CNN-TumorNet: leveraging explainability in deep learning for precise brain tumor diagnosis on MRI images

Frontiers in Oncology, Journal Year: 2025, Volume and Issue: 15

Published: March 26, 2025

Introduction The early identification of brain tumors is essential for optimal treatment and patient prognosis. Advancements in MRI technology have markedly enhanced tumor detection yet necessitate accurate classification appropriate therapeutic approaches. This underscores the necessity sophisticated diagnostic instruments that are precise comprehensible to healthcare practitioners. Methods Our research presents CNN-TumorNet, a convolutional neural network categorizing images into non-tumor categories. Although deep learning models exhibit great accuracy, their complexity frequently restricts clinical application due inadequate interpretability. To address this, we employed LIME technique, augmenting model transparency offering explicit insights its decision-making process. Results CNN-TumorNet attained 99% accuracy rate differentiating from scans, underscoring reliability efficacy as instrument. Incorporating guarantees model’s judgments comprehensible, enhancing adoption. Discussion Despite overarching challenge interpretability persists. These may function ”black boxes,” complicating doctors’ ability trust accept them without comprehending rationale. By integrating LIME, achieves elevated alongside transparency, facilitating environments improving care neuro-oncology.

Language: Английский

---

A Critical Review on Segmentation of Glioma Brain Tumor and Prediction of Overall Survival

Archives of Computational Methods in Engineering, Journal Year: 2024, Volume and Issue: unknown

Published: Oct. 17, 2024

Language: Английский

---

Predicting hepatocellular carcinoma survival with artificial intelligence

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: Feb. 20, 2025

Despite the extensive research on hepatocellular carcinoma (HCC) exploring various treatment strategies, survival outcomes have remained unsatisfactory. The aim of this was to evaluate ability machine learning (ML) methods in predicting probability HCC patients. study retrospectively analyzed cases patients with stage 1-4 HCC. Demographic, clinical, pathological, and laboratory data served as input variables. researchers employed feature selection techniques identify key predictors patient mortality. Additionally, utilized a range model rates. included 393 individuals For early-stage (stages 1-2), models reached recall values ​​of up 91% for 6-month prediction. advanced-stage (stage 4), achieved accuracy 92% 3-year overall To predict whether are ex or not, 87.5% when using all 28 features without best performance coming from implementation weighted KNN. Further improvements accuracy, reaching 87.8%, were by applying medium Gaussian SVM. This demonstrates that can reliably probabilities across disease stages. also shows AI accurately high proportion surviving assessing clinical pathological factors.

Language: Английский

---

Explainable AI analysis for smog rating prediction

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: March 8, 2025

Smog poses a direct threat to human health and the environment. Addressing this issue requires understanding how smog is formed. While major contributors include industries, fossil fuels, crop burning, ammonia from fertilizers, vehicles play significant role. Individually, vehicle's contribution may be small, but collectively, vast number of has substantial impact. Manually assessing each vehicle impractical. However, advancements in machine learning make it possible quantify contribution. By creating dataset with features such as model, year, fuel consumption (city), type, predictive model can classify based on their impact, rating them scale 1 (poor) 8 (excellent). This study proposes novel approach using Random Forest Explainable Boosting Classifier models, along SMOTE (Synthetic Minority Oversampling Technique), predict individual vehicles. The results outperform previous studies, proposed achieving an accuracy 86%. Key performance metrics Mean Squared Error 0.2269, R-Squared (R2) 0.9624, Absolute 0.2104, Explained Variance Score 0.9625, Max 4.3500. These incorporate explainable AI techniques, both agnostic specific provide clear actionable insights. work represents step forward, was last updated only five months ago, underscoring timeliness relevance research.

Language: Английский

---

FGA-Net: Feature-Gated Attention for Glioma Brain Tumor Segmentation in Volumetric MRI Images

Communications in computer and information science, Journal Year: 2024, Volume and Issue: unknown, P. 66 - 87

Published: Dec. 26, 2024

Language: Английский

---

SE-ResNeXt-50-CNN: A Deep Learning Model for Lung Cancer Classification

Applied Soft Computing, Journal Year: 2025, Volume and Issue: unknown, P. 112696 - 112696

Published: Jan. 1, 2025

Language: Английский

---

Breast Lesion Detection Using Weakly Dependent Customized Features and Machine Learning Models with Explainable Artificial Intelligence

Journal of Imaging, Journal Year: 2025, Volume and Issue: 11(5), P. 135 - 135

Published: April 28, 2025

This research proposes a novel strategy for accurate breast lesion classification that combines explainable artificial intelligence (XAI), machine learning (ML) classifiers, and customized weakly dependent features from ultrasound (BU) images. Two new feature classes are proposed to improve the diagnostic accuracy diversify training data. These based on image intensity variations area of bounded partitions provide complementary rather than overlapping information. ML classifiers such as Random Forest (RF), Extreme Gradient Boosting (XGB), Classifiers (GBC), LASSO regression were trained with both classes. To validate reliability our study results obtained, we conducted statistical analysis using McNemar test. Later, an XAI model was combined tackle influence certain features, constraints selection, interpretability capabilities across various models. LIME (Local Interpretable Model-Agnostic Explanations) SHAP (SHapley Additive exPlanations) models used in process enhance transparency interpretation clinical decision-making. The revealed common relevant malignant class, consistently identified by all benign class. However, observed importance rankings different classifiers. Furthermore, demonstrates correlation between does not impact explainability.

Language: Английский

---

Recent trends on mammogram breast density analysis using deep learning models: neoteric review

Artificial Intelligence Review, Journal Year: 2025, Volume and Issue: 58(8)

Published: May 10, 2025

Abstract Breast cancer is a globally prevalent and potentially fatal illness affecting women. Timely identification of screening mammography may decrease the occurrence incorrect positive results enhance rate patient survival. Nevertheless, density breast tissue in mammograms can impact precision effectiveness detecting cancer. This paper examines existing body research on analysis utilising advanced deep learning models, including convolutional neural networks (CNN), transfer (TL), ensemble (EL). Additionally, it various datasets evaluation measures employed investigations. The study demonstrates that models attain exceptional accuracy categorising density. However, they encounter obstacles such as limited data availability, intricate model structures, difficulties interpreting results. asserts an essential undertaking order to survival rates Further investigation warranted examine most effective augmentation methods, interpretable for this undertaking.

Language: Английский

---

Novel Metaheuristic Algorithms and Their Applications to Efficient Detection of Diabetic Retinopathy

Journal of Artificial Intelligence and Soft Computing Research, Journal Year: 2024, Volume and Issue: 15(2), P. 167 - 195

Published: Dec. 1, 2024

Abstract It is an extremely important to have AI-based system that can assist specialties correctly identify and diagnosis diabetic retinopathy (DR). In this study, we introduce accurate approach for DR using machine learning (ML) techniques a modified golf optimization algorithm (mGOA). The mGOA optimizes ML classifiers through finding the best available parameters with respect objective functions, hence decreases number of features increases classifier’s accuracy. A fitness function employed minimize feature medical dataset. obtained results showed superiority higher convergence speeds without extra processing costs across datasets compared several competitors. Also, attained maximum accuracy optimally reduced in binary multi-class achieving CEC’2022 benchmark other metaheuristic algorithms. Based on findings, three optimized called mGOA-SVM, mGOA-radial SVM,and mGOA-kNN were introduced as tools classification disease their performance was assessed Messidor EyePACS1 datasets. Experimental demonstrated mGOA-SVM SVM achieved remarkable average 98.5% precision 97.4%.

Language: Английский

---