FDA clears 3 modules of cloud-based platform for chest images

An approach for bringing artificial intelligence to computed tomography studies to help radiologists better interpret chest images has received federal clearance for use.

The Food and Drug Administration this week cleared three modules of AI-Rad Companion Chest CT from Siemens Healthineers. The modules—the first intelligent assistant of the company’s AI-Rad Companion platform—are intended to help radiologists interpret images of the thorax quickly and then automatically document the findings as structured reports.

The software thus enables automated enhanced visualization of CT images of the lungs, heart and aorta, automatically highlighting abnormalities and pathological findings. AI-Rad Companion Chest CT is a cloud-based solution that has been tested and validated for CT scanners from Siemens Healthineers, GE Healthcare and Philips Healthcare.

Siemens4-CROP.jpg

Although CT chest images are mainly assessed by radiologists with regard to the primary indication, these images contain additional clinically relevant information. Siemens Healthineers executives say the algorithms in AI-Rad Companion Chest CT were trained on extensive datasets and annotated by qualified clinical specialists to provide segmentation, measurement, and highlighting of key anatomical structures, to support quantitative and qualitative analysis.

Based on the AI-supported analysis, AI-Rad Companion Chest CT generates standardized, reproducible and quantitative reports in DICOM SC format. In addition to reducing time spent on manual results documentation, these reports can be accessed by radiologists on the picture archiving and communication system (PACS) in the clinical routine. The algorithms also highlight potentially clinically relevant changes that might otherwise remain unnoticed because they were not the primary indication for the exam.

Using CT images of the chest, the modules are able to differentiate among various structures in that region, highlight them individually, and mark and measure potential abnormalities, such as coronary calcifications. It supports a variety of tasks that include:

  • Automated detection of lesions, localization of abnormalities and measurement of lung lesions.
  • Quantification of per-lobe low-attenuation parenchyma.
  • Enhanced visualization of lung lesions.
  • Automated segmentation of lung lobes and enhanced visualization of low-attenuation parenchyma.
  • Segmentation and measurement of maximum diameters of the thoracic aorta.
  • Quantification of the total calcium volume in the coronary arteries.
  • Detection of nine anatomical landmarks as identified by American Heart Association guidelines.
For reprint and licensing requests for this article, click here.