Enhancing Prediction of Brain Tumor Classification Using Images and Numerical Data Features

Diagnostics, Journal Year: 2023, Volume and Issue: 13(15), P. 2544 - 2544

Published: July 31, 2023

Brain tumors, along with other diseases that harm the neurological system, are a significant contributor to global mortality. Early diagnosis plays crucial role in effectively treating brain tumors. To distinguish individuals tumors from those without, this study employs combination of images and data-based features. In initial phase, image dataset is enhanced, followed by application UNet transfer-learning-based model accurately classify patients as either having or being normal. second research utilizes 13 features conjunction voting classifier. The classifier incorporates extracted deep convolutional layers combines stochastic gradient descent logistic regression achieve better classification results. reported accuracy score 0.99 achieved both proposed models shows its superior performance. Also, comparing results supervised learning algorithms state-of-the-art validates

Language: Английский

---

AI-driven estimation of O6 methylguanine-DNA-methyltransferase (MGMT) promoter methylation in glioblastoma patients: a systematic review with bias analysis

Journal of Cancer Research and Clinical Oncology, Journal Year: 2024, Volume and Issue: 150(2)

Published: Jan. 31, 2024

Abstract Background Accurate and non-invasive estimation of MGMT promoter methylation status in glioblastoma (GBM) patients is paramount clinical importance, as it a predictive biomarker associated with improved overall survival (OS). In response to the need, recent studies have focused on development artificial intelligence (AI)-based methods for estimation. this systematic review, we not only delve into technical aspects these AI-driven but also emphasize their profound implications. Specifically, explore potential impact accurate GBM patient care treatment decisions. Methods Employing PRISMA search strategy, identified 33 relevant from reputable databases, including PubMed, ScienceDirect, Google Scholar, IEEE Explore. These were comprehensively assessed using 21 diverse attributes, encompassing factors such types imaging modalities, machine learning (ML) methods, cohort sizes, clear rationales attribute scoring. Subsequently, ranked established cutoff value categorize them low-bias high-bias groups. Results By analyzing 'cumulative plot mean score' 'frequency curve' studies, determined 6.00. A higher score indicated lower risk bias, scoring above mark categorized (73%), while 27% fell category. Conclusion Our findings underscore immense AI-based deep (DL) non-invasively determining status. Importantly, significance advancements lies capacity transform by providing timely information However, translation practice presents challenges, need large multi-institutional cohorts integration data types. Addressing challenges will be critical realizing full AI improving reliability accessibility lowering bias decision-making.

Language: Английский

---

WU-Net++: A novel enhanced Weighted U-Net++ model for brain tumor detection and segmentation from multi-parametric magnetic resonance scans

Multimedia Tools and Applications, Journal Year: 2024, Volume and Issue: 83(28), P. 71885 - 71908

Published: Feb. 8, 2024

Language: Английский

---

External evaluation of a deep learning-based approach for automated brain volumetry in patients with huntington’s disease

Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)

Published: April 22, 2024

Abstract A crucial step in the clinical adaptation of an AI-based tool is external, independent validation. The aim this study was to investigate brain atrophy patients with confirmed, progressed Huntington's disease using a certified software for automated volumetry and compare results manual measurement methods used practice as well volume calculations caudate nuclei based on segmentations. Twenty-two were included retrospectively, consisting eleven nucleus age- sex-matched control group. To quantify head atrophy, frontal horn width intercaudate distance ratio inner table obtained. mdbrain volumetry. Manually measured ratios automatically volumes groups compared two-sample t-tests. Pearson correlation analyses performed. relative difference between manually determined calculated. Both significantly different groups. showed high level agreement mean discrepancy − 2.3 ± 5.5%. group lower variety supratentorial structures. highest degree shown nucleus, putamen, pallidum (all p < .0001). found be strongly correlated both In conclusion, disease, it that correlates commonly practice. allowed clear differentiation collective. additionally allows radiologists more objectively assess involvement structures are less accessible standard semiquantitative methods.

Language: Английский

---

Effect of learning parameters on the performance of the U-Net architecture for cell nuclei segmentation from microscopic cell images

Microscopy, Journal Year: 2022, Volume and Issue: 72(3), P. 249 - 264

Published: Nov. 21, 2022

Nuclei segmentation of cells is the preliminary and essential step pathological image analysis. However, robust accurate cell nuclei challenging due to enormous variability staining, sizes, morphologies, adhesion or overlapping nucleus. The automation process find cell's a giant leap in this direction has an important toward bioimage analysis using software tools. This article extensively analyzes deep U-Net architecture been applied Data Science Bowl dataset segment nuclei. undergoes various preprocessing tasks such as resizing, intensity normalization data augmentation prior segmentation. complete then rigorous training validation optimized hyperparameters model selection. mean (m) ± standard deviation (SD) Intersection over Union (IoU) F1-score (Dice score) have calculated along with accuracy during process, respectively. results IoU 0.94 0.16 (m SD), 0.17 95.54 95.45. With model, we completely independent test cohort obtained IOU 0.93, 0.9311, 94.12, respectively measure performance.

Language: Английский

---