Analog In-Memory Computing with Uncertainty Quantification for Efficient Edge-based Medical Imaging Segmentation
This addresses the need for power-efficient and robust medical imaging segmentation at the edge, though it appears incremental as it builds on existing AIMC paradigms with specific architectural comparisons.
This work tackles the problem of enabling efficient and certain medical AI analysis at the edge by investigating Analog In-memory computing (AIMC), showing that isotropic architectures have a minimal accuracy drop of 0.04 in analog-aware training compared to up to 0.15 for pyramidal structures in tasks like brain tumor analysis and spleen segmentation.
This work investigates the role of the emerging Analog In-memory computing (AIMC) paradigm in enabling Medical AI analysis and improving the certainty of these models at the edge. It contrasts AIMC's efficiency with traditional digital computing's limitations in power, speed, and scalability. Our comprehensive evaluation focuses on brain tumor analysis, spleen segmentation, and nuclei detection. The study highlights the superior robustness of isotropic architectures, which exhibit a minimal accuracy drop (0.04) in analog-aware training, compared to significant drops (up to 0.15) in pyramidal structures. Additionally, the paper emphasizes IMC's effective data pipelining, reducing latency and increasing throughput as well as the exploitation of inherent noise within AIMC, strategically harnessed to augment model certainty.