Laser and AI pave the way for virtual biopsies

Researchers at Stanford Medicine have developed a non-invasive method for analysing possible cancer cells. The so-called virtual biopsy uses a laser to penetrate tissue and create a high-resolution three-dimensional reconstruction from which cross-sectional images can be made. These images mimic those generated by a standard biopsy, in which a tissue sample is cut into thin layers and placed on slides to be examined under a microscope.

The method, published in the April issue of Science Advances, has the potential to examine tissues non-invasively for abnormal cells, as well as to provide rapid results in biopsies carried out on other parts of the body. It could also provide more information than current diagnostic approaches, according to the researchers.

‘We have not only created something that can replace the current gold standard pathology slides for diagnosing many conditions, but we have also greatly improved the resolution of examinations in order to gather information that would be extremely difficult to have otherwise,’ explains Adam de la Zerda, PhD associate professor of structural biology and senior author of the published article. The method was developed by Yonatan Winetraub, a former graduate student in the lab previously led by Zerda.

Laser vision

When a dermatologist or surgeon takes a biopsy, whether of the skin, liver, breast or other organ, the tissue is usually sent to a pathologist, who cuts it into thin layers and applies reactive chemicals to each layer to identify patterns, shapes and structures of the cells more easily. These slides are routinely used to diagnose cancer and other diseases and require a lot of work that is irreversible, i.e. once the tissue is cut in one direction, it cannot be sliced in another way to offer a different view.

For this reason, la Zerda’s team has been studying a different technique called optical coherence tomography for almost a decade. Normally used by ophthalmologists, this test measures how light waves from a laser reflect off a tissue to create a representation, in a similar way to ultrasound equipment.

The optical coherence scans were improved to work with organs other than the eye. The quality of the images improved over the course of the work, allowing us to see more and more details of the tissues until they became very similar to those obtained by the conventional methods used by pathologists.

‘All hospital doctors are used to reading traditional pathological test results, so it was important to translate the optical coherence tomography images into something comfortable for them, rather than a totally new type of image,’ says de la Zerda.

A pinch of AI couldn’t be missing

The researchers also used Artificial Intelligence (AI) to help convert the digitised optical coherence tomography images into images similar to those of conventional biopsy slides. According to them, the richness of the work lay in the method developed to align pairs of virtual and real biopsy images to train machine learning algorithms.

As a test, three Stanford dermatologists analysed random groupings of real and virtual images and managed to detect cellular structures at a similar rate.

Until now, when a doctor noticed an unusual-looking spot on a patient’s skin, there were two options: wait and see if it grows or remove it and send it to a pathologist. With the virtual biopsy, a third way has emerged: take an optical coherence tomography scan and analyse the images.

In the same way, surgeons until now removed possible breast tumours and sent the material to pathologists, having to wait several days for the result and to determine whether a second surgery was necessary to widen the safety margin and then eliminate more cancerous cells. With the virtual biopsy produced by an optical coherence camera in the operating theatre, it will be possible to make this decision instantly in the future.

More work will still be needed to advance these applications, but the researchers are confident that these approaches will give doctors a new way of performing biopsies.