A chat about new AI-enhanced technologies to diagnose cancer in the healthcare setting.
Interview by Letizia Diamante
In the charming city of Jena, the gifted hands of Orlando Guntinas-Lichius perform delicate surgeries on patients’ heads and necks to remove tumours and bring back hope. Based at the Jena University Hospital in Germany, Guntinas-Lichius is responsible for the clinical aspects of the CHARM project. Today, he tells us how CHARM’s microscope powered by AI could benefit healthcare providers.
How are you testing CHARM technology in the clinic?
We use parts of patients’ biopsies with head or neck cancer to compare the results of the standard diagnostic procedure with the ones provided by CHARM technology.
We have involved our pathologists very early on, so we can shape the technology in a way that they will accept it. The standard histopathology procedure requires staining the biopsies, while CHARM technology works without staining, so it can provide results more quickly.
How is it going so far?
The images look very good already. The information from the CHARM microscope, in combination with an artificial intelligence (AI) algorithm, provides an image of the tissues that looks as if the pathologist performed the staining. Therefore, the final images are directly interpretable by the pathologist and even by clinicians like me.
What’s your view on the application of AI in cancer diagnosis?
’AI technologies have already entered our hospital. The Radiology Department uses an AI algorithm to look for metastasis in computer tomography images. The AI technology highlights all potential spots in the image and tells the radiologist whether it seems that each spot is a metastasis or something else.‘Explainability’ is a very important topic in medicine and AI application: AI technology has to explain to the healthcare provider why it decided that a certain spot is cancerous or not. The diagnosis will still be made by the radiologist, but the procedure becomes much faster. We wish to achieve the same with CHARM technology. Pathologists will maintain full responsibility, but AI will help them speed up their work.‘AI technologies have already entered our hospital. The Radiology Department uses an AI algorithm to look for metastasis in computer tomography images. The AI technology highlights all potential spots in the image and tells the radiologist whether each spot seems to be a metastasis or something else. ‘Explainability’ is a very important topic in medicine and AI application: AI technology has to explain to the healthcare provider why it decided that a certain spot is cancerous or not. The diagnosis will still be made by the radiologist, but the procedure becomes much faster. We wish to achieve the same with CHARM technology. Pathologists will maintain full responsibility, but AI will help them speed up their work.
What specific challenges or limitations have you encountered in your current methods of diagnosing cancer using existing microscopes and staining technologies?
In nine out of 10 cases, the histopathology staining of the biopsies is sufficient to provide a diagnosis, for example, to say whether the tumour is benign or not. But, in one out of 10 cases, the staining cannot answer this question, and we need more molecular information, such as the expression of certain biomarkers, to take decisions. This means that after the first histopathology staining, we require a different staining technique to confirm the appropriate therapy. This increases the time and costs involved.
For these reasons, it would be interesting to use a technology that does not require tissue staining.
Why is the CHARM project focusing on head and neck cancer?
Head and neck cancers serve as models for a large variety of cancers, the so-called solid cancers, such as breast cancer and colon cancer. However, compared to other types of solid cancers, head and neck tumours often require more biopsies. Confirming complete removal of the tumour from the neck or head may necessitate as many as 10 or even 20 biopsies, placing a tremendous workload on pathologists.
Another reason is that there are several subtypes of head and neck cancers with high variability. Thus, it will be easy to transfer this method to cases where the variability is smaller.
What are the critical features you believe the new CHARM technology should have to be most effective in diagnosing cancer?
As the population ages and the number of cancer diagnoses rises, pathologists’ workload becomes very intense. To assist pathologists, the first step is to demonstrate that the CHARM technology can diagnose head and neck cancer without the need for standard histology and other staining methods. Then, we can explore the cancer further and pinpoint the best therapy.
As a researcher, I also hope that we can use CHARM technology to discover new cancer signatures that have not been identified until now.
