Skip to main content
. 2023 Mar 16;13:4336. doi: 10.1038/s41598-023-29949-3

Figure 2.

Figure 2

(a) Schematic Illustration of the Similar Patient Search Web Service. (1) The user selects an ROI on an axial CT slice using the host Application (syngo.via VA40A). The slice, along with the ROI, is sent to the cloud service (2). A deep-learning-based feature representation (fingerprint) for the slice/ROI combination is computed in the cloud (3). The fingerprint is then compared to pre-generated fingerprints from the reference database (4). The tool returns a webpage that shows the most similar image patches from the reference database and their corresponding diagnosis (5). The clinical reference content from Thieme is accessible by a single click on the disease detail page (6). (b) Example SPS result from the study: Left: Input Slice (case with Mounier Kuhn Syndrome) with the ROI chosen by the study participant (orange outline) Right: SPS User interface with top 3 results of the similarity search for this case.