Interview with Emilian Gheorghe on using AI models for breast cancer grading

Pathologist Emilian Gheorghe discusses his experience building an AI model for breast cancer grading and the shift toward digitizing the field.
Written by Aiforia

While breast cancer is a leading cause of cancer death in women worldwide, current approaches for diagnosis and therapy are in need of an upgrade. Early detection and treatment of breast cancer considerably improves outcomes. Hence, many countries have adopted mammography screening programmes which include a critical component – accurate pathological evaluation of the tumor.

Image analysis results often vary widely between pathologists and laboratories as current techniques are bias-prone and not easily reproducible. A decrease in specialists worldwide is also resulting in a workload increase, causing pathologists to work longer hours. False positives and negatives lead to unnecessary invasive diagnostic procedures or missed malignant tumors that become less treatable as they advance.

Artificial intelligence (AI) is a unique solution to these problems, improving speed and accuracy of mammography screenings and medical-image analysis in general. With its scalability, high-quality care becomes more accessible and the data reproducible. As AI can continuously evolve, getting better with each input of annotations, it is a crucial high-precision assistive tool aiding pathologists’ toward objective and accurate final decisions.

Pathologist Emilian Gheorghe is building an AI model for breast cancer grading with Aiforia Create, through Aiforia’s aiForward program. We interviewed him about the beginning stages of his project and what intrigued him about using AI.

Interview with Emilian Gheorghe, Pathologist, University of Medicine and Pharmacy of Timisoara

Describe the project or research you are working on.

Emilian: I'm trying to develop with the help of the Aiforia Create platform a custom AI algorithm that will eventually be able to grade breast cancer.

What is your motivation in using AI for this project?

Grading breast cancer represents a very important step in the pathological evaluation of a malignant breast tumor. Unfortunately, this step can sometimes be subjective and has a low reproducibility. My motivation in using AI stems from these issues, and in time AI can have a powerful impact on the whole pathological evaluation in breast cancer.

How has it been getting started with Aiforia? Have you had experience with digital image analysis in the past?

Tricky! You need to learn to do very fine annotations, these represent the pillar of the future AI algorithm you are trying to create. Fortunately, Aiforia's platform is very intuitive, and most importantly, for me at least, error-free. Yes, I had experience with digital image analysis, but nothing on this scale of in-depth analysis. I think once you start dealing with an AI image analysis, well, everything else seems analogic.

How is the AI model performing compared to using traditional methods?

I don't think there are any comparisons to be made, an AI model can continuously evolve, and if it scares the old school establishment, AI models are the future in pathology. They will bring better diagnoses!

Has there been a shift towards digitization in your field? How do you see AI fitting into that?

Like I said before, digitization is here right now for pathology, and it has and will, in a profound way, change the practice of pathology. AI will, I believe in this strongly, be at the forefront of this change.

References

JAMA, 2019
Nature, 2020

QUESTIONS?

Contact us to find out more!