Magnetic Resonance Imaging Classification Research Articles

Evaluation of anterolateral ligament healing in patients with anterior cruciate ligament injury without anterior cruciate ligament reconstruction and its relationship with clinical examination tests.

This study radiologically investigated the potential of preoperative recovery of the anterolateral ligament (ALL) in patients who did not undergo anterior cruciate ligament reconstruction (ACLR) surgery in the early stages despite surgical indication. The secondary aim was to investigate the relationship between ALL injuries in magnetic resonance imaging (MRI) and the results of Lachman and pivot shift tests performed preoperatively and postoperatively under anaesthesia amongpatients who underwent ACLR in acute and chronic stages. The study examined 100 patients aged 19-51 years who underwent anatomic single bundle ACLR using hamstring autografts for the ACL. Based on the time between the injury and ACLR, they were divided into acute (≤6 months) and chronic (>6 months) groups. The chronic group had three subgroups of ACLR times: 7-12, 13-24and >24 months. Complete recovery was defined as a change from grade B or C upon injury to grade A before surgery according to the Muramatsu MRI classification. The relationship between preoperative and postoperative Lachman grades and pivot shifting test results and the degree of ALL injury was evaluated. In the chronic group, 19/61 patients (31.1%) had preoperative grade A, and 4/61 patients (6.5%) had grade A at the time of injury. In the acute group, 22/29 patients with grades B and C (76%) had grade 3 preoperative pivot shifting. In the acute group, grades 1 and 2 were more common in patients with grades B and C (10/29, 35%) than grade A (0/10, 0.0%). The improvement rate was 24.6% amongpatients who did not receive ACLR in the acute phase and underwent ACLR in the chronic phase. In the acute group, the degree of ALL damage was related to the pivot shift results. However, the results in the chronic phase were not associated with the degree of injury. Level IV.

Clinical characteristics and rehabilitation potential in children with cerebral palsy based on MRI classification system.

The correlation of clinical characteristics of cerebral palsy (CP) and the magnetic resonance imaging classification system (MRICS) for (CP) is inconsistent. Specifically, the variance in rehabilitation potential across MRICS remains underexplored. To investigate the clinical characteristics and potential for rehabilitation in children with CP based on MRICS. Children with CP admitted to the Department of Rehabilitation, Children's Hospital of Chongqing Medical University between 2017 and 2021 were included in the study. Qualified cases underwent a follow-up period of at least one year. The clinical characteristics of CP among different MRICS were analyzed, then the rehabilitation potential was explored by a retrospective cohort study. Among the 384 initially enrolled children, the male-to-female ratio was 2.3:1, and the median age of diagnosis was 6.5 months (interquartile range: 4-12). The most prevalent MRICS categorization was predominant white matter injury (40.6%), followed by miscellaneous (29.2%) and predominant gray matter injury (15.6%). For the predominant white matter injury and miscellaneous categories, spastic diplegia emerged as the leading subtype of CP, with incidences of 59.6% and 36.6%, respectively, while mixed CP (36.7%) was the most common type in children with predominant gray matter. Notably, 76.4% of children with predominant white matter injury were classified as levels I-III on the gross motor function classification system (GMFCS), indicating significantly less severity than other groups (χ2 = 12.438, p = 0.013). No significant difference across MRICS categories was observed for the manual ability classification system (MACS) (H = 8.176, p = 0.085). Rehabilitation potential regarding fine motor function and adaptability based on Gesell assessment was dependent on MRICS over the follow-up period. Children with normal MRI scans exhibited superior rehabilitation outcomes. Commencing rehabilitation at an earlier stage produced consistent and beneficial results in terms of fine motor function and adaptability across all MRICS categories. Moreover, participants below 2 years of age demonstrated enhanced rehabilitation potential regarding fine motor outcomes and adaptability within the MRICS framework. MRICS displayed a significant association with clinical characteristics and rehabilitation efficacy in children with CP.

Early Detection of Alzheimers Disease using Deep Learning

Alzheimer’s disease (AD) poses a significant challenge to global healthcare systems due to its progressive nature and impact on patient’s lives. Accurate and early detection of AD is crucial for timely intervention and management. In this paper, we propose the use of deep learning models, including Convolutional Neural Networks (CNNs), MobileNet, and VGG16 for the classification of Magnetic Resonance Imaging (MRI) scans into different AD stages. Index Terms—Alzheimer’s disease , Deep Learning , Convolu- tional Neural Networks(CNN) , MobileNet , VGG16 , MRI scans

Magnetic resonance imaging features and classification of intraspinal echinococcosis

ObjectiveThis study aimed to investigate the magnetic resonance imaging (MRI) features and classification of echinococcosis of the spinal canal. MethodsThe clinical manifestations and MRI findings and classification of 19 patients diagnosed with intraspinal echinococcosis over 10 years (2011–2020) were retrospectively analyzed. ResultsThe mean age of the nine males and 10 females was 39 years (range 28–65 years). Among these, the number of cases with thoracic, lumbar, sacral, thoracolumbar, and lumbosacral, cervical, and lumbar segments was nine (47 %), five (26 %), one (5 %), one (5 %), two (11 %), and one (5 %) cases, respectively. Furthermore, 13 cases (69 %) involved adjacent vertebral bodies, accessories, and surrounding soft tissues. The lesion was confined to the intramedullary, extramedullary subdural, extramedullary epidural, and multiple spaces in one (Type I) (5 %), four (Type II) (21 %), one (Type III) (5 %), and thirteen (69 %) cases, respectively. Moreover, nine cases (47 %) had a history of hydatid disease in the spine or other tissues. The clinical manifestations were chest and lumbosacral pain in 18 cases (95 %) and chest and lumbosacral pain accompanied by lower limb dysfunction in four cases (21 %). The MRI revealed different sizes of T1WI low-signal and T2WI high-signal vesicles with a “grape-like” appearance, with 16 cases (84 %) showing low signals on the edge of the cyst wall. ConclusionIntraspinal echinococcosis is rare, even in endemic areas. However, intraspinal echinococcosis should be considered when there is a history of echinococcosis in other sites or when there are clear MRI characteristics for the disease.

Enhancing Robust Liver Cancer Diagnosis: A Contrastive Multi-Modality Learner with Lightweight Fusion and Effective Data Augmentation

This article explores the application of self-supervised contrastive learning in the medical domain, focusing on classification of multi-modality Magnetic Resonance (MR) images. To address the challenges of limited and hard-to-annotate medical data, we introduce multi-modality data augmentation (MDA) and cross-modality group convolution (CGC). In the pre-training phase, we leverage Simple Siamese networks to maximize the similarity between two augmented MR images from a patient, without a handcrafted pretext task. Our approach also combines 3D and 2D group convolution with a channel shuffle operation to efficiently incorporate different modalities of image features. Evaluation on liver MR images from a well-known hospital in Taiwan demonstrates a significant improvement over previous methods. This work contributes to advancing multi-modality contrastive learning, particularly in the context of medical imaging, offering enhanced tools for analyzing complex image data.

The Australian Traumatic Brain Injury Initiative: Systematic Review of Predictive Value of Biological Markers for People With Moderate-Severe Traumatic Brain Injury.

The Australian Traumatic Brain Injury Initiative (AUS-TBI) aims to co-design a data resource to predict outcomes for people with moderate-severe traumatic brain injury (TBI) across Australia. Fundamental to this resource is the data dictionary, which is an ontology of data items. Here, we report the systematic review and consensus process for inclusion of biological markers in the data dictionary. Standardized database searches were implemented from inception through April 2022. English-language studies evaluating association between a fluid, tissue, or imaging marker and any clinical outcome in at least 10 patients with moderate-severe TBI were included. Records were screened using a prioritization algorithm and saturation threshold in Research Screener. Full-length records were then screened in Covidence. A pre-defined algorithm was used to assign a judgement of predictive value to each observed association, and high-value predictors were discussed in a consensus process. Searches retrieved 106,593 records; 1,417 full-length records were screened, resulting in 546 included records. Two hundred thirty-nine individual markers were extracted, evaluated against 101 outcomes. Forty-one markers were judged to be high-value predictors of 15 outcomes. Fluid markers retained following the consensus process included ubiquitin C-terminal hydrolase L1 (UCH-L1), S100, and glial fibrillary acidic protein (GFAP). Imaging markers included computed tomography (CT) scores (e.g., Marshall scores), pathological observations (e.g., hemorrhage, midline shift), and magnetic resonance imaging (MRI) classification (e.g., diffuse axonal injury). Clinical context and time of sampling of potential predictive indicators are important considerations for utility. This systematic review and consensus process has identified fluid and imaging biomarkers with high predictive value of clinical and long-term outcomes following moderate-severe TBI.

Brain MR Image Classification Using Superpixel-Based Deep Transfer Learning.

Nowadays, brain MR (Magnetic Resonance) images are widely used by clinicians to examine the brain's anatomy to look into various pathological conditions like cerebrovascular incidents and neuro-degenerative diseases. Generally, these diseases can be identified with the MR images as "normal" and "abnormal" brains in a two-class classification problem or as disease-specific classes in a multi-class problem. This article presents an ensemble transfer learning-inspired deep architecture that uses the simple linear iterative clustering (SLIC)-based superpixel algorithm along with convolutional neural network (CNN) to classify the MR images as normal or abnormal. Superpixel algorithm segments the input MR images into clusters of regions defined by similarity measures using perceptual feature space. These superpixel images are beneficial as they can provide a compact and meaningful role in computationally demanding applications. The superpixel images are then fed to the deep convolutional neural network (CNN) to classify the images. Three brain MR image datasets, NITR-DHH, DS-75, and DS-160, are used to conduct the experimentation. Through the use of deep transfer learning, the model achieves performance accuracy of 88.15% (NITR-DHH), 98.15% (DS-160), and 98.33% (DS-75) even with the small-scale medical image dataset. The experimentally obtained results demonstrate that the proposed method is promising and efficient for clinical applications for diagnosing different brain diseases via MR images.

Benchmarking the influence of pre-training on explanation performance in MR image classification.

Convolutional Neural Networks (CNNs) are frequently and successfully used in medical prediction tasks. They are often used in combination with transfer learning, leading to improved performance when training data for the task are scarce. The resulting models are highly complex and typically do not provide any insight into their predictive mechanisms, motivating the field of "explainable" artificial intelligence (XAI). However, previous studies have rarely quantitatively evaluated the "explanation performance" of XAI methods against ground-truth data, and transfer learning and its influence on objective measures of explanation performance has not been investigated. Here, we propose a benchmark dataset that allows for quantifying explanation performance in a realistic magnetic resonance imaging (MRI) classification task. We employ this benchmark to understand the influence of transfer learning on the quality of explanations. Experimental results show that popular XAI methods applied to the same underlying model differ vastly in performance, even when considering only correctly classified examples. We further observe that explanation performance strongly depends on the task used for pre-training and the number of CNN layers pre-trained. These results hold after correcting for a substantial correlation between explanation and classification performance.

Kiosk 8R-FB-10 - Exploration of Geometric Deep Learning FBom 3D Cardiac Shape Models for the Classification of Cardiac Amyloid Sub-type FBom 2D Cine Cardiovascular Magnetic Resonance Imaging

Kiosk 8R-FB-10 - Exploration of Geometric Deep Learning FBom 3D Cardiac Shape Models for the Classification of Cardiac Amyloid Sub-type FBom 2D Cine Cardiovascular Magnetic Resonance Imaging

Adversarial Learning for MRI Reconstruction and Classification of Cognitively Impaired Individuals.

Game theory-inspired deep learning using a generative adversarial network provides an environment to competitively interact and accomplish a goal. In the context of medical imaging, most work has focused on achieving single tasks such as improving image resolution, segmenting images, and correcting motion artifacts. We developed a dual-objective adversarial learning framework that simultaneously 1) reconstructs higher quality brain magnetic resonance images (MRIs) that 2) retain disease-specific imaging features critical for predicting progression from mild cognitive impairment (MCI) to Alzheimer's disease (AD). We obtained 3-Tesla, T1-weighted brain MRIs of participants from the Alzheimer's Disease Neuroimaging Initiative (ADNI, N=342) and the National Alzheimer's Coordinating Center (NACC, N = 190) datasets. We simulated MRIs with missing data by removing 50% of sagittal slices from the original scans (i.e., diced scans). The generator was trained to reconstruct brain MRIs using the diced scans as input. We introduced a classifier into the GAN architecture to discriminate between stable (i.e., sMCI) and progressive MCI (i.e., pMCI) based on the generated images to facilitate encoding of disease-related information during reconstruction. The framework was trained using ADNI data and externally validated on NACC data. In the NACC cohort, generated images had better image quality than the diced scans (Structural similarity (SSIM) index: 0.553 ± 0.116 versus 0.348 ± 0.108). Furthermore, a classifier utilizing the generated images distinguished pMCI from sMCI more accurately than with the diced scans (F1-score: 0.634 ± 0.019 versus 0.573 ± 0.028). Competitive deep learning has potential to facilitate disease-oriented image reconstruction in those at risk of developing Alzheimer's disease.

GHA-DenseNet prediction and diagnosis of malignancy in femoral bone tumors using magnetic resonance imaging

Background and objectiveDue to their aggressive nature and poor prognosis, malignant femoral bone tumors present considerable hurdles. Early treatment commencement is essential for enhancing vital and practical outcomes. In this investigation, deep learning algorithms will be used to analyze magnetic resonance imaging (MRI) data to identify bone tumors that are malignant. MethodologyThe study cohort included 44 patients, with ages ranging from 17 to 78 (22 women and 22 males). To categorize T1 and T2 weighted MRI data, this paper presents an improved DenseNet network model for the classification of bone tumor MRI, which is named GHA-DenseNet. Based on the original DenseNet model, the attention module is added to solve the problem that the deep convolutional model can reduce the loss of key features when capturing the location and content information of femoral bone tumor tissue due to the limitation of local receptive field. In addition, the sparse connection mode is used to prune the connection mode of the original model, so as to remove unnecessary and retain more useful fast connection mode, and alleviate the overfitting problem caused by small dataset size and image characteristics. In a clinical model designed to anticipate tumor malignancy, the utilization of T1 and T2 classifier output values, in combination with patient-specific clinical information, was a crucial component. ResultsThe T1 classifier's accuracy during the training phase was 92.88% whereas the T2 classifier's accuracy was 87.03%. Both classifiers demonstrated accuracy of 95.24% throughout the validation phase. During training and validation, the clinical model's accuracy was 82.17% and 81.51%, respectively. The clinical model's receiver operating characteristic (ROC) curve demonstrated its capacity to separate classes. ConclusionsThe proposed method does not require manual segmentation of MRI scans because it makes use of pretrained deep learning classifiers. These algorithms have the ability to predict tumor malignancy and shorten the diagnostic and therapeutic turnaround times. Although the procedure only needs a little amount of radiologists' involvement, more testing on a larger patient cohort is required to confirm its efficacy.

Introducing a deep learning method for brain tumor classification using MRI data towards better performance

Magnetic resonance imaging (MRI) is widely used to detect brain tumors, but its accuracy depends on the physician's expertise and often requires biopsy confirmation. Deep learning, especially in the field of computer vision, has revolutionized the diagnosis and classification of brain tumors using MRI. This study aims to design a sequential brain tumor detection and classification model based on deep learning and using fully convolutional neural networks. The proposed model consists of two steps: distinguishing non-neoplastic brain from neoplastic brain and determining the tumor type of the latter. Two models were trained using the Brain Tumor MRI Dataset. Four optimizers are studied for three classification tasks (Adam, Nesterov momentum, root-mean-square propagation, and adaptive gradient) to achieve the best results. Adam performed best at distinguishing tumor from non-tumor brains, with 100 % training accuracy and 98 % validation and test accuracy. Nesterov Momentum performed best at differentiating the three tumor types, with 100 % training accuracy and 92 % validation and testing accuracy. Nesterov also performed best on the third classification task, with 100 % training accuracy and 95 % validation and test accuracy. Nesterov-based sequence models show significant results compared to literature works. The proposed Nesterov momentum-based sequence model achieves high accuracy in MRI brain tumor detection and classification.

To Help the Practitioner: Imaging of Ovarian Masses According to the O-RADS MRI Ovarian Malignancy Categorical Risk Scale

Objective: to study the informativity of the Ovarian-Adnexal Imaging-Reporting-Data System Magnetic Resonance Imaging (O-RADS MRI) in detection of ovarian cancer in the practice of a radiologist.Material and methods. The study included 271 women who underwent pelvic magnetic resonance imaging (MRI) in the period from August to November, 2021. Two radiologists with 6-year (Researcher 1) and 2-year (Researcher 2) experience retrospectively analyzed MR images, in which all ovarian masses were evaluated according to O-RADS MRI categorical risk scale from 0 to 5. The analysis of true positive, true negative, false positive and false negative rates according to MRI data compared to the reference data (histology or 1-year follow-up) was performed.Results. O-RADS MRI categorial risk scale had high diagnostic performance for the characterization of adnexal lesions (sensitivity 87.5% and 87.5%, specificity 97.84% and 96.75%, area under curve (AUC) 0.991 and 0.986 for Researchers 1 and 2, respectively), as well as good interreader agreement (Cohen’s kappa coefficient 0,83). Researchers’ mistakes were analyzed, as well as controversial issues of categorial affiliation of ovarian masses that caused false positive and false negative results. Visual aids for ovarian MRI in accordance with O-RADS MRI classification were made for quick orientation in the system to simplify creating the radiology report.Conclusion. O-RADS MRI categorial risk scale proved to be a reliable working tool for the communication between radiologist and gynecologist-oncologist. However, the question of categorial affiliation of a number of nosological names that are not reflected in O-RADS MRI and require further study continues to be debated.

Brain MRI Image Analysis and Segmentation using Machine Learning

The brain magnetic resonance imaging (MRI), analysis and segmentation plays one of the crucial roles in medical diagnosis and facilitates in an early detection of diseases in critical medical conditions, Due to the structural complexity and type of the tumor, radiologists are facing difficulties in extracting essential features of the image which are crucial in treating the patient. Therefore, correct, and meaningful segmentation of brain MRI is a challenging task and is required for further processing. This article proposes machine learning based automatic brain MRI segmentation and classification. The pre-processing step is the vital part of the algorithm, where the discrete wavelet transforms (DWT) and median filtering help in identifying and pointing the exact location of the tumor. The preprocessed image is further segmented by an improved original Fuzzy C-means (FCM) clustering technique. The feature extraction and classification is performed by support vector machine (SVM) classifier. It is found that the simulation associated with ground truth data provides better segmentation results in terms of accuracy, sensitivity, and dice coefficient.

Myocardial salvage and infarct characteristics in patients with ST-elevation myocardial infarction classified by ORBI risk score

Abstract Background The Observatoire Régional Breton sur l'Infarctus (ORBI) [1] risk score was developed to predict in-hospital cardiogenic shock (CS) development in patients with ST elevation myocardial infarction (STEMI) treated with primary percutaneous coronary intervention. The association between infarct characteristics determined by cardiac magnetic resonance imaging (CMR) and ORBI risk classification have not previously been reported in patients with STEMI. Methods The Third DANish Study of Optimal Acute Treatment of Patients With ST-Segment Elevation Myocardial Infarction (DANAMI 3) trial investigated multiple revascularization strategies in patients with STEMI. CS on admission was an exclusion criterion, thus only low and intermediate scores were included. Stratification into ‘low’ (0-7) and ‘intermediate-high’ (>=8) ORBI risk score was performed (n=1,396), and CMR data from admission and three months follow-up at the largest participating centre were analysed using logistic- and Cox-regression (n=534, 38%). Results A total of 1,149 (82%) were ‘low’ risk, and 247 (18%) ‘intermediate-high’ risk. Patients with ‘intermediate-high’ risk were older (median age 71 (interquartile range (IQR) 61-78) vs. 60 (52-68), p<0.001), and more had an admission systolic blood pressure < 90 mm Hg (4.9% vs. 1.4%, p=0.002), and heart rate > 100 bpm (30% vs. 5%, p<0.001). Left ventricular ejection fraction was lower at discharge (40% (30-50%) vs. 50% (45-55%), p<0.001), and more had an anterior infarction (61% vs. 38%, p<0.001) compared to ‘low’ risk patients. Moreover, median peak troponin was higher (5010 ng/L (2195-9930) vs. 2820 ng/L (1138-5805), p<0.001). Aside from hypertension (52% vs. 39%, p<0.001), there were no differences in factors not covered by the ORBI score. In patients with ‘intermediate-high’ risk, both median acute and final infarct size were larger compared to those with ‘low’ risk (21% (12-28%) vs. 15% (7-24%), p=0.004, and 15% (10-26%) vs. 10% (4-17%), p<0.001). Median final salvage index was lower (0.66 (0.47-0.76) vs. 0.69 (0.55-0.87), p=0.048). ‘Intermediate-high’ risk was associated with higher odds of an above-median acute infarct size (odds ratio (OR) 2.02 (95% confidence interval (CI) 1.13-3.68) p=0.019), final infarct size (OR 3.00 (1.65-5.37), p<0.001), and lower odds of an above-median final salvage index (OR 0.52 (0.29-0.91), p=0.026). One-year mortality was higher in patients with ‘intermediate-high’ (9.7% (95%CI 5.9-13)) compared to "low" risk patients (1.9% (1.1-2.7), p<0.001). In univariate analyses, ‘intermediate-high’ risk was associated with a higher mortality risk (hazard ratio 5.31 (95%CI 2.98-9.48), p<0.001). Conclusions In patients with STEMI, an intermediate-high ORBI risk score of >=8 was associated with larger infarct size and less myocardial salvage, and a 5-time higher mortality risk relative to the overall favorable outcomes in those with a low-risk score.Kaplan-Meier Plot, one-year survival

MRI of Endometriosis: Single institute experience including MRI classification and Laparoscopic correlation with ENZIAN classification

BackgroundEndometriosis is a chronic gynecological condition affecting women of reproductive age and may cause pelvic pain and infertility. MRI is the best imaging technique for mapping and preoperative staging of endometriosis as it provides a tool for the assessment of deep infiltrating endometriosis. The aim of this study was to evaluate the role of magnetic resonance imaging (MRI) in detection and characterization of endometriosis in correlation between it and the laparoscopic findings according to ENZIAN classification.MethodsWe assessed 100 patients by MRI dedicated pelvis protocol for endometriosis with an assessment of the ovaries, uterus, uterosacral ligaments (USL), broad ligaments, and round ligaments, cul-de-sac, rectosigmoid colon, bladder, ureters, and rectovaginal septum. All these patients are submitted for laparoscopy, and then, data from both investigations were correlated.ResultsMRI provided a road map for surgeons before laparoscopy. The sensitivity of MRI was 87.5% in small deep infiltrating endometriosis (DIE) (grade A1) and 100% in large DIE (A2). MRI detected 82.3% of uterosacral ligament involvement, specificity 80%. Detection rate in rectal involvement was 71.4%, specificity 75%. There was 100% sensitivity and specificity in adenomyosis and urinary bladder endometriosis.ConclusionsPreoperative detection of endometriotic lesions is recommended using a dedicated MRI endometriosis protocol to choose the surgical approach and for proper multidisciplinary teamwork planning.

Transfer‐learning‐based classification of pathological brain magnetic resonance images

AbstractDifferent diseases occur in the brain. For instance, hereditary and progressive diseases affect and degenerate the white matter. Although addressing, diagnosing, and treating complex abnormalities in the brain is challenging, different strategies have been presented with significant advances in medical research. With state‐of‐art developments in artificial intelligence, new techniques are being applied to brain magnetic resonance images. Deep learning has been recently used for the segmentation and classification of brain images. In this study, we classified normal and pathological brain images using pretrained deep models through transfer learning. The EfficientNet‐B5 model reached the highest accuracy of 98.39% on real data, 91.96% on augmented data, and 100% on pathological data. To verify the reliability of the model, fivefold cross‐validation and a two‐tier cross‐test were applied. The results suggest that the proposed method performs reasonably on the classification of brain magnetic resonance images.

Robust clinical applicable CNN and U-Net based algorithm for MRI classification and segmentation for brain tumor

Early diagnosis of brain tumors is critical for enhancing patient prognosis and treatment options, while accurate classification and segmentation of brain tumors are vital for developing personalized treatment strategies. Despite the widespread use of Magnetic Resonance Imaging (MRI) for brain examination and advances in AI-based detection methods, building an accurate and efficient model for detecting and categorizing tumors from MRI images remains a challenge. To address this problem, we proposed a deep Convolutional Neural Network (CNN)-based architecture for automatic brain image classification into four classes and a U-Net-based segmentation model. Using six benchmarked datasets, we tested the classification model and trained the segmentation model, enabling side-by-side comparison of the impact of segmentation on tumor classification in brain MRI images. We also evaluated two classification methods based on accuracy, recall, precision, and AUC. Our developed novel deep learning-based model for brain tumor classification and segmentation outperforms existing pre-trained models across all six datasets. The results demonstrate that our classification model achieved the highest accuracy of 98.7% in a merged dataset and 98.8% with the segmentation approach, with the highest classification accuracy reaching 97.7% among the four individual datasets. Thus, this novel framework could be applicable in clinics for the automatic identification and segmentation of brain tumors utilizing MRI scan input images.

Brain tumour MRI detection and classification based on the convolutional neural network

Magnetic Resonance Imaging (MRI) has emerged as a widely used diagnostic technique for brain tumour detection. However, the diagnosis of brain tumours poses significant challenges due to their occurrence in diverse locations and various types. Furthermore, MRI generates images that require manual analysis by physicians, which can be laborious and prone to errors. To enhance the efficacy and accuracy of brain tumour detection, recent advances in artificial intelligence have led to the development of machine learning algorithms. In this study, a convolutional neural network (CNN) based method was proposed for brain tumour detection and classification through the preprocessing of raw MRI images. The customized CNN model achieves an accuracy of 98% on a dataset consisting of four types of MRI images, including three types of brain tumours and healthy brain images, with preprocessing applied to all images. The CNN model demonstrates an accuracy of 95% in classifying raw MRI images from the dataset. The CNN model's performance is further improved by training the model with preprocessed images that have been transformed into the same colour space and object area zoomed in. These findings provide a promising avenue for the development of automated and efficient brain tumour detection systems using CNN and MRI.

Multi-class classification of brain tumour magnetic resonance images using multi-branch network with inception block and five-fold cross validation deep learning framework

The expertise of radiologists plays a pivotal role in the intricate task of diagnosing brain tumors. However, the escalating number of patients has rendered traditional diagnostic methods increasingly formidable and expensive. Consequently, researchers have diligently delved into the realm of medical image analysis to seek more efficient and precise approaches for the detection and classification of brain malignancies. In recent times, the surge in interest surrounding Deep Learning (DL) techniques has been notable, particularly in the development of automated systems with the capability to swiftly and reliably categorize brain tumors. The essence of this study revolves around the endeavor to confront the technical challenges inherent in this domain, with a specific focus on the intricate task of classifying brain tumors through the analysis of magnetic resonance images (MRI). The crux of the proposed model for brain tumor classification rests upon the foundation of Convolutional Neural Networks (CNNs), further fortified by a multi-branch network enriched with an inception block, elegantly incorporating a five-fold cross-validation methodology. This comprehensive approach was meticulously calibrated and evaluated using the Br35H dataset, a repository encompassing 395 images devoid of tumors, 826 images depicting gliomas, 822 portraying meningiomas, and 827 capturing pituitary tumors. This research diligently concentrates on discerning among three tumor types, including scenarios where no tumors are present, and the outcomes are meticulously juxtaposed against an array of cutting-edge classification models, such as the Support Vector Classifier (SVC), Decision Tree Classifier, Logistic Regression, and Gaussian Naive Bayes. To provide a thorough assessment of the proposed model's prowess, a battery of performance metrics was subjected to intense scrutiny, spanning accuracy, loss, and variance. The meticulous analysis culminated in the revelation that the CNN with Multi-Branch Network and Inception block exhibited unparalleled excellence, achieving an awe-inspiring accuracy rate of 99.30% and a scarcely perceptible variance of 0.0005 in its classification of the four distinct tumor classes, firmly establishing its supremacy over the existing state-of-the-art models in the field.