Govind Ballabh Pant Institute of Engineering & Technology (GBPIET),
Ghurdauri PRESENTATION
On
Deep Learning-Based Brain Tumor Detection Using Keras and
Submitted in partial fulfillment of the requirements of the degree of the
TensorFlow
course Mini Project-II (AIP - 102)
Submitted Submitted By:
To: Harshit Saini(Group
Mr V K Leader)
Tamta Uttkarsh Bhatt
Saksham Negi
Akshit Anand
Introduction & Problem
Statement
• Brain tumors: a critical neurological disorder with
significant mortality.
• Early detection vital for improved treatment outcomes
and patient survival.
• Manual MRI analysis: time-consuming, subjective, prone
to human error.
Project
Abstract
This project leverages Keras and TensorFlow frameworks to develop an automated
system for brain tumor detection.
Advanced deep learning techniques enable precise interpretation of MRI images.
Goal: Achieve high accuracy tumor identification to support clinical
decisions.
Project
Objectives
• Develop a Convolutional Neural Network (CNN) for effective tumor
detection.
• Preprocess MRI scans through resizing, normalization, and augmentation.
• Classify images into tumor and non-tumor categories accurately.
• Evaluate model performance using accuracy, precision, and recall metrics.
“ Machine learning approaches, especially deep learning, can also play a
vital role in the analysis, segmentation, and classification of cancer
images, especially brain tumors. Furthermore, the use of such methods
paves the way for accurate and error-free identification of tumors to
recognize and distinguish them from other similar diseases.”
- from research paper on ‘MRI-based brain tumor detection using convolutional deep learning methods’
by Soheila Saeedi1, Sorayya Rezayi1*, Hamidreza Keshavarz2 and Sharareh R. Niakan Kalhori
Methodology Part 1: Data &
Preprocessing
• Datasets sourced from Kaggle and Figshare, featuring labeled MRI
scans.
• Preprocessing steps:
⚬ Resizing images to uniform dimensions for model compatibility.
⚬ Normalization to standardize pixel intensity ranges.
⚬ Data augmentation: rotation, flipping, and zooming to enhance model
robustness.
Methodology Part 2: Model Architecture &
Training
• Custom CNN built using Keras on TensorFlow backend.
• Multiple convolutional layers with ReLU activation and max pooling.
• Training with cross-validation to prevent overfitting.
• Evaluation metrics: accuracy, precision, recall tracked
during training and validation.
“ The first convolutional layer of the CNN is used to learn features such as color and edges, the second
convolutional layer is used to learn more complex information, such as the borders of brain tumors. The
succeeding (deeper) convolutional layers build up their features by merging the features learned by the earlier
convolutional layers”
‘A hybrid deep CNN model for brain tumor image multi‑classification’
by Saravanan Srinivasan1 , Divya Francis , Sandeep Kumar Mathivanan , Hariharan
Rajadurai
System Workflow
Input MRI
Preprocessing
CNN Model
Classification
This pipeline enables seamless automation from raw images to diagnosis-ready
results.
Expected
Outcomes
• High accuracy detection of brain tumors from MRI scans.
• Decision support for radiologists, reducing diagnostic workload.
• Solution scalable for deployment in hospitals and clinics
worldwide.
Advantages of the Proposed
System
• Rapid and reliable tumor diagnosis improves patient care timelines.
• Reduces human error by standardizing image interpretation.
• Accessible for resource-limited hospitals with streamlined model
deployment.
“Full machine learning and deep learning algorithms can be implemented as future
enhancements. Furthermore, the proposed techniques can be used to detect different forms of
cancers in MRI or Computed Tomography (CT) scan.”
Applications & Future
Vision
• Integration in hospitals and diagnostic centers for routine screening.
• Telemedicine platforms enabling remote diagnosis and second opinions.
• Support for research and academic projects, advancing brain tumor
studies.
References and
Citations
1. “A hybrid deep CNN model for brain tumor image multi-classification“ by Saravanan Srinivasan, Divya Francis, Sandeep
kumar Mathivan, Hariharan Rajadurai, Basu Dev Shivhare and Mohd. Asif Shah
2. “MRI-based brain tumor detection using convolutional deep learning methods and chosen machine learning techniques” by Soheila
Saeedi, Sorayya Rezayi, Hamidreza Keshavarz and Sharareh R. Niakan Kalhori
3. “Detection and Classification of Brain tumors Using Deep Convolutional Neural Networks” (Balaji et
al., 2022)
THANK
YOU!!!