AI can diagnose breast cancer in mammograms better than licensed radiologists suggest recent studies. AI would analyze the images of scans and detect masses, breast density and mass segmentation better than professionals.
A new computer tool, called Mia, developed by Kheiron Medical Technologies and Imperial College London, can find more breast cancers in screenings than humans. The tool was tested in a European healthcare setting and showed that it could spot up to 13% more cases of breast cancer early. This information comes from a study titled “Prospective implementation of AI-assisted screen reading to improve early detection of breast cancer,” published in Nature Medicine.
Breast Cancer is a Significant Health Concern
Screening at the earliest stages is supposed to result in overcoming the disease. But of those who get screened 20% cases are missed through conventional screening methods.
While mortality rates have been decreasing in many developed countries due to advances in early detection and treatment, breast cancer still contributes significantly to cancer-related deaths. In some regions, it remains a leading cause of cancer mortality among women. Various factors contribute to the risk of developing breast cancer, including age, gender, family history, genetic mutations (such as BRCA1 and BRCA2), hormonal factors, and lifestyle choices.
Increased awareness and early detection through regular screening, mammograms, and self-examinations have been crucial in improving outcomes. However, access to screening and healthcare resources can vary, impacting the ability to detect and treat breast cancer early.
Research on Breast Cancer Screening Using AI
Previously, in an extensive examination involving thousands of mammograms, AI algorithms demonstrated superior performance compared to the conventional clinical risk model in forecasting the five-year risk for breast cancer. The findings from this research were disseminated in the Radiology journal.
Traditionally, a woman’s susceptibility is assessed through established clinical models like the Breast Cancer Surveillance Consortium (BCSC) risk model. This model relies on various factors such as self-reported data, patient age, family history of the disease, childbirth history, and breast density to compute a comprehensive risk score.
The shift towards AI-driven risk assessment in this context implies a more nuanced and potentially accurate prediction of breast cancer risk. This advancement could have significant implications for early detection and intervention, potentially improving outcomes for individuals at risk. The integration of AI into healthcare practices, as evidenced by this study published in Radiology, reflects a growing trend towards leveraging technology to enhance diagnostic and prognostic capabilities in the field of medical research and patient care.
According to Dr. Arasu, in an article from the Radiological Society of North America, “AI for cancer risk prediction offers us the opportunity to individualize every woman’s care, which isn’t systematically available,” he said. “It’s a tool that could help us provide personalized, precision medicine on a national level.” The Lancet has also stated that AI has been proposed to reduce false positive cases in the screening and flag the false negatives.
Concerns raised in Using AI in Cancer Diagnosis
AI relies on data from diverse populations, but disparities may arise due to variations in socioeconomic conditions. Cancer incidences differ among races, adding complexity. Studies assessing AI’s effectiveness should have transparent outcomes for credibility. To gain acceptance, AI must be replicable, requiring a shared code accessible to everyone. Equal data sharing is crucial for achieving this common ground.
Using AI in healthcare, raises ethical concerns like data confidentiality, privacy, patient autonomy, and consent. Measures and legislation are in place to prevent violations. Additionally, the use of radiomics in regular clinical practice is not widespread, and most studies on this topic are small and look back at past cases. This makes the current findings less reliable or trustworthy.
