AI's Emerging Role in Advanced Medical Diagnostics Highlighted at International Myeloma Society Meeting
Rio De Janeiro, Brazil – September 25, 2024
The application of artificial intelligence (AI) in the realm of modern medicine holds significant promise, especially in the advanced screening, diagnostics, and treatment of diseases. This was a key focus at the recent International Myeloma Society 21st Annual Meeting, where notable biomedical engineering professor, Anant Madabhushi, PhD, from Emory University and the Georgia Institute of Technology, presented groundbreaking insights and future opportunities for AI in the treatment of multiple myeloma (MM).
Madabhushi's discussion underscored the potential of AI to improve diagnosis, prognosis, and treatment response predictors. In his presentation, he highlighted the issue of overdiagnosis and the subsequent overtreatment of diseases such as prostate cancer, emphasising the financial and physical toll aggressive treatments impose on patients. For instance, more than 42% of cancer patients in a study published in the American Journal of Medicine reportedly depleted their life savings within two years of their diagnosis due to treatment costs.
Madabhushi showcased how AI could mitigate such issues by extracting intricate tumor characteristics, such as proteomic and genomic data, alongside mappings of tumor architecture. Machine learning models can then assimilate this information to create better predictors for diagnostic and therapeutic outcomes, potentially reducing the need for invasive methods and promoting more targeted treatments.
"Initially, we were astounded by how powerful these algorithms could be," Madabhushi revealed. "We stood up an AI algorithm to identify cancer cells on new, unseen images successfully. However, the challenge lies in their 'black box' nature, where the exact workings of the models remain opaque."
This "black box" phenomenon presents a significant obstacle, making reproducibility a challenge across diverse clinical settings and data sets. AI models sometimes produce compelling but inaccurate outputs, indicating a pressing need for ongoing refinement and integration of clear, interpretable algorithms. Including AI tools in blinded clinical trials could generate high-quality evidence necessary for advancing these technologies.
Madabhushi elaborated on AI’s broader diagnostic and predictive capabilities beyond multiple myeloma. He cited a study involving patients with early-stage breast cancer where AI analysis of breast pathology images differentiated between low-risk and high-risk patients by examining collagen fibre patterns in the tumor microenvironment. This AI-driven differentiation suggested some patients who didn't receive necessary chemotherapy could benefit from it, while others may have been spared from excessive treatment.
Highlighting MM-specific findings, Madabhushi’s team identified that the tortuosity of blood vessels in eye fundus images could predict MM onset within a decade. This insight points to the potential of routine eye imaging and AI applications to identify at-risk patients early. Moreover, efforts are ongoing to leverage machine learning for the rapid prediction of M-spike values, a critical biomarker in MM monitoring.
Madabhushi stressed that AI-driven advancements must be equitable and accessible globally, particularly in low-middle-income countries. "We need to develop low-cost tools that can have a substantial impact and ensure these AI systems are effective across diverse populations," he concluded.
The potential for AI in MM treatment is vast yet remains in nascent stages. Continued investigation and development are essential to harness its capabilities fully, ensuring it provides precise, actionable insights and addresses treatment disparities and equity issues effectively.
Source: Noah Wire Services