From the Human Eye to AI: Computational Pathology and the Future of Cancer Care

Cancer is deeply personal, and at AstraZeneca, we believe that treatment should be too. The evolving field of computational pathology is helping to redefine how we classify some tumour types and better predict the right treatment for each patient. Our aim is to push the boundaries of pathology with AI to help revolutionise cancer care.

Evolution of pathology: Embracing computational advances

Traditional pathology is reliant on microscopy and human interpretation.  The technology to digitise pathology images of patient tissue samples opens the potential for advances in the diagnosis of cancer and treatment decisions. We can now take this even further with computational pathology. By leveraging computer vision and AI, we can augment pathologists’ evaluation of these samples to identify features beyond the perception of the human eye, with the aim of accurately selecting the patients that could see the greatest benefit from a treatment while also increasing our understanding of tumour biology.

Unveiling the unseen with AI

AI can help drive insights beyond human capability. For example, look at the image below. This illusion demonstrates that the human brain relies on surrounding cues to interpret visual messages. Although the two labelled squares look different to our eyes, they are actually the same colour. 

Computational approaches can remove this reliance on contextual cues and thereby improve precision and accuracy. AI can also detect the presence of a biomarker in a sample beyond the acuity of human vision, and it can assess the distribution of the biomarker pixel by pixel in each cell, and the total quantity of a biomarker in a sample in each individual cell scattered throughout the digitised WSI, taking into account the spatial organisation of the tumour.

A powerful pairing: Computational pathology and new medicines

Our novel computational pathology technology, called Quantitative Continuous Scoring (QCS), harnesses AI to assess biomarkers in digitised WSIs derived from patient tissue samples. QCS is our fully supervised AI-based solution to assess the number and distribution of biomarker-positive tumour cells and calculate the staining intensity of biomarkers both on the surface of and inside every tumour cell in a WSI. It provides human-interpretable measures that enable enhanced decision making by pathologists, in a more precise, objective, reproducible manner compared to traditional methods. 

Computational pathology has already confirmed new discoveries, such as the recognition that even the low, or ultra-low presence of some biomarkers on and in tumour cells may yield positive responses to targeted treatments. This observation sheds light on the potential of antibody drug conjugates (ADCs), which can have an effect not only on tumour cells expressing the ADC’s target, but also other nearby tumour cells, which is known as the bystander effect. We have also demonstrated that computational pathology can predict the internalisation of ADCs whose antitumour activity is dependent on being taken into tumour cells, thus identifying tumours that may respond to these highly targeted molecules.  We believe that computational pathology can be a powerful companion to new medicines, helping doctors decide on the most effective available treatment for each patient, at the right time.

What’s next?

We are investing in and advancing computational pathology, because we know that if we are to one day eliminate cancer as a cause of death, identifying the right patients to treat with each medicine is just as important as our work to develop innovative new medicines. We need to champion the adoption of computational pathology as a standard practice in oncology to help us take full advantage of the advances being made in cancer treatments. This requires working with the global healthcare community to establish necessary computational pathology infrastructure and increase understanding and acceptance of its benefits.

Computational pathology has the potential to transform the landscape of cancer diagnosis, treatment and survivorship. It represents a new era in oncology.