High‑Performance Computing‑Based Brain Tumor Detection Using

Parallel Quantum Dilated Convolutional Neural Network

Tummuluri Sree Pujitha, Kolli Rohitha, Karnati Nalini, Guttikonda Madhulatha,

B. Sudharani and Jawad Ahmad Dar

Department of ACSE, VFSTR Deemed to be University, Guntur, Andhra Pradesh, India

Keywords: Brain Tumor Detection, Parallel Quantum Dilated Convolutional Neural Networks (PQDCNN),

High‑Performance Computing (HPC), Deep Learning, Fuzzy Local Information C- Means (FLICM), Fast

Retina Keypoint (FREAK), Gray Level Co- Occurrence Matrix (GLCM), Feature Extraction, Medical Image

Processing, BraTS2020, Figshare.

Abstract: We present a high-performance quantum computing-based model for brain tumor detection based on a Parallel

Quantum Dilated Convolutional Neural Network (PQDCNN) framework. It uses Fuzzy Local Information C-

Means (FLICM) clustering for better segmentation, to execute better than the normal K-means. The data

preprocessing processes are median filtering and data augmentation. Deep fusion of FREAK descriptors,

GLCM texture features, and deep CNN representations Our PQDCNN model achieves high classification

performance on both datasets, outperforming state-of-the-art CNNs on BraTS2020 and Figshare datasets,

showing the potential of quantum-inspired deep learning.

1 INTRODUCTION

Brain tumor detection can be a daunting task in

medical diagnostic because of its timeliness and

accuracy. Traditional tumor diagnosis relies heavily

on the manual inspection of magnetic resonance

imaging (MRI) scans, which is an extrapolative and

subjective process associated with inter-observer

variability. With the application of deep learning

specifically, mainly through CNNs, the accuracy of

brain tumor detection has been improved by enabling

automation in identifying and characterizing these

features. However, ordinary CNN models are

computationally rigid since they consume higher

memory and are unable to model long-range spatial

relations in medical images. However, there have

been solutions like quantum-inspired neural

segmentations or high-performance computing

(HPC) frameworks that have emerged to solve these

problems. This paper presents a PQDCN-based brain

tumor detection method. This means that image

pixels are not overlapped and regions through the

image segmentation process can be identified, we

are applying Fuzzy local information c-means

clustering (FLICM), where we use these in the

identification of the image classes. Preprocessing

techniques like median filtering and image

augmentation are employed to enhance the model

generalizability. A hybrid feature representing

tumor scenes is an ensemble of Fast retina key

point (FREAK) descriptors, Gray-level

cooccurrence matrix (GLCM) texture features, and

CNN based deep features. Then, the features obtained

are fed into the PQDCNN model to achieve

multiscale contextual feature extraction with high

inference efficiency using quantum-inspired dilation

strategies.

Utilizing quantum-inspired techniques provides

enhanced analysis of medical images, and quantum

dilated convolutions allow Layer CNN to utilize

receptive field size without losing performance in

comparison to classic versions. This is

advantageous for tumor detection with wide variation

in shape and texture. This model predicts these with

better accuracy by using parallel quantum dilations.

FLICM-based segmentation and multi-feature fusion

complement each other well and provide a potential

solution for automatic brain tumor detection.

The remainder of this paper is as follows: In

Section II, a review of previous work regarding deep

learning of brain tumor detection is provided. Section

III presents the FLICM and the proposed PQDCNN

model. Experimental results, comparison, and

analysis are presented in Section IV.

836

Pujitha, T. S., Rohitha, K., Nalini, K., Madhulatha, G., Sudharani, B. and Dar, J. A.

High-Performance Computing-Based Brain Tumor Detection Using Parallel Quantum Dilated Convolutional Neural Network.

DOI: 10.5220/0013906600004919

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies (ICRDICCT‘25 2025) - Volume 3, pages

836-843

ISBN: 978-989-758-777-1

2 RELATED WORK

The QCNN model achieves 99.67% validation

accuracy and shows excellent generalization in brain

tumor classification. Over the 20 epochs, accuracy

increases, yet distinguishing between benign,

meningioma and malignant; glioma is diffi- cult.

Evaluations using real images demonstrate that it can

be integrated into clinics due to its high accuracy and

robustness against overfitting (Khan et al., 2024).

The accuracy of the optimized YOLOv7 model for

detecting glioma, meningioma, and pituitary tumors

from MRI images was 99.5%. With data

augmentation, detection quality im- proved further,

resulting in 497 correct detections, including three

false positive detections. The model achieves 99.5%

precision and 99.3% recall, outperforming state-of-

the-art techniques, though improvement is necessary

for small and noncircular tumors (Abdusalomov et

al., 2023).

In this work, three hybrid CNN-based high-

accuracy clas- sification models are developed for

brain tumor classification. The first one gets 99.53%

on the accuracy, the second one on the

classification of tumors into five types at 93.81%,

and the last one on gliomas grading at 98.56%.

Optimizing these through grid search and with access

to extensive clinical data allows for these models to

greatly outperform traditional practices in early

detection and diagnosis (Srinivasan et al., 2024).

The 16-layer CNN achieved an impressive

accuracy of 98.88% in binary classification and

97.83% in classifying tumors into three categories

using MRI datasets. By in- corporating hybrid

oversampling, we were able to enhance performance

greatly, outshining traditional machine learning

models like random forest, SVM, and k-NN when it

comes to accuracy, sensitivity, specificity, and F1

score (Singh et al., 2023).

The PDCNN model showed important results,

hitting 97.33% accuracy on dataset-I, 97.60% on

Figshare dataset-

II, and an impressive 98.12% on

Kaggle dataset-III. By integrating two CNNs with

differing window sizes, we were able to enhance

feature extraction, surpassing the performance of

existing methods (Rahman, T., & Islam, M. S. 2023).

The EDN-SVM classifier demonstrated an

impressive accu- racy of 97.93%, with a sensitivity of

92% and specificity of 98 in MRI brain tumor

detection. By using ACEA, median filtering, fuzzy c-

means segmentation, and GLCM, it not only

surpassed traditional methods in terms of precision

but also greatly improved speed, establishing itself as

a strong tool for automated diagnosis (Anantharajan

et al., 2024).

This study dives into CNN-based brain tumor

classification using a dataset of 7,022 MRI images,

exploring models like VGG, ResNet, DenseNet, and

SqueezeNet. DenseNet deliv- ered an impressive

accuracy of 85% when paired with SVM, while a

hybrid model achieved 83% with LDA (Gu¨ler, M.,

& Namlı, E. 2024).

Saeedi and colleagues took a deep dive into using

deep learning for classifying brain tumors based on

3,264 MRI scans. Their 2D CNN model hit an

impressive accuracy of 96.47%, along with a recall

rate of 95%. Meanwhile, the autoencoder performed

admirably as well, achieving 95.63% accuracy and a

94% recall. On the conventional front, K-NN stood

out with an accuracy of 86% (Saeedi et al., 2023).

The A-GRU model, enhanced with ADAM and

data aug- mentation techniques, achieved a

remarkable accuracy of 99.32% in classifying brain

tumors. It outperformed the CNN, A-CNN, LSTM,

A-LSTM, and GRU models. These results were

further improved through careful hyperparameter

tuning (Saboor et al., 2024).

In this study, we explored using YOLOv3 through

YOLOv7 models for classifying meningioma

firmness. Among these, YOLOv7 stood out with

impressive results: a specificity of 97.95%, a

balanced accuracy of 98.97%, and an F1-score of

99.24%. It outperformed both SVM and KNN

techniques (Alhussainan et al., 2024). By analyzing

3,762 MRI images from Kaggle, we found that

ResNet-50 achieved an impressive 99.82%

accuracy during training and 99.5% during testing

when using the SGD opti- mizer. Through

preprocessing, pixel reduction, and optimizing with

binary cross-entropy, we saw a boost in

performance, finally achieving a 96.10% F1-score,

96.50% precision, and 95.62% recall (Asad et al.,

2023).

In this study, we looked at how deep transfer

learning can help diagnose brain tumors using

models like ResNet152, VGG19, DenseNet169, and

MobileNetv3 on a Kaggle dataset. MobileNetv3

stood out with the highest accuracy, hitting

99.75%, while ResNet152 followed closely with

98.5%. (Mathivanan et al., 2024) The research

achieved an average entropy of 7.32 bits, which

helped in reducing saturation effects. It also recorded

a PSNR of 29.07 dB and a contrast level of 39.47 dB,

surpassing earlier techniques like GHE and BBHE.

With the enhanced Inception V3 model, we reached

an impressive accuracy of 98.89%, outperforming

AlexNet, VGG-16, and GoogLeNet in tumor

classification tasks (Agarwal et al., 2024).

High-Performance Computing-Based Brain Tumor Detection Using Parallel Quantum Dilated Convolutional Neural Network

837

This research explores deep learning models for

detecting brain tumors using 3,264 MRI images. The

newly developed CNN achieved an impressive

accuracy of 93.3%, an AUC of 98.43%, a recall of

91.1%, and a loss of 0.260. These results surpass

those of established models like ResNet-50, VGG16,

and Inception V3 (Mahmud et al., 2023).

This study emphasizes the impressive capabilities

of U-Net when it comes to segmenting brain tumors,

particularly show- ing superior outcomes in Dice

score, sensitivity, specificity, and accuracy. Notably,

ACMINet took the top spot on the BraTS2020

leaderboard, which emphasizes how effective U- Net

really is. Between 2020 and 2024, U-Net not only set

new benchmarks but also played an important role in

advancing the diagnostics and treatment of neuro-

oncology (Umarani et al., 2024).

3 METHODOLOGY

In our proposed method, we begin by using FLICM

clustering to effectively segment tumors. After that,

we apply median filtering to reduce any noise, along

with some augmentation techniques aimed at making

our model stronger. For feature extraction, we draw

on a combination of FREAK descriptors, GLCM, and

features derived from deep CNNs to provide a

thorough representation of the data. The PQDCNN

model uses quantum-inspired dilated convolutions to

enhance processing efficiency. We’ve trained our

model on the BraTS2020 and Figshare datasets,

evaluating it based on accuracy, precision, recall, and

F1 score. Plus, by integrating HPC, we ensure that

our approach is scalable and can operate in real time

for medical applications.

3.1 Dataset Description

Datasets The datasets used in this study are medical

images, targeting the detection of brain tumors. We

then split the dataset into training validation and test

sets to train and evaluate the model efficiently

Hence, we utilized the Figshare information suitable

for classification, and we exploited the BraTS2020

information suitable for tumor segmentation to give

an extensive evaluation of our proposed Parallel

Quantum Dilated Convolutional Neural Network

(PQDCNN).

Figshare Dataset:

The Figshare dataset consists

of brain MRI images categorized into three tumor

types:

•

Glioma Tumor

•

Meningioma Tumor

•

Pituitary Tumor

To enhance model performance and consistency, the

dataset undergoes preprocessing, including:

•

Standardizing image dimensions to ensure uniform

input sizes.

•

Normalizing pixel values to improve model

convergence and stability.

•

Applying noise reduction using Median Filtering to

pre- serve tumor structures.

Figure 1: Sample Figshare image dataset.

Table 1: Summary of the Figshare brain tumor MRI dataset.

Attribute Description

Dataset Name

Figshare Brain Tumor MRI

Dataset

Modality Types MRI Images

Target Prediction Brain Tumor Classification

Instances 3,064

Image Dimensions 256 × 256 × 1

Number of Classes 3

Shape of Train Data

Split

(2,451, 256, 256, 1)

Shape of Test Data

Split

(613, 256, 256, 1)

BraTS2020 Dataset: The BraTS2020 dataset includes

multi-modal MRI scans with t1ce and segment

sequences, offering detailed annotations of tumor

regions. This dataset is particularly valuable for

training deep learning models in brain tumor

segmentation. The images in this dataset undergo

preprocessing steps such as:

•

Size standardization for uniform input

representation.

•

Intensity normalization to minimize variations

across different MRI machines.

•

Fuzzy Local Information C-Means (FLICM)

clustering

for effective segmentation of the tumor

region.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

838

Figure 2: Sample BraTS2020 Image dataset.

Table 2: Summary of the BraTS2020 dataset.

Attribute Description

Dataset Name BraTS2020 Dataset

Modality Types MRI (t1ce, segment)

Target Prediction

Brain Tumor

Segmentation

Instances 369

Image

Dimensions

240 × 240 × 155

Number of

Classes

Shape of Train

Data S

lit

(293, 240, 240, 155)

Shape of

Validation Data

Split

(76, 240, 240, 155)

Table 1 gives an overview of the FigShare dataset,

while Table 2 details the BraTS2020 dataset used in

this research. By combining these datasets, we ensure

that the PQDCNN model is trained on diverse, high-

quality medical imaging data, achieving the best

accuracy and reliability in automatically detecting

brain tumors.

3.2 Preprocessing

In this paper, we introduce the preprocessing of brain

magnetic resonance images from our PQDCNN. With

the conversion of detailed MRI images to PQDCNN

and passing validation of pre-trained models, we will

get a very accurate training result. Here’s what we

did, step by step:

Image Segmentation: For Image segmentation

we implemented a clustering algorithm called Fuzzy

Local Information c-Means (FLICM). This approach

successfully maintains local spatial context,

diminishes noise, and promotes better feature

extraction which ultimately contributes to the

improved differentiation of tumour from non-tumour

regions.

Noise Reduction: To reduce noise while

preserving the important tumor structures, we

performed median filtering. This initial step of

processing an image not only improves the quality of

the images but also allows us to extract more relevant

features and increase the judiciousness of our

classification efforts with deep learning up to this

point.

Data Augmentation: We applied several data

augmentations to our model to improve

generalization and decrease overfitting. These

included rotating 15-degree rotation, applied shifts,

shear, zooming in and zooming out, flip the images.

With this, we created realistic variations for the

head’s orientation, shapes and sizes of the tumors,

which allowed the model to learn from a wider variety

of MRI scans.

Normalization: All MRI scans were normalized

to the same intensity range [0,1]. This eliminate

difference in intensity which may occur while

different MRI machines are being used or the

machine is under different settings. This also

normalizes the intensity, helping to stabilize the

training process while reducing internal covariate

shifts in deep networks.

Resize: To have a consistency among our dataset

and to achieve the best computational efficiency in

high- performance computing (HPC) setups, we

applied resize to all MRI scans. This should keep vital

tumor-specific information invariant while still

maintaining everything needed for running the

PQDCNN model (the internal structure of the model

does not have to change) as the large amount of rival

information is destroyed.

3.3 Pre-Trained Model Architectures

In this paper, we introduce an innovative deep

learning architecture designed specifically for

identifying brain tumors. This architecture merges

Parallel Convolutional Neural Net- works (PCNN)

with Quantum Dilated Convolutional Neural

Networks (QDCNN). Our approach, known as the

Parallel Quantum Dilated Convolutional Neural

Network (PQDCNN), emphasizes superior feature

extraction, precise tumor localization, and enhanced

High-Performance Computing-Based Brain Tumor Detection Using Parallel Quantum Dilated Convolutional Neural Network

839

Figure 3: Model Architecture.

classification accuracy. To improve segmentation

efficiency, we use Fuzzy Local Information C- Means

(FLICM) within the model. Besides, we use a

combination of fast retina key point (FREAK)

descriptors, Gray Level Co-occurrence Matrix

(GLCM) features, and deep CNN-based

representations for hybrid feature extraction. By

employing a High-Performance Computing (HPC)

strategy, the PQDCNN architecture optimizes the

analysis of brain MRI scans, resulting in highly

efficient and accurate tumor classification.

•

Parallel Convolutional Neural Network (PCNN):

This model employs parallel streams of convolutional

neural network (CNN) to learn hierarchical features

from brain MRI scans. PCNN enhances feature

diversity and classification robustness by parallel

processing of input data by multiple convolutional

streams. The parallel feature extraction function can

be expressed as:

𝑃𝐶𝑁𝑁



𝑥



∑

𝐶𝑜𝑛𝑣





𝑥





 

(1)

Where Conv

(x) represents the convolution

operation in the i-th parallel branch.

•

Quantum Dilated Convolutional Neural

Network (QDCNN): QDCNN applies quantum-

inspired dilated convolutions to dilate the receptive

field without sacrificing spatial resolution. The

method effectively captures multiscale relations in

brain MRI images, and it results in improved tumor

segmentation and classification. The dilation function

is defined as:

𝑦[𝑖] =

∑

𝑥[𝑖 + 𝑟. 𝑘]





.𝑤[𝑘] (2)

Where x[i] is the input, w[k] is the convolutional

filter, K represents the kernel size, and r denotes the

dilation rate.

•

Parallel Quantum Dilated Convolutional Neural

Net- work (PQDCNN): The PCNN and QDCNN

together form the PQDCNN model, which employs

the power of parallel convolutional feature extraction

and multi-scale dilated convolutions to achieve

highly accurate detection of brain tumors. The

PQDCNN model is defined as:

PQDCNN(x) = PCNN(x)+QDCNN(x) (3)

where the two components operate in harmony to

enhance classification accuracy without

compromising computational speed.

The PQDCNN model enhances brain tumor

detection by cleverly combining parallel

convolutional layers with quantum dilated

convolutions. This powerful integration allows it to

effectively capture both the local and global features

of tumors. As a result, it greatly improves tumor

localization, feature extraction, and classification

accuracy, making PQDCNN a solid choice for

automatic brain tumor detection.

3.4 Fine-Tuning Pre-Trained Models

for Brain Tumor Detection

To effectively detect tumors, we need to fine-tune our

pre- trained models so they can better recognize

patterns in brain MRI images. In this section, we’ll

explore how we fine-tuned the Parallel Quantum

Dilated Convolutional Neural Network (PQDCNN)

to achieve optimal classification performance.

Loading Pre-trained Weights: The PQDCNN

architec- ture combines Parallel Convolutional

Neural Networks (PCNN) with Quantum Dilated

Convolutional Neural Networks (QDCNN). The

PCNN block is responsible for learning hierarchical

multi-scale features, while the QDCNN boosts

feature representation using quantum- inspired

dilated convolutions. We pre-train these modules on

extensive medical imaging datasets to capture univer-

sal spatial and structural patterns associated with

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

840

brain tumors. After pre-training, we fine-tune the

weights on specific datasets like BraTS2020 and

Figshare to enhance classification performance

representations. This process helps the PQDCNN

pick up on MRI-specific textural and spatial patterns

while keeping valuable pre-trained knowledge from

extensive datasets.

Algorithm: Steps Involved in the Proposed Approach

forBrain Tumor Detection

1. Load the brain MRI dataset (BraTS2020 and

Figshare).

2. Apply data partitioning using Fuzzy Local

Information C-Means Clustering (FLICM).

3. Perform data preprocessing using Median

Filtering to remove noise.

4. Apply image augmentation techniques (Rotation,

shifting, shearing, zooming, flipping) to enhance

model generalization.

5. Extract features using:

FREAK with statistical features.

GLCM features.

CNN-based features.

6. Select the proposed deep learning model:

PCNN if model 1 is chosen,

Model =QDCNN if model 2 is chosen,

PQDCNN if model 3 is chosen.

7. Split the dataset into training, validation, and

testing sets.

8. Train the selected model on the training set.

Tune hyperparameters using the validation set to

improve classification performance.

9. Evaluate the trained model on the test set using

performance metrics such as accuracy, precision,

recall, and f1-score.

Freezing Layers: To maintain the valuable feature

rep- resentations we’ve already trained and to prevent

overfit- ting, we start by freezing the entire set of

convolutional layers during the early training phase.

We kick things off by initializing the new fully

connected layers with a higher learning rate,

allowing the model to grasp the specific patterns

present in brain MRI scans. Once we see some

progress, we gradually tune the earlier layers at a

slower rate.

Progressive Unfreezing: After training the output

layer, we gradually unfreeze the deeper layers,

allowing the model to improve its low-level and

mid-level feature representations. This process helps

the PQDCNN pick up on MRI-specific textural and

spatial patterns while keeping valuable pre-trained

knowledge from extensive datasets.

The PQDCNN uses both parallel and

quantum dilated convolutions to effectively

capture the local and global spa- tial

relationships in brain MRI images. This leads to

better accuracy in segmentation and

classification. The fine-tuning strategy we’ve

introduced helps the model generalize well with

fewer chances of overfitting while also ensuring

it runs efficiently on high-performance

computing systems.

4 RESULTS AND DISCUSSIONS

Let’s dive into the outcomes of using the Parallel

Quantum Dilated Convolutional Neural Network

(PQDCNN) for brain tumor classification. This

model employs FLICM for data splitting and

incorporates various feature extraction techniques like

FREAK descriptors, GLCM features, CNN-based

embed- dings, and of course, PQDCNN itself for an

accurate diagnosis of brain tumors. We evaluate how

well PQDCNN performs in classifying these tumors

and explore whether applying Progressive Unfreezing

in transfer learning can enhance its effectiveness.

4.1 Performance Comparisons

Figshare Dataset: PQDCNN outperforms both PCNN

and QDCNN across all metrics, hitting an impressive

accuracy of 93.53%. It also achieves precision at

93.48%, recall at 93.53%, and an F1-score of 93.46%,

displaying its effective- ness in tumor classification.

Besides, its ROC AUC score of 0.9924 emphasizes

its superior ability to distinguish between different

outcomes. While QDCNN does surpass PCNN in

precision (88.38%) and F1-score (88.34%), it still

lags when

compared to PQDCNN. These results

underline the major advantages of quantum

dilation and parallel computation in medical

imaging. Table 3 gives the dataset comparison of

Figshare while table 4 gives the comparison of

BraTS2020 dataset.

Table 3: Comparing PCNN, QDCNN, and PQDCNN

models for Figshare dataset.

Metric PCN

QDCNN PQDCNN

Accurac

84.39% 88.64% 93.53%

Precision 85.44% 88.38% 93.48%

Recall 84.39% 88.664% 93.53%

F1 Score 84.75% 88.34% 93.46%

ROC AUC 0.9551 0.9753 0.9924

High-Performance Computing-Based Brain Tumor Detection Using Parallel Quantum Dilated Convolutional Neural Network

841

Figure 4: PCNN, QDCNN and PQDCNN confusion

matrices.

Figure 5: ROC Curve.

BraTS2020 Dataset: When we look at how the three

models perform, PCNN, QDCNN, and PQDCNN, it

turns out that all have an accuracy of 75%. However,

PQDCNN stands out with a better recall rate (75%)

and an F1-score of 70.59%, suggesting it strikes a

better balance between precision and recall. While

PCNN and QDCNN are a tad more precise (71.43%),

their lower recall (62.5%) brings down their F1- score

(66.67%). This tells us that PQDCNN does a better

job of identifying positive cases, making it the

preferred model when recall is a key factor.

Table 4: Comparing PCNN, QDCNN, and PQDCNN

models for BraTS2020 Dataset.

Metric PCN

QDCNN PQDCNN

Accurac

84.39% 88.64% 93.53%

Precision 85.44% 88.38% 93.48%

Recall 84.39% 88.664% 93.53%

F1 Score 84.75% 88.34% 93.46%

ROC AUC 0.9551 0.9753 0.9924

Figure 6: PCNN, QDCNN and PQDCNN Confusion

Matrices.

Figure 7: ROC curve.

5 CONCLUSIONS

When comparing the performance of the PCNN,

QCNN, and PQDCNN models on two datasets,

Figshare and BraTS2020, it is clear that PQDCNN

stands out consistently. For the Figshare dataset,

PQDCNN achieves the highest accu- racy at 93.53%

and excels in other metrics as well, including

precision (93.74%), recall (93.53%), F1-score

(93.46%), and ROC AUC (0.9924). Thus, it emerges

as the top performer for this dataset. Similarly, in the

case of the BraTS2020 dataset, all models show

identical accuracy at 75%; however, PQDCNN has a

higher recall rate at 75% and a better F1-score of

70.58%. This points to a more favorable balance

between precision and recall. Overall, these findings

suggest that PQDCNN is the strongest model,

particularly in scenarios where recall and F1-score

play critical roles in classification performance.

REFERENCES

Abdusalomov, A. B., Mukhiddinov, M., & Whangbo, T. K.

(2023). Brain tumor detection based on deep learning

approaches and magnetic resonance imaging. Cancers,

15(16), 4172.

Agarwal, M., Rani, G., Kumar, A., Kumar, P., Manikandan,

R., & Gandomi, A. H. (2024). Deep learning for

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

842

enhanced brain tumor detection and classification.

Results in Engineering, 22, 102117.

Alhussainan, N. F., Ben Youssef, B., & Ben Ismail, M. M.

(2024). A deep learning approach for brain tumor

firmness detection based on five different YOLO

Versions: YOLOv3–YOLOv7. Computation, 12(3), 44.

Anantharajan, S., Gunasekaran, S., Subramanian, T., &

Venkatesh, R. (2024). MRI brain tumor detection using

deep learning and machine learning approaches.

Measurement: Sensors, 31, 101026.

Asad, R., Rehman, S. U., Imran, A., Li, J., Almuhaimeed,

A., & Alzahrani, A. (2023). Computer-aided early

melanoma brain-tumor detection using deep-learning

approach. Biomedicines, 11(1), 184.

Gu¨ler, M., & Namlı, E. (2024). Brain tumor detection with

deep learning methods’ classifier optimization using

medical images. Applied Sciences, 14(2), 642.

Khan, M. A. Z., Innan, N., Galib, A. A. O., & Bennai, M.

(2024). Brain tumor diagnosis using quantum

convolutional neural networks. arXiv preprint

arXiv:2401.15804.

Mahmud, M. I., Mamun, M., & Abdelgawad, A. (2023). A

deep analysis of brain tumor detection from mr images

using deep learning networks. Algorithms, 16(4), 176.

Mathivanan, S. K., Sonaimuthu, S., Murugesan, S.,

Rajadurai, H., Shivahare, B. D., & Shah, M. A. (2024).

Employing deep learning and transfer learning for

accurate brain tumor detection. Scientific Reports,

14(1), 7232.

Rahman, T., & Islam, M. S. (2023). MRI brain tumor

detection and classification using parallel deep

convolutional neural networks. Measurement: Sensors,

26, 100694.

Saboor, A., Li, J. P., Ul Haq, A., Shehzad, U., Khan, S.,

Aotaibi, R. M., & Alajlan, S. A. (2024). DDFC: deep

learning approach for deep feature extraction and

classification of brain tumors using magnetic resonance

imaging in E-healthcare system. Scientific reports,

14(1), 6425.

Saeedi, S., Rezayi, S., Keshavarz, H., & R. Niakan Kalhori,

S. (2023). MRI-based brain tumor detection using

convolutional deep learning methods and chosen

machine learning techniques. BMC Medical In-

formatics and Decision Making, 23(1), 16.

Singh, K., Kaur, A., & Kaur, P. (2023). Computer Aided

Detection of Brain Tumors using Convolutional Neural

Network based Analysis of MRI Data.

Srinivasan, S., Francis, D., Mathivanan, S. K., Rajadurai,

H., Shivahare, D., & Shah, M. A. (2024). A hybrid deep

CNN model for brain tumor image multi-classification.

BMC Medical Imaging, 24(1), 21.

Umarani, C. M., Gollagi, S. G., Allagi, S., Sambrekar, K.,

& Ankali, S. (2024). Advancements in deep learning

techniques for brain tumor segmentation: a survey.

Informatics in Medicine Unlocked, 101576.

High-Performance Computing-Based Brain Tumor Detection Using Parallel Quantum Dilated Convolutional Neural Network

843