A breakthrough breast cancer diagnosis system could help clinicians understand the rationale behind artificial intelligence (AI) decisions.

Researchers from Charité University Hospital in Berlin, the Berlin Institute of Technology and the University of Oslo have developed the new tissue-section analysis system for diagnosing breast cancer based on AI.  

The system uses heatmaps to show what visual information influenced the AI decision process and to what extent, enabling doctors to understand and assess the plausibility of the results.

Also, for the first time, morphological, molecular and histological data are integrated in a single analysis.

The results of the research have now been published in Nature Machine Intelligence and the researchers are now seeking certification and further clinical validations - including tests in tumour routine diagnostics.

WHY IT MATTERS

Machine learning algorithms that are not transparent about to how they have come to a decision have been dubbed ‘blackbox’ algorithms.

It can be difficult for clinicians and regulators to trust in AI tools that make crucial decisions without explaining the rationale followed. The researchers say their new tissue-section analysis system can help solve this problem, representing “a decisive and essential step forward for the future regular use of AI systems in hospitals.”

THE LARGER CONTEXT

Meanwhile, GE Healthcare and Turkish digital healthcare firm CUREA are developing AI-based applications for the automatic detection and classification of breast lesions through contrast-enhanced magnetic resonance imaging (CESM). The new tools aim to aggregate, standardise, and make sense of data quickly.

ON THE RECORD

Professor Dr Klaus-Robert Müller, professor of machine learning at TU Berlin, said: “The problem we have is the following: we have good and reliable molecular data and we have good histological data with high spatial detail. What we don’t have as yet is the decisive link between imaging data and high-dimensional molecular data.

“Our system facilitates the detection of pathological alterations in microscopic images. Parallel to this, we are able to provide precise heatmap visualisations showing which pixel in the microscopic image contributed to the diagnostic algorithm and to what extent.

“Our analysis system has been trained using machine learning processes so that it can also predict various molecular characteristics, including the condition of the DNA, the gene expression as well as the protein expression in specific areas of the tissue, on the basis of the histological images.”

Professor Dr Frederick Klauschen of Charité’s Institute of Pathology, said: “We know that in the case of breast cancer, the number of immigrated immune cells, known as lymphocytes, in tumour tissue has an influence on the patient’s prognosis. There are also discussions as to whether this number has a predictive value - in other words if it enables us to say how effective a particular therapy is.

“The methods we have developed will make it possible in the future to make histopathological tumour diagnostics more precise, more standardised and qualitatively better.”