Reducing the Hausdorff Distance in Medical Image Segmentation with\n Convolutional Neural Networks

2019/04/22 by Davood Karimi, Septimiu E. Salcudean, Karimi, Davood +1 · 10 citations
Computer Science · Medicine · #Advanced Neural Network Applications #FOS: Computer and information sciences #FOS: Electrical engineering #Image and Video Processing (eess.IV) #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Medical Image Segmentation Techniques #Radiomics and Machine Learning in Medical Imaging #electronic engineering #information engineering

paper · pdf · doi:10.48550/arxiv.1904.10030

openalex publication_date 2019/04/22 · openalex created_date 2022/07/24 · openalex updated_date 2026/07/28

Abstract

The Hausdorff Distance (HD) is widely used in evaluating medical image\nsegmentation methods. However, existing segmentation methods do not attempt to\nreduce HD directly. In this paper, we present novel loss functions for training\nconvolutional neural network (CNN)-based segmentation methods with the goal of\nreducing HD directly. We propose three methods to estimate HD from the\nsegmentation probability map produced by a CNN. One method makes use of the\ndistance transform of the segmentation boundary. Another method is based on\napplying morphological erosion on the difference between the true and estimated\nsegmentation maps. The third method works by applying circular/spherical\nconvolution kernels of different radii on the segmentation probability maps.\nBased on these three methods for estimating HD, we suggest three loss functions\nthat can be used for training to reduce HD. We use these loss functions to\ntrain CNNs for segmentation of the prostate, liver, and pancreas in ultrasound,\nmagnetic resonance, and computed tomography images and compare the results with\ncommonly-used loss functions. Our results show that the proposed loss functions\ncan lead to approximately 18-45 % reduction in HD without degrading other\nsegmentation performance criteria such as the Dice similarity coefficient. The\nproposed loss functions can be used for training medical image segmentation\nmethods in order to reduce the large segmentation errors.\n

Cited by

• Agentic AI in medicine: architectures, applications, evaluation, and challenges for clinical translation
• Model Agnostic Preference Optimization for Medical Image Segmentation
• Hard Spatial Gating for Precision-Driven Brain Metastasis Segmentation: Addressing the Over-Segmentation Paradox in Deep Attention Networks
• FocusSDF: Boundary-Aware Learning for Medical Image Segmentation via Signed Distance Supervision
• Towards the Automatic Segmentation, Modeling and Meshing of the Aortic Vessel Tree from Multicenter Acquisitions: An Overview of the SEG.A. 2023 Segmentation of the Aorta Challenge
• AI-powered printability evaluation framework for 3D bioprinting using Hausdorff distance metrics
• Prostate Capsule Segmentation from Micro-Ultrasound Images using Adaptive Focal Loss
• Artificial Intelligence–Based Approaches for Brain Tumor Segmentation in MRI: A Review
• XBusNet: Text-Guided Breast Ultrasound Segmentation via Multimodal Vision-Language Learning
• Comparison of Segmentation Methods in Remote Sensing for Land Use Land Cover

Related

• A Generalized Surface Loss for Reducing the Hausdorff Distance in Medical Imaging Segmentation
• High-level Prior-based Loss Functions for Medical Image Segmentation: A Survey
• Distance Map Loss Penalty Term for Semantic Segmentation
• Improving Medical Image Segmentation Using Machine Learning: The Role of Optimization of Extracted Features in Convolutional Neural Networks
• V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation