetric MRI for displaying perineural spread(PNS)of cranial nerve(CN)in nasopharyngeal carcinoma.Methods Eighty-seven patients with pathologically proven nasopharyngeal carcinoma were prospectively enrolled.MR scanning,including three-dimensional liver acquisition with volume acceleration-flexible(3D LAVA_Flex)image,T2WI with fat suppression(T2WI-FS),T1WI,contrast enhancement(CE)T1WI-FS of nasopharynx and neck region were performed.The displaying rates of CN PNS were evaluated and compared between 3D LAVA_Flex and T2WI-FS,T1WI,CE-T1WI-FS at patient level,CN group level and neural level,respectively.Results The displaying rate of CN PNS in all 87 nasopharyngeal carcinoma patients by 3D LAVA_Flex sequence was 49.43%(43/87),higher than that of conventional MRI(30/87,34.48%,P=0.001).Among 59 patients with advanced nasopharyngeal carcinoma diagnosed with conventional sequences,the displaying rate of CN PNS was 71.19%(42/59)by 3D LAVA-Flex sequence,higher than that of conventional MRI(30/59,50.85%,P=0.001).At both patient level and posterior CN level,significant differences of the displaying rate of CN PNS were found between 3D LAVA-Flex sequence and T2WI-FS,T1WI,CE-T1WI-FS,while at CN level,the displaying rates of mandibular nerve PNS,CNⅨ—ⅪPNS in jugular foramen(P<0.05)and CNⅨ—ⅫPNS in carotid space of 3D LAVA_Flex sequence were all significantly higher than that of T2WI-FS,T1WI and CE-T1WI-FS(all P<0.05),of PNS of CNⅢ—Ⅴin cavernous sinus were higher than that of T2WI-FS(P<0.05),while of PNS of hypoglossal nerve were significantly higher than that of T2WI-FS and T1WI(both P<0.05).Conclusion 3D LAVA_Flex sequence could be used to effectively display CN PNS of nasopharyngeal carcinoma.

Preoperative CT radiomics models for predicting composition of in vivo urinary calculi

Objective To observe the value of preoperative CT radiomics models for predicting composition of in vivo urinary calculi.Methods Totally 543 urolithiasis patients were retrospectively enrolled and divided into calcium oxalate monohydrate stone group(group A,n=373),anhydrous uric acid stone group(group B,n=86),carbonate apatite group(group C,n=30),ammonium urate stone group(group D,n=28)and ammonium magnesium phosphate hexahydrate stone group(group E,n=26)according to the composition of calculi,also divided into training set and test set at the ratio of 7∶3.Radiomics features were extracted and screened based on plain CT images of urinary system.Five binary task models(model A—E corresponding to group A—E)and a quinary task model were constructed using least absolute shrinkage and selection operator algorithm for predicting the composition of calculi in vivo.Then receiver operating characteristic curves were drawn,and the area under the curves(AUC)were calculated to evaluate the predictive efficacy of binary task models,while the accuracy,precision,recall and F1 score were used to evaluate the predictive efficacy of the quinary task model.Results All binary task models had good efficacy for predicting the composition of urinary calculi in vivo,with AUC of 0.860—0.948 in training set and of 0.856—0.933 in test set.The accuracy,precision,recall and F1 score of the quinary task model for predicting the composition of in vivo urinary calculi was 82.25%,83.79%,46.23%and 0.596 in training set,respectively,while was 80.63%,75.26%,43.48%and 0.551 in test set,respectively.Conclusion Binary task radiomics models based on preoperative plain CT had good efficacy for predicting the composition of in vivo urinary calculi,while the quinary task radiomics model had high accuracy but relatively poor stability.

Oriented Bounding Box Object Detection Model Based on Improved YOLOv8

In the study of oriented bounding boxes(OBB)object detection in high-resolution remote sensing images,the problem of missed and wrong detection of small targets occurs because the targets are too small and have different orientations.Existing OBB object detection for remote sensing images,although making good progress,mainly focuses on directional modeling,while less consideration is given to the size of the object as well as the problem of missed detection.In this study,a method based on improved YOLOv8 was proposed for detecting oriented objects in remote sensing images,which can improve the detection precision of oriented objects in remote sensing images.Firstly,the ResCBAMG module was innovatively designed,which could better extract channel and spatial correlation information.Secondly,the innovative top-down feature fusion layer network structure was proposed in conjunction with the Efficient Channel Attention(ECA)attention module,which helped to capture inter-local cross-channel interaction information appropriately.Finally,we introduced an innovative ResCBAMG module between the different C2f modules and detection heads of the bottom-up feature fusion layer.This innovative structure helped the model to better focus on the target area.The precision and robustness of oriented target detection were also improved.Experimental results on the DOTA-v1.5 dataset showed that the detection Precision,mAP@0.5,and mAP@0.5:0.95 metrics of the improved model are better compared to the original model.This improvement is effective in detecting small targets and complex scenes.

Automatic road extraction framework based on codec network

Road extraction based on deep learning is one of hot spots of semantic segmentation in the past decade.In this work,we proposed a framework based on codec network for automatic road extraction from remote sensing images.Firstly,a pre-trained ResNet34 was migrated to U-Net and its encoding structure was replaced to deepen the number of network layers,which reduces the error rate of road segmentation and the loss of details.Secondly,dilated convolution was used to connect the encoder and the decoder of network to expand the receptive field and retain more low-dimensional information of the image.Afterwards,the channel attention mechanism was used to select the information of the feature image obtained by up-sampling of the encoder,the weights of target features were optimized to enhance the features of target region and suppress the features of background and noise regions,and thus the feature extraction effect of the remote sensing image with complex background was optimized.Finally,an adaptive sigmoid loss function was proposed,which optimizes the imbalance between the road and the background,and makes the model reach the optimal solution.Experimental results show that compared with several semantic segmentation networks,the proposed method can greatly reduce the error rate of road segmentation and effectively improve the accuracy of road extraction from remote sensing images.

Evaluation of corneal parameter changes under different accommodative stimuli with Scheimpflug imaging-based tomography

AIM:To evaluate the reliability of measurements of corneal changes with accommodation in healthy eyes using a Scheimpflug imaging-based system and how these measurements distribute in the normal population.METHODS:Prospective,non-randomized,comparative study including 27 healthy subjects(54 eyes),including emmetropia(13 eyes),myopia(17 eyes),hyperopia(4 eyes)and astigmatism(20 eyes)groups.In all cases,a complete eye examination was performed,including the analysis of corneal changes with different accommodative stimuli(+2.00,0.00 and-3.00 D)using the Pentacam AXL system.The investigation was structured in 2 phases:repeatability analysis and characterization of accommodation-related corneal changes in healthy populations.RESULTS:In the repeatability analysis,the index of height asymmetry(IHA)showed the greatest variability with the three accommodative stimuli,being the results for the rest of parameters acceptable.The group of emmetropes showed significant differences with accommodative changes in the position of maximum keratometry(Kmax;P<0.05),whereas in the astigmatism group,significant changes were not only observed in the position of Kmax,but also in minimum corneal thickness(MCT),corneal spherical aberration,and total and low order aberration root mean square(all P<0.05).Likewise,a significant difference was found in the displacement of the X position of Kmax with+2.00 D and-3.00 D in the myopia group(P=0.033)as well as in changes with+2.00 D and-3.00 D in the magnitude of the position vector of Kmax in the emmetropia group(P<0.05).No significant changes were found between accommodative stimuli in the displacement of coordinates of MCT(P≥0.109).CONCLUSION:The position of Kmax and MCT in healthy corneas can change significantly when presenting different accommodative stimuli using the accommodation mode of the Pentacam system,with different trends in these accommodation-related corneal changes between refractive errors.Likewise,the consistency of the measurements obtained with Scheimpflug has been confirmed.

Ex⁃situ Measurement of Internal Deformation in Ball Grid Array Package with Digital Volume Correlation

In spacecraft electronic devices,the deformation of solder balls within ball grid array(BGA)packages poses a significant risk of system failure.Therefore,accurately measuring the mechanical behavior of solder balls is crucial for ensuring the safety and reliability of spacecraft.Although finite element simulations have been extensively used to study solder ball deformation,there is a significant lack of experimental validation,particularly under thermal cycling conditions.This is due to the challenges in accurately measuring the internal deformations of solder balls and eliminating the rigid body displacement introduced during ex-situ thermal cycling tests.In this work,an ex-situ three-dimensional deformation measurement method using X-ray computed tomography(CT)and digital volume correlation(DVC)is proposed to overcome these obstacles.By incorporating the layer-wise reliability-guided displacement tracking(LW-RGDT)DVC with a singular value decomposition(SVD)method,this method enables accurate assessment of solder ball mechanical behavior in BGA packages without the influence of rigid body displacement.Experimental results reveal that BGA structures exhibit progressive convex deformation with increased thermal cycling,particularly in peripheral solder balls.This method provides a reliable and effective tool for assessing internal deformations in electronic packages under ex-situ conditions,which is crucial for their design optimization and lifespan predictions.