EmirMuhammetARAN/BraTS2020-Brain-Tumor-Segmentation — reverse-engineered prompt

Build me a complete Python notebook for a brain MRI tumor segmentation research project using the BraTS 2020 training dataset from Kaggle. I want it to load the four MRI scan types, prepare the slices, train a U Net style model that can mark necrotic tumor tissue, active enhancing tumor, and edema, then validate it on a held out set.

Please include clear charts and saved images so I can see the predicted tumor areas compared with the real masks, plus confusion matrices, Dice, IoU, sensitivity, specificity, and precision for each tumor class. Add a simple final medical style report that explains the results in plain English and says this is for research or education only, not clinical diagnosis without radiologist review.

Make the notebook easy to run from top to bottom, include the needed requirements file, and save the best model weights. If anything has changed in the libraries, look up current docs online and adjust.