Synthetic Data Generated from CT Scans for Patient Pose Assessment

Manuel Laufer; Dominik Mairhöfer; Malte Maria Sieren; Hauke Gerdes; Fabio Leal dos Reis; Árpád Bischof; Thomas Käster; Erhardt Barth; Jörg Barkhausen; Thomas Martinetz

Synthetic Data Generated from CT Scans for Patient Pose Assessment

Manuel Laufer, Dominik Mairhöfer, Malte Maria Sieren, Hauke Gerdes, Fabio Leal dos Reis, Árpád Bischof, Thomas Käster, Erhardt Barth, Jörg Barkhausen, Thomas Martinetz

Abstract

An adequate diagnostic quality of radiographs is essential for reliable diagnoses and treatment planning. The patient's pose during radiography is one of the most important factors determining the diagnostic quality. Since patient positioning is difficult and not standardized, an automated AI-based approach using depth images to automatically assess the patient's pose before the radiograph has been taken would be helpful. Due to regulatory hurdles, however, it is difficult in practice to acquire the required depth images and corresponding radiographs. In this paper, we present a framework that can generate such training data synthetically from Computer Tomography scans. We further show that by pretraining on our generated synthetic dataset consisting of 3077 image pairs of upper ankle joints, the pose assessment of real upper ankle joints can be improved by up to 11 percentage points.

Original language	German
Number of pages	14
Publication status	Published - 27.03.2025
Event	MIDL 2025: Medical Imaging with Deep Learning - University of Utah, Salt Lake City, United States Duration: 09.07.2025 → 11.07.2025 https://2025.midl.io/

Conference

Conference	MIDL 2025
Country/Territory	United States
City	Salt Lake City
Period	09.07.25 → 11.07.25
Internet address	https://2025.midl.io/

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 3 Good Health and Well-being
SDG 9 Industry, Innovation, and Infrastructure

Research Areas and Centers

Centers: Center for Artificial Intelligence Luebeck (ZKIL)

DFG Research Classification Scheme

2.22-33 Nuclear Medicine, Radiotherapy, Radiobiology
2.22-32 Medical Physics, Biomedical Technology
4.43-04 Artificial Intelligence and Machine Learning Methods

Cite this

@conference{a3c9b067e9c842e3a5a90721d06390cc,

title = "Synthetic Data Generated from CT Scans for Patient Pose Assessment",

abstract = "An adequate diagnostic quality of radiographs is essential for reliable diagnoses and treatment planning. The patient's pose during radiography is one of the most important factors determining the diagnostic quality. Since patient positioning is difficult and not standardized, an automated AI-based approach using depth images to automatically assess the patient's pose before the radiograph has been taken would be helpful. Due to regulatory hurdles, however, it is difficult in practice to acquire the required depth images and corresponding radiographs. In this paper, we present a framework that can generate such training data synthetically from Computer Tomography scans. We further show that by pretraining on our generated synthetic dataset consisting of 3077 image pairs of upper ankle joints, the pose assessment of real upper ankle joints can be improved by up to 11 percentage points.",

author = "Manuel Laufer and Dominik Mairh{\"o}fer and Sieren, \{Malte Maria\} and Hauke Gerdes and \{Leal dos Reis\}, Fabio and {\'A}rp{\'a}d Bischof and Thomas K{\"a}ster and Erhardt Barth and J{\"o}rg Barkhausen and Thomas Martinetz",

year = "2025",

month = mar,

day = "27",

language = "Deutsch",

note = "MIDL 2025 : Medical Imaging with Deep Learning ; Conference date: 09-07-2025 Through 11-07-2025",

url = "https://2025.midl.io/",

}

TY - CONF

T1 - Synthetic Data Generated from CT Scans for Patient Pose Assessment

AU - Laufer, Manuel

AU - Mairhöfer, Dominik

AU - Sieren, Malte Maria

AU - Gerdes, Hauke

AU - Leal dos Reis, Fabio

AU - Bischof, Árpád

AU - Käster, Thomas

AU - Barth, Erhardt

AU - Barkhausen, Jörg

AU - Martinetz, Thomas

PY - 2025/3/27

Y1 - 2025/3/27

N2 - An adequate diagnostic quality of radiographs is essential for reliable diagnoses and treatment planning. The patient's pose during radiography is one of the most important factors determining the diagnostic quality. Since patient positioning is difficult and not standardized, an automated AI-based approach using depth images to automatically assess the patient's pose before the radiograph has been taken would be helpful. Due to regulatory hurdles, however, it is difficult in practice to acquire the required depth images and corresponding radiographs. In this paper, we present a framework that can generate such training data synthetically from Computer Tomography scans. We further show that by pretraining on our generated synthetic dataset consisting of 3077 image pairs of upper ankle joints, the pose assessment of real upper ankle joints can be improved by up to 11 percentage points.

AB - An adequate diagnostic quality of radiographs is essential for reliable diagnoses and treatment planning. The patient's pose during radiography is one of the most important factors determining the diagnostic quality. Since patient positioning is difficult and not standardized, an automated AI-based approach using depth images to automatically assess the patient's pose before the radiograph has been taken would be helpful. Due to regulatory hurdles, however, it is difficult in practice to acquire the required depth images and corresponding radiographs. In this paper, we present a framework that can generate such training data synthetically from Computer Tomography scans. We further show that by pretraining on our generated synthetic dataset consisting of 3077 image pairs of upper ankle joints, the pose assessment of real upper ankle joints can be improved by up to 11 percentage points.

M3 - Abstracts

T2 - MIDL 2025

Y2 - 9 July 2025 through 11 July 2025

ER -