Att använda AI för att detektera bröstcancer : En explorativ studie kring användning av bildanalys inom svensk sjukvård

Sammanfattning: Breast cancer is the most common form of cancer for women around the world. In an attempt to decrease the mortality, women in Sweden between the ages of 40-74 years are called to regular mammography screenings to detect the disease as early as possible. Despite this, around 1400 die from the disease every year in Sweden. Every mammography image has to be analyzed by two radiologists. Despite this and regular screening, there are cases that go unnoticed. The factors that lessen the effectiveness of the system are that some cases go unnoticed and analyzing the mammography images is time consuming. This paper has investigated whether AI can be used to help solve these issues. Earlier research examines both of these aspects. Algorithms performing at approximately the same level of accuracy as radiologists and lessening the workload for examining radiologists has been developed [1]. This paper examined how to develop a similar simplified algorithm, how it can be implemented in healthcare and what the consequences of that would be. Hopefully, usage of similar technology will lead to a decrease in mortality and more accurate assessments. The study was conducted by interviewing two experts within the subject, and an attempt to develop an algorithm that through image analysis can classify tumours from mammography images.  The result shows that there is a big potential for using AI within healthcare, and by that enabling more accurate diagnosis and reducing mortality. During development of the algorithm a deeper understanding of the difficulties was given, such as the need for adequate processing power, processing and organization of image databases and the complexity in developing such a ML-algorithm for image analysis. The developed algorithm performed slightly better than random when detecting breast cancer on mammography images.