All classes Brain Tumor Segmentation with U-Net Syllabus · Episode 4 of 6

Class U-Net Brain Tumor Segmentation Loss #4: Loss for Segmentation: Dice, IoU & BCE-Dice, then Training U-Net Baseline

4 views

Segmentation loss tutorial: Dice coefficient, IoU, BCE-Dice loss, and training U-Net baseline with PyTorch. Episode 4 of the brain tumor segmentation series with U-Net and metaheuristics.

Software: Deep Learning (PyTorch U-Net)
Level: Advanced
Duration: ± 11 min
Series total: 1 h 10 min · 6 videos
Published: October 24, 2026

Episode 4 of 6 · view syllabus

Download PDF (6 PDF pages)

0 downloads

Download Project File· 95 MB Unlock All Premium Classes ▶ Watch the video lesson0 plays

▶ Video Lesson

Watch every step above performed live, with English narration.

0 plays

Class book

Page 1 / 6

Class U-Net Brain Tumor Segmentation Loss #4: Loss for Segmentation: Dice, IoU & BCE-Dice, then Training U-Net Baseline: Cover: Class U-Net Brain Tumor Segmentation #4: Loss for Segmentation: Dice, IoU & BCE-Dice, then Training U-Net Baseline

Class U-Net Brain Tumor Segmentation Loss #4: Loss for Segmentation: Dice, IoU & BCE-Dice, then Training U-Net Baseline: Before you start, steps list, and Step 1

Class U-Net Brain Tumor Segmentation Loss #4: Loss for Segmentation: Dice, IoU & BCE-Dice, then Training U-Net Baseline: Step 2–3: BCE-Dice loss

Class U-Net Brain Tumor Segmentation Loss #4: Loss for Segmentation: Dice, IoU & BCE-Dice, then Training U-Net Baseline: Step 4–5: Evaluation function

Class U-Net Brain Tumor Segmentation Loss #4: Loss for Segmentation: Dice, IoU & BCE-Dice, then Training U-Net Baseline: Step 6: Training U-Net baseline

Class U-Net Brain Tumor Segmentation Loss #4: Loss for Segmentation: Dice, IoU & BCE-Dice, then Training U-Net Baseline: Done: Final Result and Further Practice

Cover: Class U-Net Brain Tumor Segmentation #4: Loss for Segmentation: Dice, IoU & BCE-Dice, then Training U-Net Baseline

Click a page or use ◀ ▶ to turn · swipe on mobile · to enlarge

Want to go deeper?

Join one of our courses or ask us directly on WhatsApp.

Download PDF· 0 downloads View the full 6-episode syllabus View Medical Image Segmentation Courses Ask on WhatsApp

Recommended classes for you

Picked from the software, topic, and level of “Class U-Net Brain Tumor Segmentation Loss #4: Loss for Segmentation: Dice, IoU & BCE-Dice, then Training U-Net Baseline”.

18 episodesFree preview

Python · Beginner–Advanced

Machine Learning Python

18 video lessons with PDF guides, 3 h 41 min in total. Starts with: Introduction to Machine Learning & First Model.

3 h 41 min 61 0

View class

ANFIS Metaheuristics Time Series Forecasting

6 episodesFree preview

Python · Intermediate–Advanced

ANFIS Metaheuristics Time Series Forecasting

6 video lessons with PDF guides, 59 min in total. Starts with: Non-Linear Time Series: Dataset, Stationarity, and Patterns.

59 min 13 0

View class

Deep Learning CNN-LSTM Pneumonia Detection

6 episodesFree preview

Python · Intermediate–Advanced

Deep Learning CNN-LSTM Pneumonia Detection

6 video lessons with PDF guides, 57 min in total. Starts with: Chest X-ray Dataset: Loading, Exploration & Data Splitting.

57 min 22 0

View class

50 episodesFree preview

Agentic AI · Beginner–Intermediate

Agentic AI Without Coding

50 video lessons with PDF guides, 14 h 29 min in total. Starts with: Chatbots, Automation, and AI Agents: What's the Difference?.

14 h 29 min 66 0

View class

100 episodesFree preview

Claude · Beginner–Advanced

Claude from Zero to Expert

100 video lessons with PDF guides, 15 h 1 min in total. Starts with: What Claude Is and How It Differs from ChatGPT and Gemini.

15 h 1 min 62 0

View class

6 episodesFree preview

Python · Intermediate–Advanced

Transfer Learning for Leaf Disease

6 video lessons with PDF guides, 1 h 5 min in total. Starts with: PlantVillage Dataset: Folder Structure, Class Distribution & Data Split.

1 h 5 min 9 0

View class

NSGA-II MOPSO Hybrid Multi-Objective Optimization

6 episodesFree preview

Python · Intermediate–Advanced

NSGA-II MOPSO Hybrid Multi-Objective Optimization

6 video lessons with PDF guides, 59 min in total. Starts with: Multi-Objective Optimization: Pareto Front, Dominance, and Test Functions ZDT/DTLZ.

59 min 9 0

View class

EKG Arrhythmia Classification Wavelet Deep Learning

6 episodesFree preview

Python · Intermediate–Advanced

EKG Arrhythmia Classification Wavelet Deep Learning

6 video lessons with PDF guides, 56 min in total. Starts with: EKG & EEG Signals: Characteristics and Public Datasets (MIT-BIH).

56 min 2 0

View class

Need more than a video?

Other ways we can help with Python.

Learn with a mentor

Scheduled AI & Machine Learning courses with live guidance and a certificate.

Course

Multi-Objective Optimization Course: NSGA-II, MOPSO & Hybrid Algorithms (Python)

Course

ECG Arrhythmia Classification Course with Wavelet & Deep Learning (PyTorch)

Course

ANFIS + PSO/GA Course for Non-Linear Time Series Forecasting (Python)

See all in AI & Machine Learning →

Have it built for you

No time to learn Python? Our team can do the project for you.

Service

Python Machine Learning — Prediction Models, Customer Segmentation, Fraud Detection & AI Apps

Service

AI Image Creation with Stable Diffusion — Unlimited Custom Visual Production

★ 4.9

Service

AI Image & Video Creation with Leonardo AI — All-in-One Visuals for Business

★ 4.9

See all in AI & Machine Learning →

Get notified about new Python classes

At most one email a week. Unfollow anytime with one click.

Transcript

The full narration of this video. The current sentence is highlighted while the video plays; click any sentence to jump there.

Show transcript

Introduction

This is the final result, another UNED-BASE training curve, with LOS-BASE-E DAIS continuously decreasing and DAIS validation increasing up to around 0.76. We also understand why pixel accuracy is misleading, and why segmentation researchers use DAIS and IOU. Let's get started. Hello, welcome to episode four of the brain tumor segmentation series with the metaheuristically optimized UNED. Our UNED is now ready. Before training it, we must determine how to evaluate and penalize model errors. In medical segmentation, the choice of metrics and loss functions is very crucial, because tumor pixels only make up a small portion of the image.

Step 1: Dice Formula and IoU

Create file LOS-SEC.PE DAIS coefficient is twice the intersection between prediction and target, divided by the number of pixels of both. IOU, or Jaccard, is intersection divided by union. Both have values between 0-1, and only count tumor pixels, so the large black background does not raise the score. A small epsilon prevents division by zero.

Step 2: BCE-Dice Loss

Then the class-weighted loss. Binary cross entropy penalizes each pixel and makes training stable, while DAIS loss, which is 1 minus DAIS, directly optimizes tumor overlap. Both are combined with half-half weights. This combination is very common in medical segmentation research.

Step 3: Why Pixel Accuracy is Deceptive

Let's test with a small example, a 4x4 tumor mask in a 10x10 image. A perfect prediction gets DAIS and IOU of 1. Prediction shifted 2 pixels down drastically becomes DAIS 0.25 and IOU 0.14. Empty prediction gets 0. But look, the pixel accuracy of the empty prediction is actually 84%. This is why accuracy should not be used for segmentation.

Step 4: Evaluation Function

Now create the train underscore unit.py file. The evaluation function collects predictions for all validation data, at threshold 0.5, then calculates global DAIS and IOU.

Step 5: Reused Training Function

The training function we create can be reused with learning rate parameters, number of filters F, dropout, number of IPOC, and data proportion. This function will be called repeatedly by the metaheuristic algorithm in episode 5. For each IPOC we calculate the validation DAIS, and the model with the best DAIS is saved.

Step 6: Training U-Net Baseline

The main part of training the baseline unit with general settings, learning rate 0.001, 16 filters, no dropout, 15 IPOC. Run. Validation DAIS rises quickly from 0.4 in the first IPOC, then fluctuates, this is normal because small tumors are difficult and validation data comes from different patients. The best value 0.763 is achieved at IPOC 11. The left graph shows the continuously decreasing training loss, the right graph shows the validation DAIS. These numbers become the benchmark for the optimized model.

Done!

Done. We have now understood DAIS and IOU Creating base CI DAIS loss, reusable training function, and baseline unit with DAIS validation 0.76. In the next episode, we will automatically optimize hyperparameters of the unit using Particle Swarm Optimization and Grey Wolf Optimizer. Currently working on a thesis, dissertation, or AI research in the field of medical image segmentation? Cerdas Inovasi Digital helps with deep learning implementation, from UNED segmentation, metahuristic optimization PSO and Grey Wolf, up to DAIS evaluation, IOU, and result writing, as well as offering deep learning courses with PyTor. Want to master the entire series? From MRI brain datasets to metahuristically optimized UNED? All 6 episodes, complete with step-by-step PDF guides and source code, are available to premium tutorial subscribers. Subscribe starting at Rp69.000 per month, save 28% with the 12-month package, or pay once Rp999.000 for lifetime access. Want just one class? Buy a single class with lifetime access. Subscribe now on the premium page cerdasinovasidigital.com Contact our WhatsApp or visit cerdasinovasidigital.com Thank you for watching, see you in the next episode.

Ask the class material

Type a question or keyword; we find the part of this episode that explains it and jump the video there.

My notes

Notes are saved with the current video time and are only visible to you.

Questions & answers

Stuck on a step? Ask here; the instructor and other learners can answer.

No questions yet.

Your instructor

Cerdas Inovasi Digital Team

Software developers, BIM & 3D engineers, and AI practitioners

53 video classes1,296 episodes

We build software, BIM models, games and AI automations for clients every day, and turn that hands-on work into step-by-step classes: every step is shown on screen and comes with a PDF guide you can keep.

See our clients →

Memuat

▶ Video Lesson

Recommended classes for you

Need more than a video?

Transcript

0:03Introduction

0:47Step 1: Dice Formula and IoU

1:57Step 2: BCE-Dice Loss

2:44Step 3: Why Pixel Accuracy is Deceptive

3:59Step 4: Evaluation Function

5:26Step 5: Reused Training Function

7:29Step 6: Training U-Net Baseline

9:40Done!

Ask the class material

My notes

Questions & answers

Your instructor

▶ Video Lesson

Recommended classes for you

Need more than a video?

Transcript

0:03Introduction

0:47Step 1: Dice Formula and IoU

1:57Step 2: BCE-Dice Loss

2:44Step 3: Why Pixel Accuracy is Deceptive

3:59Step 4: Evaluation Function

5:26Step 5: Reused Training Function

7:29Step 6: Training U-Net Baseline

9:40Done!

Ask the class material

My notes

Questions & answers

Your instructor

Introduction

Step 1: Dice Formula and IoU

Step 2: BCE-Dice Loss

Step 3: Why Pixel Accuracy is Deceptive

Step 4: Evaluation Function

Step 5: Reused Training Function

Step 6: Training U-Net Baseline

Done!

Introduction

Step 1: Dice Formula and IoU

Step 2: BCE-Dice Loss

Step 3: Why Pixel Accuracy is Deceptive

Step 4: Evaluation Function

Step 5: Reused Training Function

Step 6: Training U-Net Baseline

Done!