This study characterizes the performance of AI-based software that measures the quality of patient positioning for chest radiographs.We compared expert (subjective) evaluation with algorithmic (objective) classification of three important aspects of image quality (field-of-view [FOV], inhalation (INH) and thorax rotation (ROT)), using thorax radiographs selected from routine clinical imaging. The aim was to quantify the extent to which the software can reproduce expert classifications, and therefore understand how reliably the software can detect images exhibiting an unacceptable level of patient positioning errors while correctly assessing...

We designed three AI-modules based upon radiological image quality guidelines [1]. These modules are described in more detail in a complementary abstract [2]), and measure the following qualityparametersin chest PA radiographs: collimation – the distances from the lung field to the image border in all four directions (cranial Δc, abdominal Δa, left Δl and right Δr), axial rotation via the symmetry of clavicle head positions w.r.t. medial axis of the thorax, α= (dL – dR) / (dL + dR) degree of inhalation via evaluation of...

The main reason for image re-takes (field-of-view “too-narrow”) was detected using the software for at least one edge of the image in 13.9% of the entire image set. The sensitivity in all cases with concordant ratings was 92.1%, 88.2% and 100.0% for the detection of errors rated as “unacceptable” for field-of-view (either “too-wide” or “too-narrow”), rotation and inhalation errors in 203, 34 and 42 images respectively. Concurrently, the specificity for each aspect was 87.6%, 70.0% and 73.5% for identification of images in which concordant expert...

Adequate positioning of the patient is the prerequisite for images of good diagnostic quality, and standardization of criteria for image quality and feedback to the technologist may help to improve department efficiency. The artificial intelligence modules provided accurate and reliable evaluation of the patient positioning quality. Accurate measurement of positioning quality for chest radiographs can support the improvement of clinical image quality. Demonstrating good correlation between software measurement and expert image evaluation enhances trust in this AI-based quality assessment approach.

S. Young; Hamburg/DE - Employee at Philips Research M. Kotnik; Šmartno pri Slovenj Gradcu/SI - nothing to disclose L. Lin; Leiden/NL - nothing to disclose M. Sevenster; Eindhoven/NL - Employee at Philips Research N. Wieberneit; Hamburg/DE - Employee at Philips Health Systems T. Harder; Hamburg/DE - Employee at Philips Research S. Krönke; Hamburg/DE - Employee at Philips Research D. Bystrov; Hamburg/DE - Employee at Philips Research H. J. Lamb; Leiden/NL - Research/Grant Support at Philips - research support

[1] M. Carmichael et al, European guidelines on quality criteria for diagnostic radiographic images, EUR 16260 EN, European Commission, 1996 [2] J. von Berg et al, AI-based positioning quality check for chest x-ray, Poster C-05625 ECR 2020 [3] M. Englmeier et al, Inter-rater variability in the assessment of positioning quality in chest x-ray images, Poster C-05601 ECR 2020