When you think of a “pathologist,” you might picture a person in a lab coat, looking at slides containing patient tissue under a microscope. In fact, if you search for “pathologist” online, most pictures will show that. But the job has changed a lot recently. Now, instead of just using a microscope, pathologists can take digital pictures of the slides and like photos on your phone, can view them on a computer. This helps them work faster as they are able to get second opinions quicker instead of sending a slide by post.
As the records are all digitised (in Northern Ireland since 2022) this means sample tracking has become much more robust, which makes it easier to figure out what’s wrong. This is important because pathologists help other doctors figure out about 70% of all medical diagnoses, and in Northern Ireland, there are 40 million pathological diagnostic tests every year.
Diagnosis of Barrett’s oesophagus is one of these tests (Figure 1). Barrett’s oesophagus can be found by looking at tiny pieces of tissue under a microscope, taken from your food pipe (or gullet) during a procedure called an endoscopy. Once a pathologist identifies Barrett’s oesophagus, they need to also look for the potential presence of dysplasia. But what does “dysplasia” mean? This is when cells in the body start growing in the wrong way. If enough cells become abnormal, it may lead to cancer. Pathologists sort Barrett’s into four types:
- Non-Dysplastic (normal cells).
- Indefinite for Dysplasia (not sure if the cells are abnormal).
- Low-Grade Dysplasia (cells starting to look different).
- High-Grade Dysplasia (cells look very different).

Figure 1. Diagnosis of Barrett’s Oesophagus following endoscopy (Created in https://BioRender.com).
A. Clinician performs endoscopy examination.
B. Biopsies are taken from the food pipe and processed onto a glass slide.
C.1. Traditionally a microscope would be used to view patient’s samples.
C.2. Modern pathology involves scanning and viewing slides on a computer.
D. Pathologist reviews tissue slides.
E. Pathologist grades samples.
F. Secondary tests and opinions. Digital pathology allows multicenter options for review.
G. Follow up appointments and monitoring.
The good news is, most people with Barrett’s oesophagus have the non-dysplastic type, which means they’re very unlikely to get cancer of the oesophagus. Previous research from the Northern Ireland Barrett’s oesophagus register has estimated that three people in every one hundred will progress to cancer over a ten year timeframe. For the few people with the other types of Barrett’s, it’s really important to figure out what type of dysplasia is present. To determine this, the tissue samples are sent to another pathologist to check again and because the tissue images are now digital, pathologists from different places can view these, which makes it faster to find problems. They will also use a method called Immunohistochemistry (IHC). IHC uses antibodies to detect proteins in tissues samples. One of the proteins they test for is p53 which is often expressed in dysplastic Barrett’s oesophagus and is an early marker for cancer (BSG Guidelines).
While identifying Barrett’s is straightforward, distinguishing between the four types of Barrett’s oesophagus still proves challenging for pathologists. In 2018 a large population-based study from the Northern Ireland Barrett’s register showed that around 13% of patients with high grade dysplasia or oesophageal adenocarcinoma were likely to be
misdiagnosed with Barrett’s oesophagus (non-dysplastic or low-grade) initially as they then developed high-grade dysplasia or oesophageal adenocarcinoma within the 3-12 months post-prior to their diagnosis. This is where Artificial Intelligence (AI) could help. Some studies have shown that AI can not only help doctors diagnose Barrett’s oesophagus but also help figure out how serious it is through identifying dysplasia and the types of Barrett’s.
Other ways AI can help is through surveillance, by identifying for example a morphological feature within the tissue which can stratify a patient’s risk of progression to cancer.
I got the chance to talk to Dr. Damian McManus, a pathologist with more than 30 years of experience. He told me how tools, like AI, are helping doctors find and treat diseases more quickly and accurately.
Richard: How long have you been a pathologist?
Damian: I started training in 1990, passed MRCPath in 1996 and was appointed as Consultant in 1998 (27 years consultant experience).
Richard: How have you found your job has changed over time? ie When was digital pathology introduced and has it sped-up diagnosis?
Damian: The three big changes in my professional career have been increased sub specialisation, increasing use of IHC and molecular tests for diagnosis and as predictive markers and the introduction of digital pathology for primary reporting. As I write this, I am at a conference so I can’t check when digital was implemented: it was the summer of 2021 or 2022.
Richard: What is the current standard procedure for a Barrett’s diagnosis and Grading in your work?
Damian: We currently diagnose Barrett’s oesophagus when columnar mucosa is found on biopsy in the oesophagus. This may show intestinal metaplasia or there may be columnar mucosa which is more like gastric mucosa or a mixture of both. Correlation with the endoscopic findings is mandatory and very important. Most Barrett’s oesophagus biopsies are negative for dysplasia. A minority show dysplasia; this may be classified as low grade or high grade. In line with BSG guidance all dysplasia samples have an independent second read by a second pathologist and immunohistochemistry for p53 and MIB1 is used frequently.
Richard: Day to day what is the biggest challenge when it comes to oesophageal biopsy examination and Barrett’s diagnosis?
Damian: Distinction between dysplasia and reactive atypica secondary to ulceration and active inflammation is still probably the most problematic; the category of indefinite for dysplasia is the categorisation used if the pathologist is not sure, although it is important to use this category sparingly.
Richard: Do you currently use AI in your job?
Damian: Not currently in histological diagnosis. Large language models like ChatGPT can be very useful in undergraduate and postgraduate teaching.
Richard: How do you feel about the introduction of AI to your field of work?
Damian: I think that it is inevitable that it will be introduced and that it might ultimately change the role of the pathologist.
Richard: Where do you feel AI could assist you most with Barrett’s Diagnostics?
Damian: In many NHS laboratories there are large backlogs of routine cases; models such as Paige’s Virchow model could help triage these cases and detect unsuspected cancers
Richard: How important is human intervention in the world of AI for diagnostic purposes?
Damian: Currently it is still very important, and the pathologist still has to make the final decision…. Although who knows how long that will last.
From the conversation with Dr McManus, it is evident that AI could prove helpful in a Barrett’s diagnostic setting alongside a pathologist. A major limiting factor in AI in the diagnostic setting is the vast range of data it requires to perform efficiently. A multicenter approach and agreed guidelines for Barrett’s oesophagus are required – both of which are currently not in place. For my project I aim to use Northern Ireland Barrett’s Register data and AI to identify a novel biomarker which will help stratify Barrett’s oesophagus patients on their risk of developing oesophageal cancer which will prioritise surveillance strategies.
Authors: Mr. Richard Murray and Dr Damian McManus.