BRAIN TUMOR

PROBLEM STATEMENT

Typically Brain tumor occurs due to uncontrolled and rapid growth of cells and could be fatal if not treated early. Despite many significant efforts and promising outcomes in this domain, accurate segmentation and classification remain a challenging task. A major challenge for brain tumor detection arises from the variations in tumor location, shape, and size .And the first step in treatment for any brain tumor patient is often surgery to remove as much of the mass as possible and should not be delayed. The objective of this case study is to classify brain tumor through image segmentation from the MRI. Our approach was used to deliver comprehensive results using machine learning and deep learning algorithms on brain tumor detection through image segmentation magnetic resonance imaging combined with patient medical data. After a detailed analysis and research, Ai labs developed and implemented a Proof of Concept (POC) using its proprietary engine Minsky™ to identify the presence of a tumor.

KEY BENEFITS (MINSKY™)

• User-Friendly cloud based AI platform
• Scalable across various domains/data.
• Provides you a list of % dependency features that can be used to optimize the outcomes.
• Ability to fine tune or optimize the models by trying different algorithms / prediction attributes
• Easy integration with other third party solutions such as TABLEAU for data visualization

RESULTS/BENEFITS

• Identification of brain tumor in early stages with better accuracy can save a human life.
• Can accurately diagnose the tumor shape and size.
• Patients can get more safer and effective treatment.
• Reduce time and cost involved in the process.

SOLUTION OVERVIEW:

After comprehensively evaluating the client‘s challenges/data, we used Minsky™ to accurately model the historical data which comprised of some medical parameters. Two features (First order features) like mean, variance, Standard Deviation, Skewness, kurtosis and Second order features like Contrast, Energy, ASM(Angular Second Moment), Entropy, Homogeneity and image data that were labelled & classified into two categories.
1. Tumor
2. Non-Tumor
Minsky™ modelled and analysed these features using machine learning algorithms that were mapped with image data modelled using deep learning algorithms. Historical labelled image data and the corresponding patient’s medical data was used to establish Minsky™ models that are then used with Real-time patient data to predict the presence or absence of a tumor.

Minsky™ ML Based Results for Brain Tumor

Based on the data from Fig 1 above, our analysis shows that ADA Boost Classifier (99%) and Gradient Boosting Classifier (98.5%) gave the highest model accuracy. In terms of the tumor attributes data, Entropy showed the highest dependency cause for the brain tumor followed by ASM (Angular Second Moment).

Minsky™ DL Based Results for Brain Tumor

Data in Fig 2 represent Minsky™ results showing tumor effected brain image with the best matching image. Based on the data from Fig 2 above, our analysis shows that CNN model (87%) gave the highest model accuracy based on our limited Train and Test data images.