Artificial intelligence (AI) can be trained to detect whether or not a tissue image contains a tumor. However, until now, it has remained a mystery as to how it makes its decision. A team from Ruhr-Universität Bochum’s Research Center for Protein Diagnostics (PRODI) is working on a new approach that will make an AI’s decision transparent and thus trustworthy. The approach is described in the journal Medical Image Analysis, which was published online by the researchers led by Professor Axel Mosig.
Professor Andrea Tannapfel, head of the Institute of Pathology, oncologist Professor Anke Reinacher-Schick from the Ruhr-St. Universität’s Josef Hospital, and biophysicist and PRODI founding director Professor Klaus Gerwert collaborated on the study. The group developed a neural network, i.e. an AI, that can classify whether a tissue sample contains tumour or not. To this end, they fed the AI a large number of microscopic tissue images, some of which contained tumours, while others were tumour-free.
“Neural networks are initially a black box: it’s unclear which identifying features a network learns from the training data,” explains Axel Mosig. Unlike human experts, they lack the ability to explain their decisions. “However, for medical applications in particular, it’s important that the AI is capable of explanation and thus trustworthy,” adds bioinformatics scientist David Schuhmacher, who collaborated on the study.
Neural networks are initially a black box: it’s unclear which identifying features a network learns from the training data. Unlike human experts, they lack the ability to explain their decisions. However, for medical applications in particular, it’s important that the AI is capable of explanation and thus trustworthy.Axel Mosig
AI is based on falsifiable hypotheses
The Bochum team’s explainable AI is thus based on the only kind of meaningful statements known to science: falsifiable hypotheses. If a hypothesis is false, this must be demonstrated through an experiment. Artificial intelligence typically employs the principle of inductive reasoning: using concrete observations, i.e. training data, the AI develops a general model on the basis of which it evaluates all subsequent observations.
The underlying problem was described by philosopher David Hume 250 years ago and is easily illustrated: no matter how many white swans we see, we can never conclude from this data that all swans are white and no black swans exist. As a result, science employs what is known as deductive logic. In this approach, a general hypothesis is the starting point. For example, the hypothesis that all swans are white is falsified when a black swan is spotted.
Activation map shows where the tumour is detected
“At first glance, inductive AI and the deductive scientific method appear almost incompatible,” physicist Stephanie Schörner, who also contributed to the study, says. However, the researchers discovered a solution. Their novel neural network not only classifies whether a tissue sample contains or is free of tumors, but it also generates an activation map of the microscopic tissue image.
The activation map is based on a testable hypothesis, namely that the neural network activation corresponds exactly to the tumor regions in the sample. This hypothesis can be tested using site-specific molecular methods.
“Thanks to the interdisciplinary structures at PRODI, we have the best prerequisites for incorporating the hypothesis-based approach into the development of trustworthy biomarker AI in the future, for example to be able to distinguish between certain therapy-relevant tumour subtypes,” concludes Axel Mosig.
The business world’s unyielding competition forces 86% of executives to rely on the power of Artificial Intelligence solutions. Artificial Intelligence (AI) encompasses not only the automation of numerous recurring processes, but also influences decision-making.
AI provides businesses with a full range of information and structured solutions to arising issues by eliminating human errors and analyzing vast amounts of various data quickly and continuously.