Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AACLiteNet: A Lightweight Model for Detection of Fine-Grained Abdominal Aortic Calcification

Zaid Ilyas, Afsah Saleem, David Suter, Siobhan Reid, John Schousboe, William Leslie, Joshua Lewis, Syed Zulqarnain Gilani

TL;DR

The proposed AACLiteNet is a lightweight deep learning model that predicts both cumulative and granular level AAC scores with high accuracy, and also has a low memory footprint, and computation cost (Floating Point Operations (FLOPs).

Abstract

Cardiovascular Diseases (CVDs) are the leading cause of death worldwide, taking 17.9 million lives annually. Abdominal Aortic Calcification (AAC) is an established marker for CVD, which can be observed in lateral view Vertebral Fracture Assessment (VFA) scans, usually done for vertebral fracture detection. Early detection of AAC may help reduce the risk of developing clinical CVDs by encouraging preventive measures. Manual analysis of VFA scans for AAC measurement is time consuming and requires trained human assessors. Recently, efforts have been made to automate the process, however, the proposed models are either low in accuracy, lack granular level score prediction, or are too heavy in terms of inference time and memory footprint. Considering all these shortcomings of existing algorithms, we propose 'AACLiteNet', a lightweight deep learning model that predicts both cumulative and granular level AAC scores with high accuracy, and also has a low memory footprint, and computation cost (Floating Point Operations (FLOPs)). The AACLiteNet achieves a significantly improved one-vs-rest average accuracy of 85.94% as compared to the previous best 81.98%, with 19.88 times less computational cost and 2.26 times less memory footprint, making it implementable on portable computing devices.

AACLiteNet: A Lightweight Model for Detection of Fine-Grained Abdominal Aortic Calcification

TL;DR

The proposed AACLiteNet is a lightweight deep learning model that predicts both cumulative and granular level AAC scores with high accuracy, and also has a low memory footprint, and computation cost (Floating Point Operations (FLOPs).

Abstract

Cardiovascular Diseases (CVDs) are the leading cause of death worldwide, taking 17.9 million lives annually. Abdominal Aortic Calcification (AAC) is an established marker for CVD, which can be observed in lateral view Vertebral Fracture Assessment (VFA) scans, usually done for vertebral fracture detection. Early detection of AAC may help reduce the risk of developing clinical CVDs by encouraging preventive measures. Manual analysis of VFA scans for AAC measurement is time consuming and requires trained human assessors. Recently, efforts have been made to automate the process, however, the proposed models are either low in accuracy, lack granular level score prediction, or are too heavy in terms of inference time and memory footprint. Considering all these shortcomings of existing algorithms, we propose 'AACLiteNet', a lightweight deep learning model that predicts both cumulative and granular level AAC scores with high accuracy, and also has a low memory footprint, and computation cost (Floating Point Operations (FLOPs)). The AACLiteNet achieves a significantly improved one-vs-rest average accuracy of 85.94% as compared to the previous best 81.98%, with 19.88 times less computational cost and 2.26 times less memory footprint, making it implementable on portable computing devices.
Paper Structure (4 sections, 4 equations, 4 figures, 2 tables)

This paper contains 4 sections, 4 equations, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Proposed Framework
  3. Experiments and Results
  4. Conclusion

Figures (4)

  • Figure 1: (a) The red arrows point to calcified pixels. (b) Two examples of AAC scoring. (c) DXA image with Poorly demarcated vertebral boundaries. (d) Use of guides for AAC measurement using Kauppila AAC-24 Point Scoring Method r17.
  • Figure 2: (a) CNN Encoder with multiple 2D simple and depthwise convolution layers. (b) Granular Output Block with four outputs each for anterior and posterior sections of one vertebra. (c) SAM Block for Self-Attention. (d) GFFM Blocks for Channel Attention using Gating Mechanism.
  • Figure 3: (a) Average one-vs-rest accuracy vs FLOPs. (b) Average one-vs-rest accuracy vs parameter count. (c) Confusion Matrix Reid et al. r15 (d) Confusion Matrix Gilani et al. r16 (e) Confusion Matrix AACLiteNet.
  • Figure 4: (a) Pearson Correlation of AACLiteNet AAC score with human ground truth. (b) ROC Curves for MACE prediction.