RPS 2211 - Photon-counting CT in neuroradiology: from artifact reduction to high-resolution surgical planning

Author Block: L. A. Cappellini¹, G. Ressa¹, M. De robertis¹, G. Savini¹, R. Levi¹, C. Brembilla², M. Fornari², L. S. Politi¹; ¹Pieve Emanuele/IT, ²Rozzano/IT
Purpose: The aim of this study was to qualitatively and quantitatively determine whether Photon-Counting CT (PCCT) with Virtual Monoenergetic Imaging (VMI) improves postoperative assessment of spinal implants by reducing metallic artifacts, enhancing diagnostic confidence, and better evaluating postoperative complications, in comparison with conventional CT.
Methods or Background: A total of 16 subjects who underwent PCCT (NAEOTOM Alpha, Siemens) for postoperative evaluation following anterior, posterior, or combined spinal fixation were included. Scans were performed at 120 or 140 kVp, with and without IMAR.
Images were evaluated using VMI (60–190 keV). For each examination, 13 ROIs were placed to calculate Artifact Index (AIx), HU values, and Standard Deviation (SD). Subjective evaluation was performed by three radiologists using a Likert scale (1–5) at different energy levels. Where available PCCT was compared to conventional CT.
Results or Findings: PCCT proved effective in reducing soft tissue artifacts, with optimal performance between 110 and 140 keV. At energies above 150 keV, loss of contrast were observed impairing tissue differentiation.
Images at 0.2 mm with Br76 kernel showed higher SD and AIx but were subjectively superior for evaluating screws, bone contours, and metal-bone interfaces, compared to IMAR and VMI, due to overcorrection bands, shading, and native software limitations.
The subjectively selected optimal reconstruction did not always correspond to the image with the lowest AIx. Compared to available conventional CT scans, monoenergetic PCCT images achieved higher Likert scores and lower AIx values.
Conclusion: PCCT provides high subjective and objective image quality in patients with spinal metallic implants. For the evaluation of periprosthetic bone changes, ultra-high-resolution reconstructions with bone kernel images are preferred. Virtual Monoenergetic Imaging (VMI) improves soft tissue assessment, particularly in the setting of postoperative complications.
Limitations: Sample size, Comparison with conventional CT was limited
Funding for this study: ANTHEM Project
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was approved by our institutional ethics committee.

Author Block: A. Toth, M. T. Hagar, M. Silbergleit, J. Y. Cho, M. Vecsey-Nagy, A. Varga-Szemes, M. V. Spampinato; Charleston, SC/US
Purpose: Ultra-high-resolution (UHR) photon-counting detector (PCD)-CT angiography with sharp neurovascular kernels (Hv72) and high-level quantum iterative reconstruction (QIR-4) yields excellent intracranial image quality but often suffers from elevated image noise in the lower neck. This study evaluated the added value of a post-hoc convolutional neural network (CNN) denoising algorithm on quantitative image quality metrics in head and neck UHR PCD-CTA.
Methods or Background: In this retrospective study, head and neck UHR PCD-CTA datasets were reconstructed using a sharp neurvascular kernel (Hv72) and iterative reconstruction (QIR-4), with and without post-hoc CNN–denoising. Image noise, intraluminal attenuation, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR) were measured at the ’lower–neck’ (CCA and vertebral artery segment V1), ’mid–neck’ (carotid bifurcation and V2), ’intracranially’ (ICA and V4), and compared using paired t-test.
Results or Findings: 25 consecutive patients were enrolled. CNN-denoising reduced image noise at all vascular levels (p<0.001), with the largest absolute reduction in the ’lower-neck’ (–28.6 HU; 43.3%, p<0.001). Mean intraluminal attenuation was preserved across all vascular levels with only minimal differences (<3 HU), and signal variability (SD) remained unchanged. SNR and CNR improved significantly throughout the vasculature. In the ’lower-neck’, CNR increased by 79.4% (6.3 ± 5.8 vs 11.3 ± 9.2; p<0.001). In the ’mid-neck’, SNR and CNR increased from 35.8 ± 11.3 to 74.8 ± 72.5 (p<0.001). ’Intracranially’, SNR rose from 27.8 ± 6.4 to 51.4 ± 11.6 and CNR from 25.1 ± 6.3 to 46.5 ± 11.2 (p<0.001).
Conclusion: CNN–denoising selectively suppressed image noise without altering vessel attenuation, providing SNR and CNR improvements across all vascular levels. Improvements were most pronounced in the mid neck, while intracranial regions demonstrated stable and reproducible enhancement. While lower neck values improved, variability remained higher due to artifact-prone anatomy.
Limitations: No limitations.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was approved by the Institutional Review Board (Pro00142731).

Author Block: G. D'Anna, S. Barazzetta, M. Alì, A. Licata, M. Gerli, F. Darvizeh, L. Di Palma, I. Castiglioni, D. Fazzini; Milan/IT
Purpose: To evaluate the possibility to detect brachial plexus on Total Body PCCT. We evaluated post-contrast images looking for the divisions and relations with surrounding structures.
Methods or Background: We retrospectively analyzed the first 11 patients scanned at our Institution with photon-counting CT (Siemens Naetom Alpha), comparing similar patients evaluated with our other scanner (Siemens Magnetom Flash). We established as “well-visible” a case with good delineation of roots and trunks, “medium visible” a case with weak delineation of roots and trunks, “poor visible” a case with poor delineation of the structures.
Results or Findings: In the reprocessed images at 70 keV, based on the acquisition made with Photon-Counting technology scanner, out of a total of 11 patients the brachial plexus anatomy is clearly visible in 4 patients (36.36%). In 5 patients it is moderately visible (45.45%) while in 2 cases (18.18%) it is not possible to discern the anatomical structures reliably. In images reprocessed at 140 keV brachial plexus anatomy is clearly visible in 3 patients (27.27%). In 3 patients it is moderately visible (27.27%) while in 5 cases (45.45%) it is not possible to discern the anatomical structures reliably. Aat 70 keV (No. 4), only No. 3 is also clearly visible in the reworks at 140 keV. (75%). Cases with average brachial plexus anatomy visible at 70 keV are No. 5, while at 140 keV they turn out to be No. 3 (60%). All the cases evaluated on Somatom Flash showed poor delineation of brachial plexus.
Conclusion: PCCT may evaluate anatomical structures of brachial plexus, allowing more precise and effective diagnosis and patient management.
Limitations: Small sample size
Funding for this study: None
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: A. Toth, S. Al Kasab, K. Kicielinski, M. Silbergleit, J. Y. Cho, M. T. Hagar, M. Vecsey-Nagy, A. Varga-Szemes, M. V. Spampinato; Charleston, SC/US
Purpose: Three-dimensional (3D) volume rendering (VR) reconstructions offer comprehensive illustration of anatomy and pathology. We hypothesized that combining the high spatial resolution of ultra-high-resolution (UHR) photon-counting detector (PCD)-CTA with 3D VR post-processing could provide significant benefits for the neurosurgical evaluation and management of intracranial aneurysms.
Methods or Background: In this IRB-approved retrospective study, consecutive patients with diagnosed intracranial aneurysms were included. Image acquisition was performed using a clinical PCD-CT scanner in UHR mode. Two neurosurgeons independently evaluated both conventional 2D images and 3D VR reconstructions. A 13-item questionnaire was used to assess image quality, aneurysm characterization, and the clinical usefulness of the 3D VR datasets. Aneurysm neck and dome dimensions were measured on both 2D source images and 3D VR reconstructions to determine inter-reader agreement.
Results or Findings: 21 patients with 32 aneurysms were enrolled. Image quality and clinical usefulness of 3D VR images were rated as excellent, with percentage agreement ranging from 58% to 100%, primarily differing between ratings of 4 and 5. Inter-reader reliability for aneurysm size measurements on 3D VR reconstructions was good for dome size (ICC=0.74) and moderate for neck size (ICC=0.55). Bland–Altman analysis showed measurement differences were generally minor, with a few larger discrepancies for bigger aneurysms.
Conclusion: UHR CTA-based 3D VR reconstructions provided high subjective image quality, allowing for the precise depiction of aneurysm morphology and reproducible size measurements with good inter-reader agreement. UHR CTA-based 3D VR reconstructions may complement or potentially replace confirmatory invasive angiography in selected cases by providing reliable morphological information, streamlining preoperative workflows, and supporting surgical or endovascular planning.
Limitations: Digital subtraction angiography (DSA), MR angiography, or surgical records were not used as reference standard, precluding direct confirmation of aneurysm size and morphology.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was approved by the Institutional Review Board (Pro00142731).

Author Block: I. Conti, G. De Paula, D. Turilli, L. Piscopo, E. Solinas, L. M. Fattacciu, M. Pedde, M. Scaglione, S. A. Masala; Sassari/IT
Purpose: This study aims to develop autonomous Generative AI agents capable of integrating entropy maps derived from Photon-Counting Computed Tomography (PCCT), magnetic resonance imaging (MRI), and, when available, histopathological data, to enable personalized diagnostic profiling of brain tumors (specifically glioblastomas and high-grade gliomas) thereby facilitating advanced preoperative planning and intraoperative guidance.
Methods or Background: A prospective analysis was conducted on 25 patients affected by glioblastoma or various grade glioma who underwent PCCT brain imaging at our Radiology Department. A dedicated angiographic protocol was applied to assess perilesional vascularization, in addition to generating iodine concentration maps and entropy maps for lesion characterization and perfusion analysis. Imaging datasets underwent radiomic feature extraction and were co-registered with multiparametric MRI, including MR spectroscopy and, when available, with histopathological results.
Results or Findings: This study demonstrated that the use of PCCT provides enhanced spatial resolution and allows evaluation of post-treatment tumor activity using entropy and iodine maps, while achieving significant reduction in both radiation dose and intravenous contrast medium volume.
integrating traditional AI with advanced radiological imaging significantly enhances the quality of care for patients with glioblastoma and high-grade gliomas. The use of multimodal imaging and AI-driven interpretation enables advanced diagnosis and personalized treatment planning, resulting in increased survival rates, fewer re-interventions, and overall cost-effectiveness.
Conclusion: Since clinicians often struggle to integrate complex, heterogeneous, and large-scale data into a comprehensive view of oncologic patient management, the assistance of an autonomous Generative AI agents play a pivotal role to fully leverage all available data.
Limitations: No limitations were identified.
Funding for this study: No funding was received for this study
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information: The study is educational and is under ethical committee evaluation