Magnetic Resonance Imaging Classification Research Articles

Development of hybrid models based on deep learning and optimized machine learning algorithms for brain tumor Multi-Classification

Accurate classification of magnetic resonance imaging (MRI) images of brain tumors is crucial for early diagnosis and effective treatment in clinical studies. In these studies, many models supported by artificial intelligence (AI) have been proposed as assistant systems for experts. In particular, state-of-the-art deep learning (DL) models that have proven themselves in different fields have been effectively used in the classification of brain MRI images. However, the low accuracy of multiple classification of these images still leads researchers to conduct different studies in this field. Especially there is a need to develop models that achieve high accuracy on original images, and it is believed that this need can be met not only by DL models but also by classical machine learning (ML) algorithms. However, it is critical to choose the hyperparameters correctly for the hybrid use of ML algorithms with DL models. This study proposes a powerful new hybrid method to perform multiple classifications of brain tumors with high accuracy. This method also uses a novel convolutional neural network (CNN) model for feature extraction, and ML algorithms are used for feature classification. In addition, nine state-of-the-art CNN models are used for CNN performance comparison. The Bayesian optimization algorithm is used to obtain the optimal hyperparameter values of ML algorithms. The results obtained from the experimental studies show that the proposed hybrid model achieved 97.15% mean classification accuracy and 97% recall, precision, and F1-score values. Other hybrid models, including DarkNet19-SVM, DarkNet53-SVM, DenseNet201-SVM, EfficientNetB0-SVM, InceptionV3-SVM, NasNetMobile-SVM, ResNet50-SVM, ResNet101-SVM, and Xception-SVM, achieved mean classification accuracies of 95.01%, 95.58%, 96.87%, 97.01%, 95.3%, 95.01%, 96.3%, 95.87%, and 96.23%, respectively. Additionally, the proposed hybrid model exhibited remarkable time efficiency, accomplishing the classification process in a mere 67 min. Conversely, the model that exhibited the lowest time efficiency was the InceptionV3, with a processing time of 370 min. In terms of computational complexity, the EfficientNetB0 model is the most efficient. Despite the higher computational complexity of the proposed CNN model compared to some other models, it achieves the second-best classification accuracy. These results show that the proposed method performs better than previous studies on the same dataset. Especially in the classification problem, the optimized ML algorithms were superior to CNN classifiers. Finally, except for one, the proposed CNN model achieved better classification accuracies than the state-of-the-art CNN models.

Deep learning‐based robust hybrid approaches for brain tumor classification in magnetic resonance images

AbstractThe classification of brain images is an immensely challenging task and one of the most valuable and frequently employed procedures in the field of medical diagnostics. Deep Learning (DL), which falls under the umbrella of artificial intelligence, has introduced novel techniques for automated analysis of medical images. The objective of this study was to develop two hybrid DL models for the segmentation and classification of brain magnetic resonance imaging (MRI) data. The first hybrid model combines fully convolutional networks (FCNs) and residual networks (ResNets) for both segmentation and classification purposes. The second model is a hybrid of SegNet for segmentation and MobileNet for classification. Rest assured, it will provide you with a dependable and promising performance. The analysis was conducted on a labeled dataset comprising images of glioma, meningioma, pituitary, and no‐tumor cases. Python was utilized to implement the aforementioned hybrid models. The proposed models achieved remarkable accuracies of 93.9% and 91.3%, respectively, when evaluated on a brain tumor MRI dataset. Additionally, a comprehensive evaluation of the proposed models was conducted, comparing them with other existing hybrid models in terms of metrics such as the dice score, Jaccard index, positive predictive value (PPV), false predictive value (FPV), testing time, precision, recall, specificity, and F1 score. The results demonstrated that the proposed DL technique can significantly assist doctors and radiologists in the early detection of brain tumor cells.

Longitudinal trajectory of medial gastrocnemius muscle growth in the first years of life.

To define the longitudinal trajectory of gastrocnemius muscle growth in 6- to 36-month-old children with and without spastic cerebral palsy (SCP) and to compare trajectories by levels of gross motor function (Gross Motor Function Classification System, GMFCS) and presumed brain-lesion timing. Twenty typically developing children and 24 children with SCP (GMFCS levels I-II/III-IV = 15/9), were included (28/16 females/males; mean age at first scan 15.4 months [standard deviation 4.93, range 6.24-23.8]). Three-dimensional freehand ultrasound was used to repeatedly assess muscle volume, length, and cross-sectional area (CSA), resulting in 138 assessments (mean interval 7.9 months). Brain lesion timing was evaluated with magnetic resonance imaging classification. Linear mixed-effects models defined growth rates, adjusted for GMFCS levels and presumed brain-lesion timing. At age 12 months, children with SCP showed smaller morphological muscle size than typically developing children (5.8 mL vs 9.8 mL, p < 0.001), while subsequently no differences in muscle growth were found between children with and without SCP (muscle volume: 0.65 mL/month vs 0.74 mL/month). However, muscle volume and CSA growth rates were lower in children classified in GMFCS levels III and IV than typically developing children and those classified in GMFCS levels I and II, with differences ranging from -56% to -70% (p < 0.001). Muscle growth is already hampered during infancy in SCP. Muscle size growth further reduces with decreasing functional levels, independently from the brain lesion. Early monitoring of muscle growth combined with early intervention is needed.

Short-term results after robot-assisted surgery for primary rectal cancers requiring beyond total mesorectal excision in multiple compartments.

Rectal cancers requiring beyond total mesorectal excision (bTME) are traditionally operated using an open approach, but the use of minimally invasive robot-assisted procedures is increasing. Introduction of minimal invasive surgery for complex cancer cases could be associated with compromised surgical margins or increased complication rates. Therefore, reporting results both clinical and oncological in large series is important. Since bTME procedure reports are heterogeneous, comparing results is often difficult. In this study, a magnetic resonance imaging (MRI) classification system was used to describe the bTME surgery according to pelvic compartments. Consecutive patients with primary rectal cancer operated with laparoscopic robot-assisted bTME were prospectively included for 2 years. All patients had tumors that threatened the mesorectal fascia, invaded adjacent organs, and/or involved metastatic pelvic lateral lymph nodes. Short-term clinical outcomes and oncological specimen quality were registered. Surgery was classified according to pelvic compartments resected. Clear resection margins (R0 resection) were achieved in 95 out of 105 patients (90.5%). About 26% had Accordion Severity Grading System of Surgical Complications grade 3-4 complications and 15% required re-operations. About 7% were converted to open surgery. The number of compartments resected ranged from one to the maximum seven, with 83% having two or three compartments resected. All 10 R1 resections occurred in the lateral and posterior compartments. The short-term clinical outcomes and oncological specimen quality after robot-assisted bTME surgery were comparable to previously published open bTME surgery. The description of surgical procedures using the Royal Marsden MRI compartment classification was feasible.

Impact of lateral meniscus injury detected by preoperative magnetic resonance imaging on midterm results after unicompartmental knee arthroplasty

BackgroundThe significance of lateral meniscus injury and its impact on success rates with medial unicompartmental knee arthroplasty (UKA) is still debated among scholars. This study aims to investigate whether preoperative magnetic resonance imaging (MRI) findings of lateral meniscus injury influence midterm outcomes following UKA. MethodsThis study recruited 104 patients who underwent medial mobile-bearing UKA. Based on the extent of lateral meniscus injury indicated by the preoperative MRI and Stoller's classification system, patients were divided into two groups: the normal group (grade 0), and the lateral meniscus injury group (grade 3). Further, preoperative demographic and clinical outcome data (Hospital for Special Surgery score, lateral knee pain, squatting, and knee extension ability) were compared at least 2 years postoperatively. ResultsNo statistically significant difference in knee function or clinical outcome was found between the normal group (n = 59) and the lateral meniscus injury group (n = 45). During the 39.2 months (range: 24–64 months), no patient required any reoperation or revision procedures. ConclusionIn summary, the presence of lateral meniscus injury as determined by preoperative MRI does not affect the midterm results after UKA. Without additional treatment for lateral meniscus injury, UKA can yield desired outcomes regardless of the presence of lateral meniscus injury on preoperative MRI.

Deep Learning-Based Magnetic Resonance Image Segmentation and Classification for Alzheimer’s Disease Diagnosis

Accurate and rapid detection of Alzheimer’s disease (AD) using magnetic resonance imaging (MRI) gained considerable attention among research workers because of an increased number of current researches being driven by deep learning (DL) methods that have accomplished outstanding outcomes in variety of domains involving medical image analysis. Especially, convolution neural network (CNN) is primarily applied for the analyses of image datasets according to the capability of handling massive unstructured datasets and automatically extracting significant features. Earlier detection is dominant to the success and development interferences, and neuroimaging characterizes the potential regions for earlier diagnosis of AD. The study presents and develops a novel Deep Learning-based Magnetic Resonance Image Segmentation and Classification for AD Diagnosis (DLMRISC-ADD) model. The presented DLMRISC-ADD model mainly focuses on the segmentation of MRI images to detect AD. To accomplish this, the presented DLMRISC-ADD model follows a two-stage process, namely, skull stripping and image segmentation. At the preliminary stage, the presented DLMRISC-ADD model employs U-Net-based skull stripping approach to remove skull regions from the input MRIs. Next, in the second stage, the DLMRISC-ADD model applies QuickNAT model for MRI image segmentation, which identifies distinct parts such as white matter, gray matter, hippocampus, amygdala, and ventricles. Moreover, densely connected network (DenseNet201) feature extractor with sparse autoencoder (SAE) classifier is used for AD detection process. A brief set of simulations is implemented on ADNI dataset to demonstrate the improved performance of the DLMRISC-ADD method, and the outcomes are examined extensively. The experimental results exhibit the effectual segmentation results of the DLMRISC-ADD technique.

Segmentation-Based Analysis of T2- and T1-Weighted Dynamic Magnetic Resonance Images Provides Adequate Observer Agreement in the Evaluation of Interstitial Lung Disease.

The aim of the study is to quantify observer agreement in the magnetic resonance imaging (MRI) classification of inflammatory or fibrotic interstitial lung disease (ILD). Our study is a preliminary analysis of a larger prospective cohort. The MRI images of 18 patients with ILD (13 females; mean age, 65 years) were acquired in a 1.5 T scanner and included axial fat-saturated T2-weighted (T2-WI, n = 18) and coronal fat-saturated T1-weighted images before and 1, 3, 5, and 10 minutes after gadolinium administration (n = 16). The MRI studies were evaluated with 2 different methods: a qualitative evaluation (visual assessment and measurement of few regions of interest; evaluations were performed independently by 5 radiologists and 3 times by 1 radiologist) and a segmentation-based analysis with software extraction of signal intensity values (evaluations were performed independently by 2 radiologists and twice by 1 radiologist). Interstitial lung disease was classified as inflammatory or fibrotic, based on previously described imaging criteria. Regarding the qualitative evaluation, intraobserver agreement was excellent (κ = 0.92, P < 0.05) for T2-WI and fair (κ = 0.29, P < 0.05) for T1 dynamic study, while interobserver agreement was moderate (κ = 0.56, P < 0.05) and poor (κ = 0.11, P = 0.18), respectively. In contrast, upon segmentation-based analysis, intraobserver and interobserver agreement were excellent for T2-WI (κ = 0.886, P < 0.001; κ = 1.00, P < 0.001; respectively); for T1-WI, intraobserver agreement was excellent (κ = 0.87, P < 0.05) and interobserver agreement was good (κ = 0.75, P < 0.05). Segmentation-based MRI analysis is more reproducible than a qualitative evaluation with visual assessment and measurement of few regions of interest.

Brain MRI in Progressive Supranuclear Palsy with Richardson's Syndrome and Variant Phenotypes.

Brain magnetic resonance imaging (MRI) is used to support the diagnosis of progressive supranuclear palsy (PSP). However, the value of visual descriptive, manual planimetric, automatic volumetric MRI markers and fully automatic categorization is unclear, particularly regarding PSP predominance types other than Richardson's syndrome (RS). To compare different visual reading strategies and automatic classification of T1-weighted MRI for detection of PSP in a typical clinical cohort including PSP-RS and (non-RS) variant PSP (vPSP) patients. Forty-one patients (21 RS, 20 vPSP) and 46 healthy controls were included. Three readers using three strategies performed MRI analysis: exclusively visual reading using descriptive signs (hummingbird, morning-glory, Mickey-Mouse), visual reading supported by manual planimetry measures, and visual reading supported by automatic volumetry. Fully automatic classification was performed using a pre-trained support vector machine (SVM) on the results of atlas-based volumetry. All tested methods achieved higher specificity than sensitivity. Limited sensitivity was driven to large extent by false negative vPSP cases. Support by automatic volumetry resulted in the highest accuracy (75.1% ± 3.5%) among the visual strategies, but performed not better than the midbrain area (75.9%), the best single planimetric measure. Automatic classification by SVM clearly outperformed all other methods (accuracy, 87.4%), representing the only method to provide clinically useful sensitivity also in vPSP (70.0%). Fully automatic classification of volumetric MRI measures using machine learning methods outperforms visual MRI analysis without and with planimetry or volumetry support, particularly regarding diagnosis of vPSP, suggesting the use in settings with a broad phenotypic PSP spectrum. © 2023 The Authors. Movement Disorders published by Wiley Periodicals LLC on behalf of International Parkinson and Movement Disorder Society.

IDENTIFICATION OF NEW BIOMARKERS AND PROMISING NEURO IMAGING OF DIFFUSE AXONAL INJURY: AN OBSERVATIONAL STUDY

Objective: To determine the clinical patterns of the newly identified novel biomarkers of prognosis among adult patients with traumatic brain injury (TBI), diffuse axonal injury (DAI) and ophthalmic injury (OI) and micro bleeds in the Brain. Methods: According to the online search engine from the databases of PubMed, Embase, Ovid Science Direct, and the way of sciences of the original contribution articles which are providing with the essential information about outcome in patients with DAI in general, DAI vs. non-DAI, related to magnetic resonance imaging (MRI) classification , related to lesion and their specific , Specific location/load and their future outcomes. We have gone through the major references and checked their quality control (QC) assessments performed through International standard protocol/guidelines and our own experiences. Results: A total of 585 articles were included. TBI patients with DAI patients favorable outcomes are (75-90 %). The risk associated unfavorable outcome (RAUO) is in TBI with DAI was three to four times higher than in TBI without DAI in our Asian Indian patients.The Odds ratio (OR) for their unfavorable outcome was 2.7 as per increase of DAI grade on the MRI findings. Lesions are located in the area of the corpus callosum (CC) associated with an unfavorable outcomes reported frequently (85-95%). Conclusions: DAI patients management through MRI and patients with with TBI results unfavorable outcome are increasing in larger in numbers day by day in trauma care. MRI grading, locations, pattern of diffuse (POD) and axonal injury rates (AIR) are increasing in their three to four fold with every grade. Associated Lesions in the corpus callosum (CC) are associated with an unfavorable outcomes are those frequently associated with the DAI patients in TBI. Need appropriate diagnosis and management for DAI. Our newly developed biomarkers are quite useful for DAI patients associated within the TBI patients management and follow-up.

DEEP JOINT RP-NET-BASED SEGMENTATION ALGORITHM AND OPTIMIZED DEEP LEARNING FOR SEVERITY PREDICTION OF BRAIN TUMOR

The uncontrolled growth of cells in a particular area is referred to as a tumor. The premature and precise identification of the tumor and its level have a straight impression on the patient’s survival, treatment process, and tumor progression computation. However, in the medical field, picture segmentation and classification are more important and difficult processes. Typically, the Magnetic Resonance Imaging (MRI) modality can detect malignancy. Segmenting tumor images with respect to Cerebrospinal Fluid (CSF), Grey Matter (GM), and White Matter is the most important task in MRI identification or classification (WM). The introduction of medical image analysis based on radiology pictures is a result of the significant contributions of engineering, data sciences, and medicine. The precise and automatic segmentation of tumors affords excessive support to doctors in the medicinal area, speed detection in the treatment process, computer-aided operation, radiation treatment and so on. Thus, Remora Aquila optimization (RAO)-enabled deep learning is devised for tumor classification and its severity classification. The deep learning approach is utilized for categorizing tumors as normal or abnormal as well as their severity grades. The RAO-aided deep learning system achieved improved performance with prediction error, specificity, sensitivity and testing accuracy of 0.072, 0.905, 0.925, and 0.917, respectively.

BRAIN TUMOR SEGMENTATION AND CLASSIFICATION USING OPTIMIZED U-NET

In the brain, the abnormal growth of cells or solid intracranial neoplasm is known as brain tumor, which is one of the world’s most tedious diseases. Hence, there is a need for segmentation and classification of the brain tumor accurately. It is difficult to separate the tumor tissues and other tissues from the brain. The major aim of this research is to use magnetic resonance imaging (MRI) segment and classify the brain tumor and all the abnormalities in the brain. The MRI is initially fed into the pre-processing system and then it is segmented using the region-growing segmentation algorithm in the pre-operative MRI. It produces the segmented area and it is forwarded for classification. In the classification step, the Honey Badger Algorithm (HBA) is applied to train the U-Net classifier. The tumor tissues and the different types of tissues or abnormalities in brain tumors are classified by this algorithm. Overall, the post-operative and pre-operative MRI brain tumor segmentation and classification consist of the same steps. To find out the pixel changes, both the segmented output of pre-operative and post-operative MRI was compared. It helps in finding the emerging tumor after surgery and the success rate of surgery. Based on pre-operative MRI, the implemented scheme has maximum specificity, sensitivity, and accuracy of 0.977, 0.968, and 0.949.

Effective Brain Tumor Classification Using Deep Residual Network-Based Transfer Learning

Brain tumor classification is an essential task in medical image processing that provides assistance to doctors for accurate diagnoses and treatment plans. A Deep Residual Network based Transfer Learning to a fully convoluted Convolutional Neural Network (CNN) is proposed to perform brain tumor classification of Magnetic Resonance Images (MRI) from the BRATS 2020 dataset. The dataset consists of a variety of pre-operative MRI scans to segment integrally varied brain tumors in appearance, shape, and histology, namely gliomas. A Deep Residual Network (ResNet-50) to a fully convoluted CNN is proposed to perform tumor classification from MRI of the BRATS dataset. The 50-layered residual network deeply convolutes the multi-category of tumor images in classification tasks using convolution block and identity block. Limitations such as Limited accuracy and complexity of algorithms in CNN-based ME-Net, and classification issues in YOLOv2 inceptions are resolved by the proposed model in this work. The trained CNN learns boundary and region tasks and extracts successful contextual information from MRI scans with minimal computation cost. The tumor segmentation and classification are performed in one step using a U-Net architecture, which helps retain spatial features of the image. The multimodality fusion is implemented to perform classification and regression tasks by integrating dataset information. The dice scores of the proposed model for Enhanced Tumor (ET), Whole Tumor (WT), and Tumor Core (TC) are 0.88, 0.97, and 0.90 on the BRATS 2020 dataset, and also resulted in 99.94% accuracy, 98.92% sensitivity, 98.63% specificity, and 99.94% precision.

Self-supervised pretraining improves the performance of classification of task functional magnetic resonance imaging.

Decoding brain activities is one of the most popular topics in neuroscience in recent years. And deep learning has shown high performance in fMRI data classification and regression, but its requirement for large amounts of data conflicts with the high cost of acquiring fMRI data. In this study, we propose an end-to-end temporal contrastive self-supervised learning algorithm, which learns internal spatiotemporal patterns within fMRI and allows the model to transfer learning to datasets of small size. For a given fMRI signal, we segmented it into three sections: the beginning, middle, and end. We then utilized contrastive learning by taking the end-middle (i.e., neighboring) pair as the positive pair, and the beginning-end (i.e., distant) pair as the negative pair. We pretrained the model on 5 out of 7 tasks from the Human Connectome Project (HCP) and applied it in a downstream classification of the remaining two tasks. The pretrained model converged on data from 12 subjects, while a randomly initialized model required 100 subjects. We then transferred the pretrained model to a dataset containing unpreprocessed whole-brain fMRI from 30 participants, achieving an accuracy of 80.2 ± 4.7%, while the randomly initialized model failed to converge. We further validated the model's performance on the Multiple Domain Task Dataset (MDTB), which contains fMRI data of 26 tasks from 24 participants. Thirteen tasks of fMRI were selected as inputs, and the results showed that the pre-trained model succeeded in classifying 11 of the 13 tasks. When using the 7 brain networks as input, variations of the performance were observed, with the visual network performed as well as whole brain inputs, while the limbic network almost failed in all 13 tasks. Our results demonstrated the potential of self-supervised learning for fMRI analysis with small datasets and unpreprocessed data, and for analysis of the correlation between regional fMRI activity and cognitive tasks.

Intact revision rotator cuff repair stabilizes muscle atrophy and fatty infiltration after minimum follow up of two years

BackgroundThe extent of fatty infiltration and rotator cuff (RC) atrophy is crucial for the clinical results after rotator cuff repair (RCR). The purpose of this study was to evaluate changes in fatty infiltration and RC atrophy after revision RCR and to correlate them with functional outcome parameters.MethodsPatients who underwent arthroscopic revision RCR for symptomatic recurrent full-thickness tear of the supraspinatus tendon between 2008 and 2014 and were retrospectively reviewed with a minimum follow up of 2 years. Magnetic resonance imaging (MRI) was performed pre- and postoperatively to assess 1) tendon integrity after revision RCR according to Sugaya classification, (2) RC atrophy according to Thomazeau classification, and (3) fatty infiltration according to Fuchs MRI classification. Constant score (CS) and the American Shoulder and Elbow Surgeon (ASES) score were used to correlate functional outcome, tendon integrity, and muscle degeneration.Results19 patients (17 males and 2 females) with a mean age of 57.5 years (range, 34 to 72) were included into the study at a mean follow-up of 50.3 months (range, 24 – 101). At final evaluation, 9 patients (47%) presented with intact RCR and 10 patients (53%) suffered a re-tear after revision repair. No progress of fatty infiltration was observed postoperatively in the group with intact RC, atrophy progressed in only 1 out of 9 patient (11%). Fatty infiltration progressed in 5/10 patients (50%) and RC atrophy increased in 2/10 patients (20%) within the re-tear group. CS (42.7 ± 17.7 preop, 65.2 ± 20.1 postop) and ASES (47.7 ± 17.2 preop, 75.4 ± 23.7 postop) improved significantly from pre- to postoperatively (p < 0.001). A positive correlation between fatty infiltration and RC integrity was detected (r = 0.77, p < 0.01). No correlation between clinical outcome and tendon integrity or RC atrophy was observed.ConclusionArthroscopic revision RCR leads to reliable functional outcomes even in case of a recurrent RC retear. An intact RCR maintains the preoperative state of fatty infiltration and muscle atrophy but does not lead to muscle regeneration.Level of evidenceLevel IV; Therapeutic study.

A novel classification of Kienbock's disease based on magnetic resonance imaging.

We devised a classification system for Kienbock's disease using magnetic resonance imaging (MRI). Moreover, we compared it with the modified Lichtman classification and evaluated the inter-observer reliability. Eighty-eight patients diagnosed with Kienbock's disease were included. All patients were classified using the modified Lichtman and MRI classifications. MRI staging was based on factors including partial marrow oedema, cortical integrity of the lunate, and dorsal subluxation of the scaphoid. The inter-observer reliability was evaluated. We also evaluated the presence of a displaced coronal fracture of the lunate and investigated its association with the presence of a dorsal subluxation of the scaphoid. Seven patients were categorized into stage I, 13 into II, 33 into IIIA, 33 into IIIB, and two into IV using the modified Lichtman classification. Six patients were categorized into stage I, 12 into II, 56 into IIIA, ten into IIIB, and four into IV using the MRI classification. The greatest shift between the stages was observed in stages IIIA and IIIB when the results of the two classification systems were compared. The inter-observer reliability of the MRI classification was greater than that of the modified Lichtman classification. Fifteen cases with a displaced coronal fracture of the lunate were identified, and a dorsal subluxation of the scaphoid was significantly more present in these patients. The MRI classification system is more reliable than is the modified Lichtman classification. MRI classification reflects carpal misalignment with higher fidelity and is more appropriate for classification into stages IIIA and IIIB.

An evolutionary explainable deep learning approach for Alzheimer's MRI classification

Deep Neural Networks (DNN) are prominent Machine Learning (ML) algorithms widely used, especially in medical tasks. Among them, Convolutional Neural Networks (CNN) are well-known for image-based tasks and have shown excellent performance. In contrast to this remarkable performance, one of their most fundamental drawbacks is their inability to clarify the cause of their outputs. Moreover, each ML algorithm needs to present an explanation of its output to the users to increase its reliability. Occlusion Map is a method used for this purpose and aims to find regions of an image that have a significant impact on determining the network's output, which does this through an iterative process of occluding different regions of images. In this study, we used Magnetic Resonance Imaging (MRI) scans from Alzheimer's Disease Neuroimaging Initiative (ADNI) and trained a 3D-CNN model to diagnose Alzheimer's Disease (AD) patients from cognitively normal (CN) subjects. We tried to combine a genetic algorithm-based Occlusion Map method with a set of Backpropagation-based explainability methods, and ultimately, we found a brain mask for AD patients. Also, by comparing the extracted brain regions with the studies in this field, we found that the extracted regions are significantly effective in diagnosing AD from the perspective of Alzheimer's specialists. Our model achieved an accuracy of 87% in 5-fold cross-validation, which is an acceptable accuracy compared to similar studies. We considered a 3D-CNN model with 96% validation accuracy (on unmasked data that includes all 96 distinct brain regions of the Harvard-Oxford brain atlas), which we used in the genetic algorithm phase to produce a suitable brain mask. Finally, using lrp_z_plus_fast explainability method, we achieved 93% validation accuracy with only 29 brain regions.

Comparison of high-intensity focused ultrasound for the treatment of internal and external adenomyosis based on magnetic resonance imaging classification

Objective To compare the safety and efficacy of high-intensity focused ultrasound (HIFU) treatment for patients with internal or external adenomyosis based on magnetic resonance imaging (MRI) classification. Materials and methods A total of 238 patients with internal adenomyosis and 167 patients with external adenomyosis who received HIFU treatment were enrolled. HIFU treatment results and adverse effects between patients with internal and external adenomyosis were compared. Results The treatment time and sonication time for patients with external adenomyosis were significantly longer than that for patients with internal adenomyosis. The total energy used and EEF for patients with external adenomyosis were higher than that for patients with internal adenomyosis (p < 0.05). The pre-HIFU median dysmenorrhea score in patients with internal or external adenomyosis was 5 or 8 points, the median score decreased to 1 or 3 points in these two groups at 18-month post-HIFU (p < 0.05). The relief rate of dysmenorrhea was 79.5% in patients with internal adenomyosis, and it was 80.8% in patients with external adenomyosis. The pre-HIFU median menorrhagia score in patients with internal or external adenomyosis was 4 or 3 points, the median score decreased to 1 point in both groups at 18-month post-HIFU with a relief rate of 86.2% and 77.1%, respectively (p = 0.030). No serious complication occurred in any of these patients. Conclusions HIFU is a safe and effective treatment either for patients with internal adenomyosis or external adenomyosis. It seemed that internal adenomyosis is easier to be treated with HIFU and with a higher relief rate of menorrhagia than external adenomyosis.

Magnetic Resonance Imaging Classification For The Assessment Of Tendon Integrity Of The Rotator Cuff After Repair. A Reliability Study

Magnetic Resonance Imaging Classification For The Assessment Of Tendon Integrity Of The Rotator Cuff After Repair. A Reliability Study

Deep Learning Network with Spatial Attention Module for Detecting Acute Bilirubin Encephalopathy in Newborns Based on Multimodal MRI.

Acute bilirubin encephalopathy (ABE) is a significant cause of neonatal mortality and disability. Early detection and treatment of ABE can prevent the further development of ABE and its long-term complications. Due to the limited classification ability of single-modal magnetic resonance imaging (MRI), this study aimed to validate the classification performance of a new deep learning model based on multimodal MRI images. Additionally, the study evaluated the effect of a spatial attention module (SAM) on improving the model's diagnostic performance in distinguishing ABE. This study enrolled a total of 97 neonates diagnosed with ABE and 80 neonates diagnosed with hyperbilirubinemia (HB, non-ABE). Each patient underwent three types of multimodal imaging, which included T1-weighted imaging (T1WI), T2-weighted imaging (T2WI), and an apparent diffusion coefficient (ADC) map. A multimodal MRI classification model based on the ResNet18 network with spatial attention modules was built to distinguish ABE from non-ABE. All combinations of the three types of images were used as inputs to test the model's classification performance, and we also analyzed the prediction performance of models with SAMs through comparative experiments. The results indicated that the diagnostic performance of the multimodal image combination was better than any single-modal image, and the combination of T1WI and T2WI achieved the best classification performance (accuracy = 0.808 ± 0.069, area under the curve = 0.808 ± 0.057). The ADC images performed the worst among the three modalities' images. Adding spatial attention modules significantly improved the model's classification performance. Our experiment showed that a multimodal image classification network with spatial attention modules significantly improved the accuracy of ABE classification.

Research on classification method of multimodal magnetic resonance images of Alzheimer's disease based on generalized convolutional neural networks

Alzheimer's disease (AD) is a progressive and irreversible neurodegenerative disease. Neuroimaging based on magnetic resonance imaging (MRI) is one of the most intuitive and reliable methods to perform AD screening and diagnosis. Clinical head MRI detection generates multimodal image data, and to solve the problem of multimodal MRI processing and information fusion, this paper proposes a structural and functional MRI feature extraction and fusion method based on generalized convolutional neural networks (gCNN). The method includes a three-dimensional residual U-shaped network based on hybrid attention mechanism (3D HA-ResUNet) for feature representation and classification for structural MRI, and a U-shaped graph convolutional neural network (U-GCN) for node feature representation and classification of brain functional networks for functional MRI. Based on the fusion of the two types of image features, the optimal feature subset is selected based on discrete binary particle swarm optimization, and the prediction results are output by a machine learning classifier. The validation results of multimodal dataset from the AD Neuroimaging Initiative (ADNI) open-source database show that the proposed models have superior performance in their respective data domains. The gCNN framework combines the advantages of these two models and further improves the performance of the methods using single-modal MRI, improving the classification accuracy and sensitivity by 5.56% and 11.11%, respectively. In conclusion, the gCNN-based multimodal MRI classification method proposed in this paper can provide a technical basis for the auxiliary diagnosis of Alzheimer's disease.