Artificial intelligence to aid in the diagnosis of celiac disease

A Smart Aid for Celiac Disease Diagnosis

Scientists have developed a machine learning (or AI) model that can assist in diagnosing celiac disease. It photographically analyzes magnified images of small tissue samples (biopsies) taken from the small intestine, just as a medical specialist (pathologist) would do.

 

Why is this important?

• Faster diagnosis: Sometimes, it can take a long time for a patient to receive a diagnosis, especially because there aren’t enough medical specialists to analyze these biopsies, in both developed and developing countries. This program could speed things up.
• More consistent diagnosis: Even medical specialists can sometimes have different opinions about a biopsy. This program could help provide more uniform results.

 

How was this program “trained” and tested?

• The program’s “school”: The program “learned” by looking at over 3,300 biopsy images from patients at four hospitals in the UK, along with their actual diagnoses. It used five different types of scanners to get used to varied images.
• The program’s “exam”: To see how well it learned, the program was tested on nearly 650 new images it had not seen before, from a fifth hospital. Also, its results for 30 images were directly compared to those of gastrointestinal medical specialists.

 

How well did it perform?

• Very accurate: The program was able to correctly identify celiac disease (sensitivity) and correctly exclude the disease in healthy individuals (specificity) in over 95% of cases. This means it makes few mistakes.
• As good as experts: Its performance was as good as that of medical specialists. This shows that artificial intelligence has great potential to help automate celiac disease diagnosis.

 

What does this mean for different groups of people?

The program worked well for most people, regardless of age or gender. There was a slight drop in performance in young people between 10 and 19 years old, possibly because there weren’t enough examples from this age group for the program to learn as well.

 

Are there limitations?

Yes, it is important to know that sometimes even medical specialists can have difficulty being 100% certain about a celiac disease diagnosis solely from a biopsy, as there can be small differences in interpretation. Although efforts have been made to address this, “absolute truth” in diagnosis is hard to achieve, and this also influences how the program is trained and evaluated.

 

In conclusion:

This smart program is an important step forward. It could help doctors make celiac disease diagnoses faster and more consistently, which is good news for patients. It is a promising example of how modern technology can be used in medicine.

The study was financially supported by organizations such as Coeliac UK, Innovate UK, and the Cambridge Centre for Data-Driven Discovery.

Source Medscape