Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Embedded Deployment of Semantic Segmentation in Medicine through Low-Resolution Inputs

Erik Ostrowski, Muhammad Shafique

TL;DR

An extensive analysis is conducted to illustrate that the proposed architecture enhances existing state-of-the-art frameworks for lightweight semantic segmentation of cancer in MRI images and tests the deployment speed of state-of-the-art lightweight networks and the architecture on Nvidia's Jetson Nano to emulate deployment in resource-constrained embedded scenarios.

Abstract

When deploying neural networks in real-life situations, the size and computational effort are often the limiting factors. This is especially true in environments where big, expensive hardware is not affordable, like in embedded medical devices, where budgets are often tight. State-of-the-art proposed multiple different lightweight solutions for such use cases, mostly by changing the base model architecture, not taking the input and output resolution into consideration. In this paper, we propose our architecture that takes advantage of the fact that in hardware-limited environments, we often refrain from using the highest available input resolutions to guarantee a higher throughput. Although using lower-resolution input leads to a significant reduction in computing and memory requirements, it may also incur reduced prediction quality. Our architecture addresses this problem by exploiting the fact that we can still utilize high-resolution ground-truths in training. The proposed model inputs lower-resolution images and high-resolution ground truths, which can improve the prediction quality by 5.5% while adding less than 200 parameters to the model. %reducing the frames per second only from 25 to 20. We conduct an extensive analysis to illustrate that our architecture enhances existing state-of-the-art frameworks for lightweight semantic segmentation of cancer in MRI images. We also tested the deployment speed of state-of-the-art lightweight networks and our architecture on Nvidia's Jetson Nano to emulate deployment in resource-constrained embedded scenarios.

Embedded Deployment of Semantic Segmentation in Medicine through Low-Resolution Inputs

TL;DR

An extensive analysis is conducted to illustrate that the proposed architecture enhances existing state-of-the-art frameworks for lightweight semantic segmentation of cancer in MRI images and tests the deployment speed of state-of-the-art lightweight networks and the architecture on Nvidia's Jetson Nano to emulate deployment in resource-constrained embedded scenarios.

Abstract

When deploying neural networks in real-life situations, the size and computational effort are often the limiting factors. This is especially true in environments where big, expensive hardware is not affordable, like in embedded medical devices, where budgets are often tight. State-of-the-art proposed multiple different lightweight solutions for such use cases, mostly by changing the base model architecture, not taking the input and output resolution into consideration. In this paper, we propose our architecture that takes advantage of the fact that in hardware-limited environments, we often refrain from using the highest available input resolutions to guarantee a higher throughput. Although using lower-resolution input leads to a significant reduction in computing and memory requirements, it may also incur reduced prediction quality. Our architecture addresses this problem by exploiting the fact that we can still utilize high-resolution ground-truths in training. The proposed model inputs lower-resolution images and high-resolution ground truths, which can improve the prediction quality by 5.5% while adding less than 200 parameters to the model. %reducing the frames per second only from 25 to 20. We conduct an extensive analysis to illustrate that our architecture enhances existing state-of-the-art frameworks for lightweight semantic segmentation of cancer in MRI images. We also tested the deployment speed of state-of-the-art lightweight networks and our architecture on Nvidia's Jetson Nano to emulate deployment in resource-constrained embedded scenarios.
Paper Structure (5 sections, 6 equations, 8 figures, 4 tables)

This paper contains 5 sections, 6 equations, 8 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Our Framework
  4. Results and Discussion
  5. Conclusion

Figures (8)

  • Figure 1: Trade-off between the number of Giga Multiply-Accumulate (GMAC) Operations and Jaccard prediction score of our architecture, U-Net UNET and state-of-the-art ELU-Net ELUNET
  • Figure 2: The novelty of our architecture lies in its ability to generate high resolution predictions with very compressed inputs.
  • Figure 3: Overview of our architecture. Our architecture extends the conventional "U" shape of the network to produce higher-resolution outputs. In training, we compare the output at different sizes to the correspondingly reshaped ground truths.
  • Figure 4: Detailed illustration of our architecture up-scaling layers with skip connections for the case of a $16 \times 16$ input image with one class.
  • Figure 5: A comparison between input resolution and prediction quality in Jaccard of the different networks on Decathlon.
  • ...and 3 more figures