According to the study published in European cancer journalThe fairness and equity of the data set for the interpretation of the AI -centered breast photography can be in jeopardy by underestimating racial and ethnic diversity.
The AI shows a promise that can be improved in the area where resources are limited, but the author of the study found a warning signal on the diversity of the data set and the development of the AI model, which could affect the generalization, fairness and equity of the model.
For this study, the researchers conducted a scientific review of the studies published in 2017, 2018, 2022, and 2023 on screening or diagnostic breast photography for breast cancer detection.
Of the 5,774 studies confirmed, 264 people met the criteria.
The authors of the research said, “The number of research increased from 28 to 28 to 2022 to 2023. -311%.
In addition, almost all patient cohorts are derived from high -income countries and have no research in low -income environments. Author alliance is mainly from high -income areas, and sex imbalances were observed between the first and last authors. “
The authors concluded that “the lack of race, ethnic and geographical diversity in both data sets and researchers can weaken the generalization and fairness of AI -based breast photography interpretation.”
In addition, recognizing imbalance through various data sets and comprehensive international cooperation is important for ensuring the fair development of breast cancer treatment.
Research data shows that algorithms focused mainly on the white population can lead to incorrect diagnosis in poor results and poor population. In addition, patient results can be a threat and the current imbalance may worsen.
“The fairness of these AI tools is doubtful because there is a risk of systematically interfering with a specific race, ethnic or social population statistical group. It is essential to prioritize the diversity of the data set collection in order to alleviate these problems and to distribute the advantage of AI equally in BC imaging, and to prioritize the data set collection. It is essential to decide the ranking.
Greater trend
In February Google has partnered with the Women’s Cancer Research Institute, founded by Institute Curie, a French cancer research and treatment center, and studied how AI tools deal with cancer, share science -based health information, and support post -doctoral researchers.
The two objects have investigated how AI -based tools will help to develop cancer and predict the possibility of recurrence of cancer to develop more accurate and successful treatment.
The researchers focused on treating female cancer, including triple negative breast cancer and triple -voice breast cancer, an aggressive breast cancer that grows faster than other types.
2024 AI BIOTECH Company Owkin has partnered with Pharma Giant Astrazeneca and developed an AI -based tool designed to pre -screen GBRCA mutation (GBRCAM) of breast cancer directly from digitized pathological slides.
The goal of this tool is to increase and approach the GBRCA inspection speed that may not consider some patients.
Same year, LUNIT, an AI -based solution provider for cancer diagnosis and treatment, provides Volpara Health, a company that provides AI -based software that provides AI -based software to help you understand cancer risks, and develops a comprehensive ecosystem for early cancer detection, cancer risk prediction and independent AI. Improved.
In May of that year, LUNIT acquired Volpara and integrated AI breast health platforms, including score card breast density evaluation tools, into the AI tool line for breast cancer detection.
Lunit before acquiring Volpara Increase your Swedish cancer screening ability.
In 2023, LUNIT signed a three -year contract with Capio S: T Göran Hospital, providing and licensing AI -based breast ancient analysis software Lunit Insight MMG. The AI tool allowed the hospital to analyze the breast image of about 78,000 patients each year.
