RPS 1611 - Neuroimaging: innovation in techniques and contrast agents

Author Block: B. D. Wichtmann¹, K. Teichmann¹, N. Lehnen¹, T. Tonguc¹, F. Padormo², A. Rastogi¹, Z. Bendella¹, A. Radbruch¹, J. Faber¹; ¹Bonn/DE, ²Guilford, CT/US
Purpose: Portable ultra-low-field-(pULF)-MRI enables bedside neuroimaging in resource limited settings. However, lower signal and relaxivity differences at 64mT limit contrast enhancement (CE). This prospective study provides first evaluation of CE-pULF-MRI in a heterogeneous cohort compared to high-field-(HF)-MRI for lesion detection and diagnostic confidence.
Methods or Background: This IRB-approved prospective study included 39 patients (58.1±15.7years; 20♀) undergoing clinically indicated CE-HF-MRI at 1.5/3T for oncologic, demyelinating, inflammatory, vascular, postoperative, or degenerative conditions between 04-05/2025. Patients received native pULF-MRI (SWOOP,Hyperfine,Guilford,CT) with axial T1w standard (std) and synthetic echo train length (sETL) sequences, FLAIR, T2w, and DWI. HF-MRI followed clinical protocols with pre-/post-contrast T1w-imaging after IV-administration of 1.0mmol/mL gadobutrol (Gadovist). CE-pULF-MRI was then repeated using the same axial T1w std and sETL sequences.
4 blinded (neuro-)radiologists independently evaluated all T1w scans from both field strengths for post-contrast hyperintense lesions, rating diagnostic confidence per lesion on a 5-point Likert scale (1=very uncertain; 5=very certain).
Results or Findings: 29/39 patients had ≥ 1 lesion on CE-HF-MRI. Up to 100 lesions per rater were marked on HF-MRI.
On pULF-MRI 23% of lesions were missed, thereof 59% <6 mm, 89% <8 mm. The remaining 11% were located infratentorial, near the venous sinus, subcallosal at the frontal sinus, or outside the pULF-MRI scan volume. One lesion missed on pULF-MRI showed minimal CE, detected on HF-MRI by only two readers with low confidence (2).
Diagnostic confidence was slightly higher for pULF T1w std than sETL. In pULF T1w std, ≥2 readers marked 3 false positives; in sETL, 4 false positives, all with low confidence.
Conclusion: CE pULF MRI enables detection of enhancing brain lesions ≥6–8 mm and is feasible for bedside imaging. T1w std outperforms sETL.
Limitations: Limitations exist for smaller lesions and those in anatomically challenging locations.
Funding for this study: None.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was conducted under the ethical approval “Gehirnentwicklung von Früh- und Neugeborenen während des Aufenthaltes auf der neonatologischen Intensivstation” granted by the local institutional review board (Ethics Committee of the University Hospital Bonn). All participants provided written informed consent.

Author Block: S. Pasumarthi Venkata, T. Campbell Arnold, A. Shankaranarayanan; Menlo Park, CA/US
Purpose: Multi-contrast brain MRI images offer complimentary information that aids in differential diagnosis and treatment. The acquisition of high-quality multi-contrast images is time consuming and is often susceptible to artifacts. In this work, we propose the synthesis of T1w images from T2 and FLAIR images using a multi-contrast multi-scale vision transformer (MMT). T1w images are usually acquired at the start of any brain protocol and provide a structural baseline. Other contrasts like T2, FLAIR and T1C provide additional pathological information. It was found that the basic structure in T1 can be faithfully reconstructed from T2 and FLAIR images thus cutting down on the acquisition time (~4 mins for 3D; ~1.5 mins for 2D). We quantitatively evaluate the synthesized and acquired T1w images using PSNR, SSIM and RMSE.
Methods or Background: MMT was pre-trained on all subsets of T1,T1C,T2,FLAIR as input and the missing sequence(s) as the output. This model was further fine-tuned by fixing T2&FLAIR as inputs, and T1 as output. Reconstruction L1-loss was weighted on the lesion maps generated from T1C images using a pre-trained lesion segmentation algorithm. Multi-contrast (T1w,T2&FLAIR) brain MRI images from 45 patients were quantitatively evaluated using PSNR, SSIM and RMSE computed between real and synthesized T1.
Results or Findings: The average PSNR, SSIM and RMSE over the 45 patient studies are 33.53±1.75, 0.91±0.13 and 0.18±0.02 respectively. The synthesized T1w images had better overall image quality with less artifacts.
Conclusion: In this study, we have shown the quantitative performance of synthesizing T1w images from T2 and FLAIR to reduce acquisition time. The synthesized T1w images provide similar structural information as that of the acquired T1w images.
Limitations: Detailed qualitative assessment is missing. Future studies will evaluate different downstream tasks using the synthesized T1 image.
Funding for this study: n/a
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Internal IRB

Author Block: B. P. Liu¹, C. Deuschl², K. Kudo³, A. Liu⁴, G. Sutter⁵, B. M. Hofmann⁵, P. Palkowitsch⁵; ¹Chicago, IL/US, ²Essen/DE, ³Sapporo/JP, ⁴Whippany, NJ/US, ⁵Berlin/DE
Purpose: Compared to macrocyclic gadolinium (Gd)-based contrast agents (GBCA) used at 0.1 mmol Gd/kg body weight (bw), gadoquatrane’s tetrameric structure enables high relaxivity, allowing for a reduction of 90% molecule and 60% Gd dose. This analysis of Phase (ph) 2 and 3 data explores the potential effect of the gadoquatrane structure and larger molecular size on imaging and pharmacokinetics (PK) vs. gadoterate meglumine, gadoteridol and gadobutrol.
Methods or Background: Two controlled clinical trial studies were conducted in patients with known or suspected CNS pathologies. Patients underwent two CE-MRIs, one with gadoquatrane (0.01 mmol/kg body weight (bw), equal to 0.04 mmol Gd/kg bw) and one with comparator GBCA (0.1 mmol Gd/kg bw).
In ph3, MRIs followed clinical practice; in ph2 scans were taken at 5-, 10- and 15-min post injection. Images were evaluated in a blinded central read using 3 visualization parameters: contrast enhancement, delineation, morphology. Signal intensity was also assessed quantitatively in ph2. PK were evaluated via plasma samples taken at predefined time windows.
Results or Findings: The integrated analysis included 362 adult CNS patients, with 326 evaluable for efficacy, and 338 for pharmacokinetics.
Gadoquatrane MRI was non-inferior to the comparator MRI for all visualization parameters. Quantitative enhancement was similar for gadoquatrane and gadobutrol. Dose normalized plasma concentration-time curves showed overlapping concentrations up to 24 hours post dose, indicating dose-proportionality and essentially the same PK behavior over time for all GBCAs.
Conclusion: Gadoquatrane at a 60% reduced Gd dose is non-inferior to comparator GBCAs for CNS CE-MRI. For later imaging time points, qualitative and quantitative assessments were similar to comparator GBCAs. There is no relevant impact of the larger molecular size on imaging properties or PK behavior.
Limitations: The integrated analysis includes patients with CNS pathologies only.
Funding for this study: Both studies referenced were sponsored by Bayer AG.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The studies were approved by Institutional Review Boards at all participating investigation sites of these multicenter trials.

Author Block: G. Sutter¹, E. Vendel², P. Vis², J. Zisowsky¹, M. Feldmueller¹, B. M. Hofmann¹; ¹Berlin/DE, ²Leiden/NL
Purpose: An integrated population-pharmacokinetic (popPK) modeling approach was chosen to thoroughly evaluate the pharmacokinetics (PK) of the extracellular macrocyclic gadolinium-based contrast agent (mGBCA) gadoquatrane, in adults and children, utilizing data from all Phase 1-3 clinical trials, including a pediatric study, to support regulatory submission for the use in patients of all ages with CNS or body pathologies.
Methods or Background: Data from eight clinical Phase 1-3 studies, covering doses of 0.0025 – 0.05 mmol/kg body weight (bw) gadoquatrane (corresponding to 0.01 – 0.2 mmol Gd/kg bw), and a Phase 1/3 study in pediatric patients aged 0-<18 years (0.04 mmol Gd/kg bw) were analyzed. Non-linear mixed-effects methods were applied for popPK model development including covariate assessment. Analysis started with data from 871 adults (4054 Gd plasma concentrations), forming the basis for the pediatric analysis by applying and adjusting the adult model to pediatric data (92 children with 276 Gd plasma concentrations). Exposure parameters were compared across populations (age, gender, race/ethnicity, renal function, health status) and with other mGBCAs.
Results or Findings: The adult model was a linear three-compartment model with allometric scaling based on lean body mass (LBM) and estimated glomerular filtration rate as covariate on clearance (CL). The pediatric data allowed only for identification of two compartments, the resulting model included LBM-based allometric scaling and a maturation factor on CL based on postmenstrual age.
Gadoquatrane demonstrated consistent PK behavior across populations, closely matching the patterns observed with other mGBCAs.
Conclusion: This integrated popPK modeling approach enabled a thorough and efficient characterization of gadoquatrane’s PK in both adults and children, supporting regulatory submission across all indications and ages.
Limitations: Limitations included limited representation of some groups (Patients aged <6 months, Black participants, individuals with severe renal impairment).
Funding for this study: All studies referenced were sponsored by Bayer AG.
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: A. Rulseh, Z. Ryznarová, R. Liščák, J. Vymazal; Prague/CZ
Purpose: To demonstrate the role of gadopiclenol in MR imaging of brain metastases at standard (0.05 mmol/kg) and double doses (0.1 mmol/kg), including delayed examination (~15 minutes post-contrast).
Methods or Background: One hundred sixty-two subjects with known brain metastases underwent MRI prior to radiosurgery (75 female, mean age 63.67 years ± 12.7 SD). A standard T1W 3D SPACE sequence was acquired post-contrast in all subjects (5 subjects at 1.5T, 157 at 3T), followed by a delayed acquisition. In roughly half of subjects (n=83), the delayed acquisition was preceded by additional contrast administration (cumulative double dose). After randomization, 3 blinded readers evaluated each subject in 2 sessions, minimally 4 weeks apart. A one-sided paired t-test was used to compare the number of lesions detected in the 2 groups: single dose (early versus delayed), and double dose (early versus delayed [double]).
Results or Findings: Agreement between all blinded readers was used for evaluation. The delayed single dose exam revealed 288 metastases compared to 279 metastases on the early exam. Delayed double-dose exam revealed 258 metastases compared to 221 metastases on the early single dose exam. In subjects that received a double dose, the number of lesions detected on the early versus delayed (double) dose exams differed significantly (p=0.002), while in subjects that received only a single dose no significant difference in the number of lesions detected on early versus delayed exams was detected (p=0.24).
Conclusion: Significantly more metastases were detected on delayed double dose imaging, while no difference was detected on early versus delayed single dose acquisitions. Our results suggest that a delayed, double dose gadopiclenol acquisition may improve the detection of metastases.
Limitations: The study was performed at a single institution. Only the SPACE sequence was used for blinded reading.
Funding for this study: Ministry of Health CZ, IG244302.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Ethics committee approval, na homolce hospital ethics committee

Author Block: S. Gerevini¹, G. Hutóczki², A. G. Bagó³, A. Pichiecchio⁴, S. Gaudino⁵; ¹Cremona/IT, ²Debrecen/HU, ³Gdansk/PL, ⁴Pavia/IT, ⁵Rome/IT
Purpose: To demonstrate the non-inferiority of DSC-MRI perfusion using gadopiclenol at 0.05 mmol/kg compared to gadoterate meglumine at 0.1 mmol/kg in terms of diagnostic quality of Cerebral Blood Volume (CBV) perfusion map.
Methods or Background: In this multicentre randomized controlled trial, 138 adult patients with primary glial tumour scheduled for a follow-up DSC-MRI were randomized to receive either gadopiclenol or gadoterate meglumine. The primary outcome was the CBV map diagnostic quality assessed by two independent blinded readers, using a 4-point scale (poor, fair, good, excellent). Secondary outcomes included glioma grade differentiation by relative CBV (rCBV) quantification, and safety. Non-inferiority was defined as proportion of patients with excellent or good diagnostic quality images for gadopiclenol of no more than 12% below that for gadoterate meglumine.
Results or Findings: Per-protocol analysis included 60 patients with gadopiclenol and 64 with gadoterate meglumine. The proportion of patients with CBV map images assessed as “excellent or good” was 96.7% with gadopiclenol and 98.4% with gadoterate meglumine, respectively, with a difference of -1.8% [two-sided 95% CI: - 8.9% to 5.3%], demonstrating the non-inferiority. Overall, rCBV values were significantly different between high-grade and low-grade gliomas with both contrast agents, with a median of 3.11 vs 1.51 (p =0.003) for gadopiclenol and 2.47 vs 1.41 (p <0.001) for gadoterate meglumine.
Three patients receiving gadopiclenol and two receiving gadoterate meglumine experienced non-serious adverse events of mild intensity, none related to the contrast agent.
Conclusion: DSC-MRI perfusion using gadopiclenol at 0.05 mmol/kg was non-inferior to gadoterate meglumine at 0.1 mmol/kg on diagnostic quality of CBV perfusion map and both enabled brain glioma grade differentiation. This study additionally confirmed the safety profile of gadopiclenol.
Limitations: No limitations were identified
Funding for this study: This study was funded by Guerbet.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Approval number for the centers in Hungary : BM/3737-0/2023-EKL
Approval number for the centers in Italy : 2023-3.11/17
Approval number for the centers in Poland: KE-0254/60/03/2023

Author Block: J. S. Endrikat¹, I. Siddiqui², H. Khater¹, M. Blankenburg¹; ¹Berlin/DE, ²Reading/UK
Purpose: To systematically explore rationales for non-standard dosing of GBCAs and discuss the potential future impact of high-relaxivity contrast agents.
Methods or Background: A systematic literature review was conducted using Embase, covering 1991 to 2024. Publications were categorized by indication, GBCA dose, study design, and rationale for non-standard dosing. The dose of 0.1 mmol Gd/kg body weight was defined as the 'standard' reference for comparison.
Results or Findings: Eighty-four publications comparing non-standard dosing regimen with the standard dose were included: 43 high-dose and 55 low-dose studies. The rationales for high-dose administration were to achieve better contrast (25/43; 58%) and improve lesion detection (15/43; 35%). High-dose studies were primarily performed in CNS until 2006.

Rationales for using low-dose administration were related to 1) NSF (28/55; 51%); 2) Gd exposure (22/55; 40%); 3) cost (22/55; 40%); 4) unspecified safety (18/55; 33%); 5) Gd retention/presence (16/55; 29%); and 6) environment (6/55; 11%). From 1991 to 2006, cost was the sole rationale for lower dose administration. From 2008, NSF was noted; from 2017 onwards, Gd retention/presence, and most recently, environmental impact emerged as important rationales. Forty-six of 55 studies (84%) investigating low-dose regimen reported comparable outcomes, 7 studies (13%) inferior outcomes. However, 35 of 46 low-dose studies reporting comparable outcomes modified also other parameters or used a study design potentially impacting study strength.
Conclusion: For over 34 years, there has been a consistent demand to lower GBCA doses, with an increasing number of rationales over time. The next generation of high-relaxivity, low-dose mGBCAs shows promise for reducing Gd dose while maintaining high image quality, potentially defining a new standard dose.
Limitations: The impact of advances in scanner technology and software was beyond the scope of this review.
Funding for this study: Funding was provided by Bayer AG.
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: Z. Che; GuiYang/CN
Purpose: To evaluate the feasibility of an ultra-fast MRI protocol (scan time: 3 min 10 s) with Deep Learning Reconstruction (DeepRecon) combined with AI Compressed Sensing (ACS) in diagnosing acute stroke, and compare it with the conventional MRI protocol (scan time: 6 min 36 s).
Methods or Background: Fifty-six prospectively enrolled suspected acute stroke patients (onset < 6 h) underwent 3.0T MRI with both protocols. The conventional protocol included T2WI/T2-FLAIR/DWI/TOF-MRA (6 min 36 s); the ultra-fast one used DeepRecon+ACS for the same sequences (3 min 10 s). Three neuroradiologists performed double-blind evaluations. ICC, McNemar test, Bland-Altman analysis and Wilcoxon test were used for statistics (SPSS 26.0, α=0.05).
Results or Findings: The two protocols had 100% consistency in acute infarct detection (κ=1.0), 94.4% agreement in large vessel occlusion (LVO) detection (κ=0.89), and 97.2% in minor findings (e.g., microbleeds). The ultra-fast protocol had slightly lower but diagnostic image quality (4.1±0.4 vs 4.8±0.3, P=0.002), 63% less motion artifact (5.6% vs 27.8%, P=0.003), and 58.1% shorter scan time.
Conclusion: The DeepRecon+ACS ultra-fast MRI (3 min 10 s) maintains high diagnostic efficacy (consistency >94%) for acute infarcts/LVO, shortens scan time, reduces motion artifact, and meets the needs of stroke green channels.
Limitations: 1. Small sample size (56 patients) with no subgroup analysis (e.g., age, comorbidities), limiting external applicability.
2. Single scanning device (only 3.0T MRI); efficacy on low-field MRI (e.g., 1.5T) unproven, poor device compatibility.
3. No "gold standard" imaging (e.g., CTA, DSA) as reference; no subgroup analysis by onset time, incomplete study dimensions.
Funding for this study: Not applicable
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study has been approved by the Ethics Committee of Guiqian International Hospital.

Author Block: Y. Xiang; Guiyang/CN
Purpose: To investigate the 2D MultiBand EPI ASL imaging technique for obtaining quantitative results of Cerebral Blood Flow (CBF) and Arterial Transit Time (ATT). Subsequently, open-source post-processing software BASIL was used to analyze the data, and the results were compared with those from the conventional 3D ASL sequence to verify the quantitative accuracy of this technique.
Methods or Background: The study was conducted using a United Imaging uM880 3.0T magnetic resonance scanner equipped with a 48-channel head and neck coil. The scanning parameters were as follows: Field of View (FOV) = 220 × 220 mm², voxel size = 2.5 × 2.5 × 2.3 mm, interslice gap = 10%, number of slices = 48, scanned in a bottom-to-top sequential order. The MultiBand factor was set to 6, bandwidth = 2600 Hz/pixel, Echo Time (TE) = 16.4 ms. The labeling time was 1500 ms, and a total of 5 Post-Labeling Delays (PLDs) were used: 0.2 / 0.7 / 1.2 / 1.7 / 2.2 sOne M0 image was acquired for calibration. The total scanning time was 5 minutes and 36 seconds.After acquisition, two independent radiologists evaluated the overall image quality of both the mPLD ASL and the conventional 3D ASL sequences using a five-point Likert scale in terms of signal-to-noise ratio (SNR), clarity, and accuracy.
Results or Findings: Compared with the conventional 3D ASL sequence, the mPLD ASL demonstrated a significant improvement in overall image quality, particularly in SNR. Additionally, it reduced motion sensitivity and image blurring. Most importantly, the mPLD ASL technique enabled the acquisition of two sets of quantitative images for CBF and ATT, showing enhanced quantitative accuracy.
Conclusion: The mPLD ASL sequence not only improves SNR and image clarity but also provides superior quantitative accuracy.
Limitations: No
Funding for this study: No
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: M. R. López De La Torre Carretero, D. A. Zambrano, C. Mbongo, A. M. Delgado Brito, C. D. Solano, M. A. Fernández Seara, M. Calvo Imirizaldu; Pamplona/ES
Purpose: The main aim of this study was to assess the diagnostic value of perfusion imaging during intraoperative MRI (iMRI) for detecting residual lesions. We evaluated the image quality of intraoperative dynamic susceptibility contrast (DSC) and arterial spin labeling (ASL) sequences in a cohort of patients undergoing brain surgery, and then compared their ability to identify residual lesions with that of conventional techniques.
Methods or Background: Twenty-seven patients (13 men; mean age 60±14) undergoing intraoperative MRI during brain tumor resection were prospectively imaged with structural MRI, DSC, and ASL. Two independent readers rated image quality (1–4; 1–2 non-diagnostic, 3–4 diagnostic) and scored tumour remnants on structural MRI, DSC-CBV, and ASL-CBF as present, absent, or non-evaluable (artifacts); only perfusion studies rated diagnostic were included for the remnant analysis, and interobserver and intermodality agreement were assessed with Cohen’s kappa.
Results or Findings: Consensus diagnostic quality was slightly higher for DSC (25/27 cases) compared to ASL (23/26). Interobserver agreement was excellent for DSC (κ=1.00) and substantial for ASL (κ=0.75). Concordance between DSC and ASL in evaluable cases was moderate (κ=0.26). When compared with structural MRI, perfusion-based assessment of residual tumor demonstrated substantial agreement for both DSC (κ=0.89) and ASL (κ=0.69).
The performance of the two techniques was not interchangeable: DSC was generally more resilient to image degradation, whereas ASL enabled interpretation in one case where DSC failed.
Conclusion: Intraoperative perfusion MRI is feasible. DSC yields slightly better image quality and structural agreement, but ASL proves to be useful when contrast is contraindicated, not needed (i.e. non-enhancing tumors) or DSC is compromised, providing complementary value for the detection of tumor remnants during brain surgery. Their complementary roles support combined use to enhance intraoperative confidence.
Limitations: Tumour heterogeneity
Qualitative analysis
No histopathological gold-standard
Funding for this study: No funding was provided for this study
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: S. S. Croosu, J. Røikjer, S. M. Tahami, T. M. Hansen, J. Brøndum Frøkjær, N. Ejskjaer, C. D. Mørch; Aalborg/DK
Purpose: Diabetic peripheral neuropathy (DPN) is a frequent complication of type 1 diabetes mellitus (T1DM), causing sensory loss and pain, but reliable imaging biomarkers of nerve damage are lacking. This study investigated whether tibial and sciatic nerve integrity is affected by T1DM with painful or painless DPN using diffusion tensor imaging magnetic resonance neurography (DTI-MRN), and whether these measurements associate with clinical outcomes.
Methods or Background: Fifty-eight participants with T1DM were divided into painful DPN(n=20), painless DPN(n=19), and no DPN(n=19), along with 20 healthy controls (HC). DTI-MRN was performed at 10% and 40% above the patella to evaluate the tibial and sciatic nerves, respectively. Fractional anisotropy (FA) and apparent diffusion coefficient (ADC) were measured. Correlations with age, diabetes duration, HbA1c, nerve conduction, and pain scores were assessed.
Results or Findings: Higher FA indicates better fibre integrity and coherence, while lower ADC reflects more restricted diffusion, consistent with intact nerve fibres and myelin. Mean tibial nerve FA was lower in painful DPN(0.25±0.08) and painless DPN(0.24±0.05) versus T1DM without DPN(0.33±0.06) and HC(0.32±0.07) (all p<0.002). Tibial ADC was higher in painful DPN(912.2±338.2 ×10⁻⁶ mm²/s) compared with T1DM without DPN(622.0±152.6 ×10⁻⁶ mm²/s) and HC(673.6±185.4 ×10⁻⁶ mm²/s) (all p<0.007). No differences in the sciatic nerve were detected (all p>0.05). Lower FA and higher ADC of the tibial nerve were associated with poorer sural amplitude and velocity, impaired warm/cold detection, and greater pain intensity (all p<0.03).
Conclusion: DTI-MRN detects tibial nerve alterations in T1DM with DPN. FA reduction occurred in both painful and painless DPN, while ADC increase was specific to painful DPN. Tibial FA and ADC correlated with nerve dysfunction and pain, suggesting potential as biomarkers of neuropathy severity.
Limitations: Sample size and cross-sectional design limit causal interpretation. Longitudinal studies are needed.
Funding for this study: This work was partly supported by Augustinus Fonden, København, Denmark (grant 19-1302). The funding source did not influence the
study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The North Denmark Region Committee on Health Research Ethics granted the ethics approval (N-20190003)

Author Block: E. Armando¹, M. Porzio², G. Salmè³, E. Rachetta¹, F. Groppo Marchisio¹; ¹Caraglio/IT, ²Borgo San Dalmazzo/IT, ³Collegno/IT
Purpose: To evaluate brain CT scan doses in a radiology department with different CT scanners, with different algorythms, from filtered back projection to deep learning reconstructions and with or without camera based patient postioning.
Methods or Background: Dose data of all the brain CT scan examinations obtained by seven different scanners between January and September 2025 were included in the analysis, using the dose tracking software (Gray detector, EL.Co, Cairo Montenotte, Italy). DLP and CTDI data were analysed using SAS Studio on demand for academics using the Kruskall Wallis test. Children examinations were excluded
Results or Findings: 5064 head CT scans were included in the analysis. There were no statistical differences in CTDI and DLP among the scanners despite differences in technology and protocols. The most recent CT scanner with deep learning reconstruction and a camera for automated positioning of the patient had a 15% lower median DLP than other scanners. This improvement was not immediate but it was obtained after protocol revision, reducing the minimum mA in the protocol setting oif the current modulation.
Conclusion: Brain CT scan is the most frequent examination in most radiology departments. However it is the examination with the lowest reduction of CTDI and DLP despite the introduction of iterative and deep learning reconstruction algorythms, as low noise images are needed.
New scanners with deep learning based algorythms do not necessarily lead to dose reduction unless protocols are carefully revised to exploit these improvements
Limitations: This is a retrospective study and is the basis for further research
Funding for this study: None
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information: