According to a study conducted by the Norwegian Institute of Public Health (FHI), years before a breast Cancer diagnosis is made, Artificial Intelligence (AI) can identify which women are more likely to have the disease. FHI, the University of California, and the University of Washington researchers examined the mammograms of 116,495 women who took part in a Norwegian breast cancer screening program from 2004 to 2018. A total of 1,607 of these women went on to develop breast cancer. The AI system was able to identify high-risk individuals.
Four to six years beforehand, the study also identified the breast that was most likely to be impacted. Lead researcher Solveig Hofvind observed that breasts with cancer had AI scores around twice as high as the other breast. This implies that AI can improve early intervention, strengthen personalised detection programs, lower costs, and more successfully target at-risk groups. The results demonstrate AI’s potential for breast cancer diagnosis and were published in the Journal of the American Medical Association Network.
There are obstacles to overcome to incorporate AI into clinical practice.
To test if AI can diagnose cancer as accurately as or better than radiologists, Norway started a different research in 2023 with 140,000 participants. Deep learning techniques are the main way that artificial intelligence (AI) has emerged as a crucial tool in the identification of breast cancer. Due to their thorough training on mammography picture datasets, these AI models are able to recognise patterns and anomalies that might be signs of cancer. Google and Northwestern Medicine, for example, collaborated to create an AI model that examined de-identified mammograms with biopsy-confirmed results.
Also, according to this model, false positives decreased by 5.7% and 1.2% in the U.S. and the U.K., respectively, and false negatives decreased by 9.4% and 2.7%, respectively. Even with these developments, there are still obstacles to overcome when incorporating AI into clinical practice. One significant issue is automation bias, which occurs when medical professionals depend too much on AI results, possibly missing diagnosis or doing needless operations. Less experienced radiologists were more likely to accept inaccurate AI recommendations, according to a research.
Ethical issues pertaining to data protection must be addressed.
Furthermore, this emphasises the necessity of thorough Education and awareness campaigns to lessen these prejudices. Practical challenges to AI integration in healthcare systems include the requirement for large amounts of high-quality data in order to properly train models. AI performance may be impacted by differences in imaging methods and equipment between institutions, hence thorough validation in a range of clinical contexts is required. To foster confidence between patients and healthcare practitioners, ethical issues pertaining to Data Protection and the interpretability of AI conclusions must also be addressed.
Even though AI has demonstrated the ability to match or even outperform human radiologists in specific diagnostic tasks, it is generally agreed that AI should be used in conjunction with human expertise rather than in substitute of it. In a study comparing AI and radiologists’ performance, the AI system’s area under the receiver operating characteristic curve (AUROC) was 0.962, whereas radiologists’ average was 0.924. The best course of action, however, is a cooperative model in which AI helps radiologists, improving the precision and effectiveness of diagnosis. Globally, the use of AI in breast cancer screening is growing in popularity.
Related Article: AI expert unveils health-focused chatbot
AI-assisted screenings, for instance, identified breast cancer at a rate of 6.7%, compared to 5.7% in routine screenings, without raising false positives, according to a research conducted in Germany with over 260,000 women. This implies that AI can lessen radiologists’ workloads and increase detection rates. Applications of AI are spreading beyond breast imaging into other medical specialities where pattern recognition is essential, such as pathology and dermatology. As artificial intelligence (AI) systems develop further, their incorporation into several facets of healthcare could increase the precision of diagnoses, customise treatment regimens, and eventually improve patient outcomes.