From Images to Insights: Transforming Brain Cancer Diagnosis with Explainable AI
Md. Arafat Alam Khandaker 1 , Ziyan Shirin Raha 1 , Salehin Bin Iqbal 1 , M.F. Mridha 2 , Jungpil Shin 3 · 2025
Read From Images to Insights: Transforming Brain Cancer Diagnosis with Explainable AI by Md. Arafat Alam Khandaker 1 , Ziyan Shirin Raha 1 , Salehin Bin Iqbal 1 , M.F. Mridha 2 , Jungpil Shin 3 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-05426.txt.
Introduction
Arafat Alam Khandaker and colleagues transform brain cancer diagnosis with explainable AI. Their 2025 arXiv paper introduces a Bangladeshi MRI dataset for the task. Brain cancer detection must be fast and precise to guide treatment. Radiologist skill varies, and reads can be slow and inconsistent. The dataset holds over six thousand MRI images in three classes.
Images came from several hospitals across Bangladesh. Classes cover brain tumor, glioma, and meningioma cases. Several deep learning models were trained on the data. One model reached near-perfect scores on all main metrics. Explainable methods then visualized why models chose each class. Heat maps highlight the image regions behind decisions. The work pairs high accuracy with clinician-friendly transparency.
Catalog record
| CREATOR | Md. Arafat Alam Khandaker 1 , Ziyan Shirin Raha 1 , Salehin Bin Iqbal 1 , M.F. Mridha 2 , Jungpil Shin 3 |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 21 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-05426
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
From Images to Insights: Transforming Brain Cancer Diagnosis with Explainable AI Md. Arafat Alam Khandaker 1 , Ziyan Shirin Raha 1 , Salehin Bin Iqbal 1 , M.F. Mridha 2 , Jungpil Shin 3 Abstract Brain cancer represents a major challenge in medical diagnostics, requisite precise and timely detection for effective treatment. Diagnosis initially relies on the proficiency of radiologists, which can cause difficulties and threats when the expertise is sparse. Despite the use of imaging resources, brain cancer remains often difficult, time-consuming, and vulnerable to intraclass variability. This study conveys the Bangladesh Brain Cancer MRI Dataset, containing 6,056 MRI images organized into three categories: Brain Tumor, Brain Glioma, and Brain Menin. The dataset was collected from several hospitals in Bangladesh, providing a diverse and realistic sample for research. We implemented advanced deep learning models, and DenseNet169 achieved exceptional results, with accuracy, precision, recall, and F1-Score all reaching 0.9983. In addition, Explainable AI (XAI) methods including GradCAM, GradCAM++, ScoreCAM, and LayerCAM were employed to provide visual representations of the decision-making processes of the models. In the context of brain cancer, these techniques highlight DenseNet169’s potential to enhance diagnostic accuracy while simultaneously offering transparency, facilitating early diagnosis and better patient outcomes. [email protected], [email protected], [email protected], [email protected], [email protected] I Introduction Brain cancer remains one of the most challenging and critical medical conditions having a wide variety of tumor types that poses a lot of challenges in medical diagnosis due to the complexity and the need for precise detection. Despite being advanced in imaging technology like Magnetic Resonance Imaging (MRI), the diagnosis of brain cancer remains challenging. The traditional methodology, which is the reliance on the radiologists’ expertise can sometimes be hindered by factors such as tumor diversity, limitations of human interpretation, particularly in regions with limited specialized resources [ 1 ] . This scenario is even worsened by the need for timely, detailed and accurate classification of the cancer which is crucial for effective treatment planning and patient management [ 2 ] . Recent studies show that brain tumors, such as meningiomas, gliomas and other malignancies have significant intraclass variability making accurate identification challenging [ 3 ] . As a result, there is an increasing demand for advanced computational tools to improve diagnosis accuracy and assist radiologists in their decision-making processes. Deep learning (DL) approaches have shown great potential for improving medical image analysis by automating tumor identification and classification [ 4 ] . Convolutional Neural Networks (CNNs), a subset of DL, have been successfully applied to various imaging tasks, including brain tumor classification, by learning complex features from large datasets [ 2 ] [ 5 ] . Our study is conducted on the Bangladesh Brain Cancer MRI Dataset [ 6 ] , a comprehensive collection of 6,056 MRI images categorized into three distinct brain cancer types. As DL models are quite powerful, they frequently function as “black boxes,” which makes it challenging to understand how they make decisions. This may make it more difficult for users to accept and trust these models, particularly in crucial medical applications where knowing the reasoning behind a diagnosis is crucial [ 7 ] [ 8 ] . To address these challenges and enhance the reliability and transparency of our models, we employed Explainable AI (XAI) techniques [ 9 ] , including GradCAM, GradCAM++, ScoreCAM, and LayerCAM. These methods provide visual explanations of the model’s decision-making process, revealing how predictions are made and solving the ‘black box’ problem. The research paper’s contributions are summarized below: • This study investigated nine pre-trained CNN models (DenseNet121, DenseNet169, DenseNet201, ResNet50, ResNet101, ResNet152, MobileNetV3, Xception, and InceptionV3) for the automated categorization of brain cancer in Brain MRI images. • In order to boost interpretability, explainable AI approaches, such as Grad-CAM, Grad-CAM++, Score-CAM, and Layer CAM have been implemented. • Our research aims to address the gap in explainable AI methods for the classification of brain cancer in Brain MRI images. This study begins with an introduction in Section I , followed by a review of related works in Section II . The background study is detailed in Section III , methodology is detailed in Section IV , results are analyzed in Section V , and limitations with future work are discussed in Section VI . Finally, Section VII summarizes the study. II Related Works This section gives a compact overview of previous study on Brain Tumor Classification which is relevant to our study. II-A Brain Tumor Classification without using XAI Techniques Vinod Kumar et al. [ 4 ] , Saleh et al. [ 10 ] , Heba et al. [ 11 ] , and Sneha et al. [ 13 ] each proposed methods for detecting brain tumors using various deep learning models [ 12 ] , without integrating Explainable AI (XAI) techniques. Vinod Kumar et al. applied transfer learning using CNNs such as AlexNet, VGG-16, and ResNet-50, and introduced a hybrid model combining VGG-16 and ResNet-50. Their approach, validated on a Kaggle dataset containing 3,264 MRI images, achieved impressive accuracy, sensitivity, and specificity of 99.98%. Saleh et al. explored the classification of brain tumors by leveraging pre-trained CNNs- Xception, ResNet50, InceptionV3, VGG16, and MobileNet on a Kaggle dataset of 4,480 MRI images, resulting in F1-scores between 97.25% and 98.75%. Heba et al. suggested a technique that combines a Deep Neural Network (DNN) with Discrete Wavelet Transform (DWT) for feature extraction and uses Principal Component Analysis (PCA) for feature reduction. This method, when tested on 66 MRI images from Harvard Medical School, yields a classification rate of 96.97%. Lastly, Sneha et al. compared a basic CNN with VGG-16 architecture for brain tumor detection, training on an augmented dataset of 2,065 MRI images. Their basic CNN achieved 93.36% training accuracy and 86.45% validation accuracy, while VGG-16 reached 97.16% training accuracy and 97.42% validation accuracy. II-B Brain Tumor Classification using XAI Techniques Fahad et al. [ 8 ] , Ramy A et al. [ 14 ] , Eman Ragab et al. [ 2 ] and Burak et al. [ 16 ] each proposed methods for detecting brain tumors using various deep learning models incorporating XAI techniques. Fahad et al. used the VGG-16 model with transfer learning and Layer-wise Relevance Propagation (LRP) to detect brain tumors from 3,000 MRI images, achieving 97.33% accuracy. Ramy A et al. introduced the NeuroXAI framework, which integrated seven XAI techniques, such as GradCAM and SmoothGrad, with CNN models like ResNet-50, to enhance the interpretability of brain tumor MRI analysis. Eman Ragab et al. utilized CNNs, including VGG16, VGG19, and ResNet50, to classify pediatric posterior fossa tumors from a dataset of 300,000 MRI images. VGG16 reached the highest accuracy of 95.33%. Burak et al. presented DGXAINet, which combined DenseNet201 for feature extraction, GradCAM for visualization, and Iterative Neighborhood Component Analysis (INCA) for feature selection, achieving accuracies of 98.65% and 99.97% on two different datasets. Collectively, these studies illustrate how integrating deep learning with XAI can enhance both classification performance and model transparency in brain tumor detection. Fig. 2 : Methodological Framework for Brain Cancer Classification III BackGround Study III-A Convolutional Neural Networks (CNNs) Numerous CNN-based methods for brain cancer detection, including DenseNet [ 17 ] , ResNet [ 18 ] , MobileNet [ 19 ] , Inception [ 20 ]