Analog In-Memory Computing with Uncertainty Quantification for Efficient Edge-based Medical Imaging Segmentation
Abstract: 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.
- Bringing uncertainty quantification to the extreme-edge with memristor-based Bayesian neural networks. Nature Communications, 14(1), 11 2023. doi: 10.1038/s41467-023-43317-9. URL https://doi.org/10.1038/s41467-023-43317-9.
- Nucleus segmentation across imaging experiments: the 2018 Data Science Bowl. Nature Methods, 16(12):1247–1253, 10 2019. doi: 10.1038/s41592-019-0612-7. URL https://doi.org/10.1038/s41592-019-0612-7.
- Monai: An open-source framework for deep learning in healthcare, 2022.
- A 64-core mixed-signal in-memory compute chip based on phase-change memory for deep neural network inference. Nature Electronics, 6(9):680–693, 8 2023. doi: 10.1038/s41928-023-01010-1. URL https://doi.org/10.1038/s41928-023-01010-1.
- Swin UNETR: swin transformers for semantic segmentation of brain tumors in MRI images. In Alessandro Crimi and Spyridon Bakas (eds.), MICCAI, volume 12962, pp. 272–284, 2021.
- nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation. Nature Methods, 18(2):203–211, 12 2020. doi: 10.1038/s41592-020-01008-z. URL https://doi.org/10.1038/s41592-020-01008-z.
- A 40-nm, 2m-cell, 8b-precision, hybrid slc-mlc pcm computing-in-memory macro with 20.5 - 65.0tops/w for tiny-al edge devices. In 2022 IEEE International Solid-State Circuits Conference (ISSCC), volume 65, pp. 1–3, 2022. doi: 10.1109/ISSCC42614.2022.9731670.
- A flexible and fast pytorch toolkit for simulating training and inference on analog crossbar arrays. CoRR, abs/2104.02184, 2021. URL https://arxiv.org/abs/2104.02184.
- U-net: Convolutional networks for biomedical image segmentation. In Nassir Navab, Joachim Hornegger, William M. Wells III, and Alejandro F. Frangi (eds.), MICCAI, volume 9351, pp. 234–241, 2015.
- A large annotated medical image dataset for the development and evaluation of segmentation algorithms, 2019.
- A compute-in-memory chip based on resistive random-access memory. Nature, 608(7923):504–512, 8 2022. doi: 10.1038/s41586-022-04992-8. URL https://doi.org/10.1038/s41586-022-04992-8.
- Monolithically integrated rram- and cmos-based in-memory computing optimizations for efficient deep learning. IEEE Micro, 39(6):54–63, 2019. doi: 10.1109/MM.2019.2943047.
- Unet++: A nested u-net architecture for medical image segmentation. In Danail Stoyanov, Zeike Taylor, Gustavo Carneiro, Tanveer F. Syeda-Mahmood, Anne L. Martel, Lena Maier-Hein, João Manuel R. S. Tavares, Andrew P. Bradley, João Paulo Papa, Vasileios Belagiannis, Jacinto C. Nascimento, Zhi Lu, Sailesh Conjeti, Mehdi Moradi, Hayit Greenspan, and Anant Madabhushi (eds.), DLMIA- MICCAI, volume 11045, pp. 3–11, 2018.
Paper Prompts
Sign up for free to create and run prompts on this paper using GPT-5.
Top Community Prompts
Collections
Sign up for free to add this paper to one or more collections.