Dental AI: Deep Learning for Lesion Detection

Sometimes between the start and end of a project lies not just lines of code, but a story of transformation.

Today I want to share some exciting news with you: our academic paper published under Elsevier is now available. But in this post, I don't just want to introduce the paper—I want to tell the story behind it.

Months of Data Collection

The foundation of every AI project is data. For us, this meant annotating lesion and crown relationships on panoramic dental radiographs. During this months-long process, we held countless meetings with Buse Çebi Gül and other valuable academics. Long discussions over each image, annotation decisions, and standardization efforts...

This phase taught me something clearly: In machine learning, data quality directly determines model success.

From YOLOv5 to YOLO11: A Learning Journey

When I started the project, my experience was limited to a similar YOLOv5 project. Looking back now, I can see how far I've come. During this process:

• I learned the YOLO11 architecture in depth

• I discovered segmentation models like U-Net and DeepLab V3

• I applied model comparison and evaluation methodologies

• I experienced academic writing standards

The Value of Working with Professionals

This project gave me much more than technical knowledge. Staying in constant communication with expert academics, exchanging ideas in meetings, and constructively evaluating criticism... All of this took my understanding of work to the next level.

I'm no longer just someone who writes code. I'm on my way to becoming a researcher who can define scientific problems, collect data, train models, and present findings to academic standards.

Conclusion

This paper is not an end for me, but a beginning. It opened the doors to contributing to academia and the scientific world. From now on, I will continue to train more successful models and conduct more impactful research.

Access the paper: https://authors.elsevier.com/c/1mIU355rY~IPV