Data source: https://www.kaggle.com/mateuszbuda/lgg-mri-segmentation
Clickable link for the full list of startups in AI medical imagery: https://research.aimultiple.com/looking-for-better-medical-imaging-for-early-diagnostic-and-monitoring-contact-the-leading-vendors-here/
Feature Extraction and Convolutions: https://setosa.io/ev/image-kernels/
CNN Visualization: https://www.cs.ryerson.ca/~aharley/vis/conv/flat.html
Excellent Resource on transfer learning by Dipanjan Sarkar: https://towardsdatascience.com/a-comprehensive-hands-on-guide-to-transfer-learning-with-real-world-applications-in-deep-learning-212bf3b2f27a
Article by Jason Brownlee: https://machinelearningmastery.com/transfer-learning-for-deep-learning/
We need a custom loss function to train this ResUNet.So, we have used the loss function as it is from https://github.com/nabsabraham/focal-tversky-unet/blob/master/losses.py
NOW YOU KNOW HOW TO APPLY AI TO DETECT AND LOCALIZE BRAIN TUMORS. THIS IS A GREAT ACHIEVEMENT IN HEALTHCARE.
UNDERSTAND THE PROBLEM STATEMENT AND BUSINESS CASE IMPORT LIBRARIES AND DATASETS PERFORM DATA VISUALIZATION Plot 12 randomly selected both MRI image and the corresponding mask UNDERSTAND THE THEORY AND INTUITION BEHIND CONVOLUTIONAL NEURAL NETWORKS AND RESNETS TRAIN A CLASSIFIER MODEL TO DETECT IF TUMOR EXISTS OR NOT Change the network architecture by adding more/less dense layers ASSESS TRAINED MODEL PERFORMANCE Print out the classification report UNDERSTAND THE THEORY AND INTUITION BEHIND RESUNET MODELS BUILD A SEGMENTATION MODEL TO LOCALIZE TUMOR TRAIN A SEGMENTATION RESUNET MODEL TO LOCALIZE TUMOR ASSESS TRAINED SEGMENTATION RESUNET MODEL PERFORMANCE UNDERSTAND THE THEORY AND INTUITION BEHIND TRANSFER LEARNING List the challenges of transfer learning (external research is required)