The study, promoted by the non-profit organization The Women’s Brain Project and supported by the Bioinfo4Women programme at the BSC, has been published today in Nature Digital Medicine.
The application of artificial intelligence (AI) to the biomedical sector is leading us to a better understanding of human diseases, facilitating their prevention, diagnosis, personalised treatments and, in general, precision medicine.
The success of precision medicine largely depends on overcoming a number of challenges, many of which are inherently related to the responsible use of AI in research and healthcare. To meet these ambitious objectives, it is essential to account for the risks of neglecting differences among individuals that reflect the clinical characteristics of diseases and drugs response, such as sex and gender differences.
This call to responsibility is one of the main conclusions of the study led by Davide Cirillo, researcher of the Computational Biology group led by Alfonso Valencia, director of the Life Sciences department at the Barcelona Supercomputing Center (BSC), in collaboration with María José Rementería, head of the BSC Social Link Analytics group. The study, published today in Nature Digital Medicine, is co-authored by Silvina Catuara-Solarz, from Telefónica Innovation Alpha Health, and scientists from many other centers, such as the Universitat Oberta Catalunya (UOC), the Biomedical Informatics Research Program (GRIB), the Interactive Robots and Media Laboratory (IRML), among others. The study is promoted by the non-profit organization The Women’s Brain Project and supported by the Bioinfo4Women programme at the BSC.
The article investigates sex and gender biases both in the data used to train the algorithms and in their development and subsequent use, providing solutions and specific actions for researchers and developers to enhance awareness and eradicate biases from biomedical applications. "If we do not correctly capture differences between individuals in our algorithms, artificial intelligence will not meet the challenges of precision medicine but instead it will amplify and consolidate biases," says Davide Cirillo.
In the text, the authors highlight the need to develop tools that allow the detection and mitigation of biases in data and algorithms, while achieving explainability and interpretation of outcomes that can directly affect the well-being of the population.
The researchers involved in this analysis warn about the need for the community in its entirety, including governments and policy makers, to become involved in addressing the ethical issues associated with each stage of the technological development for health.
Article: Sex and gender differences and biases in artificial intelligence for biomedicine and healthcare
DOI: 10.1038/s41746-020-0288