Energy Efficiency of Quantized Neural Networks in Medical Imaging

Priyanshu Sinha, Sai Sreya Tummala, Saptarshi Purkayastha, Judy Gichoya

Keywords: medical imaging, segmentation, object detection, TPU, Jetson Nano, Apple M1

TL;DR: Comparing energy usage of different SoCs in medical imaging.

Abstract:

The main goal of this paper is to compare the energy efficiency of quantized neural networks to perform medical image analysis on different processors and neural network architectures. Deep neural networks have demonstrated outstanding performance in medical image analysis but require high computation and power usage. In our work, we review the power usage and temperature of processors when running Resnet and Unet architectures to perform image classification and segmentation respectively. We compare Edge TPU, Jetson
Nano, Apple M1, Nvidia Quadro P6000 and Nvidia A6000 to infer using full-precision FP32 and quantized INT8 models. The results will be useful for designers and implementers of medical imaging AI on hand-held or edge computing devices.

Registration: I acknowledge that acceptance of this work at MIDL requires at least one of the authors to register and present the work during the conference.

Authorship: I confirm that I am the author of this work and that it has not been submitted to another publication before.

Paper Type: novel methodological ideas without extensive validation

Primary Subject Area: Transfer Learning and Domain Adaptation

Secondary Subject Area: Segmentation

Confidentiality And Author Instructions: I read the call for papers and author instructions. I acknowledge that exceeding the page limit and/or altering the latex template can result in desk rejection.

Code And Data:

Source Code:
- https://github.com/pri2si17-1997/deep_models_energy_consumption

Datasets:
1. https://www.kaggle.com/datasets/mateuszbuda/lgg-mri-segmentation
2. https://www.kaggle.com/datasets/nih-chest-xrays/data
3. https://www.kaggle.com/competitions/ultrasound-nerve-segmentation/data