AI model offers predictive insights for breast cancer progression

AI Model Offers Predictive Insights for Breast Cancer Progression

A groundbreaking development in the realm of oncology suggests significant advancements in the predictive analysis of ductal carcinoma in situ (DCIS), a preinvasive form of breast cancer. Researchers at ETH Zurich, in collaboration with experts at MIT, have harnessed the power of artificial intelligence (AI) to predict the progression of DCIS to more invasive stages of breast cancer.

DCIS is known to account for approximately 25% of breast cancer diagnoses. However, the uncertainty over which cases will escalate to invasive cancer presents a considerable challenge. Currently, about 30% to 50% of DCIS cases progress to invasive breast carcinoma. The inherent difficulty for clinicians in staging and identifying the precise type of DCIS results in the risk of overtreatment as a precautionary measure.

The research team, led by G.V. Shivashankar, PhD, from the department of health sciences and technology at ETH Zurich, introduced a promising AI model that capitalises on chromatin image analysis. This technique uses inexpensive and readily obtainable chromatin images to ascertain the DCIS stage accurately. The method offers a cost-effective alternative to more complex and expensive procedures such as multiplexed staining and single-cell RNA sequencing.

The AI model utilises an extensive database, comprising chromatin images from 560 samples collected from 122 women diagnosed with DCIS. This dataset allows the model to identify eight distinct cell states within the samples, which serves as indicators of potential progression to invasive cancer. Furthermore, the model assesses the organisational patterns of cells in the tissue microenvironment to predict disease advancement.

Xinyi Zhang, a graduate student involved in the study at MIT’s department of electrical engineering and computer science, elaborated on the AI’s functionality. “It extracts features of the neighbourhood of cells, as well as genome packing, to predict the different stages of DCIS,” Zhang explained. The AI’s analysis was cross-validated by a senior pathologist, ensuring its findings were congruent with expert evaluations in many cases. Significantly, in more ambiguous situations, the AI provided cell maintenance and distribution insights that could aid pathologists in making informed treatment decisions.

The implications of this research extend beyond breast cancer. Caroline Uhler, PhD, a prominent figure at MIT and the Broad Institute, suggested potential applications of the AI model in other cancer types and neurodegenerative diseases. However, Uhler noted that before widespread clinical implementation can occur, substantial large-scale clinical trials are necessary to confirm the efficacy of this AI-assisted approach.

The AI model represents a promising technique in cancer diagnostics, owing to its affordability and the depth of information it provides. By potentially reducing unnecessary treatments and focusing efforts on cases with a higher likelihood of progression, this method could mark a significant shift in how early breast cancer is managed. Future tests and regulatory approvals will determine its role in clinical settings.

Source: Noah Wire Services