Detection of Brain Tumor Using CNN Algorithm

Akhilesh Kumar, Ankush Singh , Himanshu Singh

Introduction: Detection of Brain Tumor Application: Voice Assistant Speech To Text Recognition for Virtual Asssitant Challenges in of Brain Cancer Detection using ML
□Users can speak naturally to their Windows virtual assistant, □ Data Availability: Access to high-quality, voice datasets is
A brain tumor denotes an accumulation of aberrant brain cells, Voice based Virtual assistant used in a wide area of issuing commands or asking questions, and the assistant will essential for training
categorized as either malignant or benign, with the potential to applications like: transcribe and interpret the spoken words. □ the machine learning model, and data collection and curation
manifest in any region of the brain or its adjacent tissues, such as the □ Smart Home Integration □ It is used to understand the voice of computer ad convert them may pose challenges.
meninges, pituitary gland, or cranial nerves. Traditionally, a □ Entertainment and Media to perform any required task. □ Model Integration: The machine learning model should be
combination of clinical examination and various medical imaging □ Office and Business Applications seamlessly integrated into the web
modalities is employed for brain tumor detection, including MRI, CT,
□ Education □ application, requiring careful design for efficient interaction.
Biopsy, PET (Positron Emission Tomography), and Neurological
Examination. Medical professionals utilize a grading system to □ Healthcare □ Model Training and Updates: Procedures for
communicate effectively within interdisciplinary teams, strategize □ Development and Programming training,
versioning, and updating the model
treatments, and prognosticate outcomes. The grading scale, ranging □ must be established to maintain accuracy and adapt to new data.
from 1 to 4, assesses the level of cancerousness in tumor cells. Grade □ Interpretability: In the medical field, model interpretability and
1 tumors, the least malignant, are associated with a favorable long-
Methodology: Detection of Brain Tumor explainability are crucial for
term prognosis. Notably, tumors in this grade, such as pilocytic
□ user trust and understanding
astrocytoma, may be effectively treated with surgery alone due to
their moderate growth rate and nearly normal cellular structure. Fig:6 Speech Recognition
Grade 2 tumors exhibit slight aberrations in cellular appearance and a Conclusion
relatively slow growth rate. Some have the potential to infiltrate
neighboring normal tissue and progress to higher grade tumors. □ The study presents a comprehensive methodology for brain tumor
Examples include grade 2 oligodendroglioma and grade 2 detection using Convolutional Neural Networks (CNN) on MRI
astrocytoma. Grade 3 tumors involve active cellular division, images. The proposed approach involves three main stages: image
generating aberrant cells that invade surrounding healthy brain tissue. acquisition, pre-processing, and post-processing. Image
These tumors have a propensity to recur or progress to grade 4 acquisition involves obtaining MRI images from various sources,
tumors. Grade 4 brain tumors pose the highest level of danger, while pre-processing includes converting images to grayscale and
characterized by rapid proliferation, distinct cellular abnormalities removing noise using a median filter. Post-processing involves
under the microscope, and invasive growth into surrounding healthy using CNN for segmentation and morphological operations for
brain tissue. Glioblastoma, a common grade 4 tumor, exhibits rapid feature extraction.
growth supported by the development of new blood vessels. In this
context, we propose the utilization of Convolutional Neural Network □ In conclusion, the proposed methodology demonstrates a
(CNN) algorithms for brain tumor segmentation in MRI images. Fig.7 System Design systematic approach to brain tumor detection using CNNs,
CNNs, with their advanced deep learning capabilities, offer superior highlighting the importance and effectiveness of CNNs in medical
segmentation precision compared to traditional methods, potentially image analysis. The study contributes to the field of medical
enhancing the accuracy and efficiency of brain tumor identification Fig. 2 Flowchar imaging by providing a detailed framework for accurate And
and classification. Convolution Nural Network (CNN): efficient brain tumor detection.
In the area of medical image processing, the Neural Network is Research Gap, Objectives, Motivation & Challenges
commonly used. Many researchers have tried to develop a model
References
over the years that can more accurately recognize the tumor. To Research Gap
□ Limited Data: Need larger and more diverse datasets for 1.Gaonkar, O. M., Jadhav, N. S., Koli, D. K., & Bhosale, P. V.
determine the utility of the proposed brain tumor classification (2022). Brain Tumour Diagnosis Using Matlab with Edge
system, training accuracy, validation accuracy, and validation loss are training. Detection. Int. J. Res. Appl. Sci. Eng. Technol, 10(5), 122-128.
measured. We attempted to come up with an example that could □ Handling Imbalanced Data: Address issues with unequal classes
2. Zulkoffli, Z., & Shariff, T. A. (2019, June). Detection of brain
correctly identify the tumor from 3D images of the brain MRI. This for accurate results. tumor and extraction of features in MRI images using K-means
tumor can be identified by a fully-connected neural network, but we □ Lack of Pre-Trained Models: Training a model requires a lot of clustering and morphological operations. In 2019 IEEE international
adopted CNN for our model because of parameter sharing and time and resources, having pre-trained weights helps with conference on automatic control and intelligent systems
connection sparsity. (I2CACIS) (pp. 1-5). IEEE.
achieving better accuracy.
Objective 3. Handore, S., & Kokare, D. (2015, January). Performance analysis
of various methods of tumour detection. In 2015 International
□ Seamless Interaction:- Enable users to interact with their Conference on Pervasive Computing (ICPC) (pp. 1-4). IEEE.
Fig.8.Working of CNN model for brain tumor detection Windows devices naturally using voice commands, reducing the 4. Kaur, N., & Sharma, M. (2017, August). Brain tumor detection
reliance on manual input methods using self-adaptive K-means clustering. In 2017 International
□ Task Automation :-Automate routine tasks and commands to Conference on Energy, Communication, Data Analytics and Soft
Computing (ICECDS) (pp. 1861-1865). IEEE.
streamline user workflows and improve overall efficiency.
Dataset:
5. Arunmozhi, S., Sivagurunathan, G., Meenakshi, P. K., Karishma,
S., & Rajinikanth, V. (2020, July). A study on brain tumor extraction
Motivation using various segmentation techniques. In 2020 International
The dataset has 556 images with different types of tumor and Conference on System, Computation, Automation and Networking
□ Voice-based interfaces cater to users with varying physical
also including images which has tissues of Fat or water. (ICSCAN) (pp. 1-4). IEEE.
abilities, making technology more accessible
1. DICOM Samples Image Sets, http://www.osirix-viewer.com/.
□ Voice-based virtual assistants empower elderly individuals to 6. Hazra, A., Dey, A., Gupta, S. K., & Ansari, M. A. (2017, August).
[3] Brain tumor detection based on segmentation using MATLAB.
2. “Brainweb:SimulatedBrainDatabase,” engage with technology. In 2017 International Conference on Energy, Communication, Data
http://brainweb.bic.mni.mcgill.ca/cgi/brainweb1.[4] Fig.3 Convolution Neural Network Architecture □ Voice-based interfaces cater to linguistic diversity. Analytics and Soft Computing (ICECDS) (pp. 425-430). IEEE.
□ Voice-based virtual assistants promote a digital-first mindset.

RESEARCH POSTER PRESENTATION DESIGN © 2019

www.PosterPresentations.com