Brain Tumor Detection in MRI Images Using Classical Image Processing Techniques

Abstract

This research introduces a stepwise method for enhancing and segmenting MRI brain scans, utilizing classical image processing techniques including anisotropic diffusion, skull extraction, and contrast enhancement strategies like top-hat and histogram-based filtering. The tumor segmentation process leverages the watershed transform for precise boundary detection. The entire workflow, developed in Python with OpenCV and scikit-image, was applied to a curated set of 80 brain MRI samples. Qualitative analysis suggests that this resource-efficient and explainable pipeline reliably identifies tumor regions with high clarity — making it practical for clinical environments with limited computational infrastructure.

Introduction

The human brain, central to cognition and body regulation, is vulnerable to tumors—either benign or malignant—that can cause severe impairments. Early detection is critical for effective treatment.

MRI in Brain Tumor Detection

• MRI is a preferred imaging tool due to its non-invasive, radiation-free, high-resolution soft tissue visualization.

• However, manual MRI interpretation is time-consuming, requires expertise, and suffers from inter-observer variability.

• Automated systems are needed to address growing imaging demands and workforce shortages.

Proposed Classical Image Processing System

Unlike deep learning approaches that require massive data and computing power, this study proposes a lightweight, transparent, and efficient classical method for tumor detection using 80 MRI images from Kaggle.

Key Advantages:

• No need for large labeled datasets

• Resource-friendly (suitable for low-compute environments)

• Highly interpretable for clinicians

Pipeline Overview

Implemented in Python (Google Colab) using libraries like OpenCV, NumPy, and scikit-image.

Preprocessing Steps:

1. Anisotropic Diffusion Filtering (ADF): Reduces noise while preserving edges.

2. Skull Stripping: Removes non-brain tissues using thresholding and morphology.

3. Top-Hat Filtering: Enhances small bright regions like tumors.

4. Histogram Equalization: Improves contrast; sometimes CLAHE is used.

5. Binarization: Converts image to binary using Otsu's thresholding.

Segmentation Phase:

• Watershed Segmentation: Models intensity as terrain to isolate tumor regions.

• Morphological Refinement: Smooths and cleans segmented tumor mask (e.g., using closing, dilation).

Literature Foundations

The methodology builds on established techniques:

• Anisotropic diffusion (Perona & Malik)

• Histogram equalization and morphology (Gonzalez, Woods, Russ)

• GLCM for texture analysis (Haralick)

• Otsu’s method for thresholding

• Watershed algorithm (Vincent & Soille)

• Adaptive histogram equalization (Zhang)

Experimental Setup

• Dataset: 80 labeled MRI scans from Kaggle.

• Platform: Google Colab for reproducibility.

• Tools: Python, OpenCV, scikit-image, NumPy.

• Focus: Efficiency, interpretability, and clinical relevance.

Conclusion

This study introduced a structured image processing framework for detecting and segmenting brain tumors from MRI scans using classical techniques. The proposed method follows a multi-step workflow including noise reduction, tissue isolation, contrast enhancement, and region-based segmentation — all implemented with open-source Python libraries. Its lightweight design makes it well-suited for clinical environments with limited computational resources. Qualitative evaluation of selected MRI images showed that each processing step contributed meaningfully to the identification of tumor regions. While the approach performed reliably in most cases, challenges such as reduced contrast or image noise occasionally affected the segmentation accuracy. Beyond segmentation, feature extraction based on shape, intensity, and texture provides a foundation for potential classification tasks in future studies. The interpretability and modular nature of the pipeline also support its integration with machine learning components to enhance performance further. Overall, this work reinforces the relevance of classical image processing in medical imaging applications, particularly where transparency and cost-effectiveness are key. It offers a reproducible and adaptable baseline that can guide future research and the development of practical diagnostic tools.

References

[1] P. Perona and J. Malik, “Scale-space and edge detection using anisotropic diffusion,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 12, no. 7, pp. 629–639, 1990. [2] R. C. Gonzalez and R. E. Woods, *Digital Image Processing*, 4th ed. Pearson, 2018. [3] J. C. Russ, *The Image Processing Handbook*. CRC Press, 2016. [4] D. Jha et al., “An automated method for brain tumor detection and segmentation using watershed and morphological operations,” IEEE Access, vol. 7, pp. 145928–145941, 2019. [5] N. Paul, “Brain MRI Images for Brain Tumor Detection,” *Kaggle*, [Online]. Available: https://www.kaggle.com/navoneel/brain-mri-images-for-brain-tumor-detection [Accessed: July 21, 2025]. [6] R. M. Haralick, K. Shanmugam, and I. Dinstein, “Textural features for image classification,” IEEE Transactions on Systems, Man, and Cybernetics, vol. SMC-3, no. 6, pp. 610–621, 1973. [7] N. Otsu, “A threshold selection method from gray-level histograms,” IEEE Transactions on Systems, Man, and Cybernetics, vol. 9, no. 1, pp. 62–66, 1979. [8] L. Vincent and P. Soille, “Watersheds in digital spaces: An efficient algorithm based on immersion simulations,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 13, no. 6, pp. 583–598, 1991. [9] Y. Zhang, “An adaptive histogram equalization algorithm for the image gray level mapping,” *Procedia Engineering*, vol. 15, pp. 3738–3742, 2011.

Copyright

Copyright © 2025 Manoj N Sogi, Dr. S. Lokesh. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.