METHODS: A 3D-printed cardiac insert and Catphan 500 phantoms were scanned using CCTA protocols at 120 and 100 kVp tube voltages. All CT acquisitions were reconstructed using filtered back projection (FBP) and Adaptive Statistical Iterative Reconstruction (ASIR) algorithm at 40% and 60% strengths. Image quality characteristics such as image noise, signal-noise ratio (SNR), contrast-noise ratio (CNR), high spatial resolution, and low contrast resolution were analyzed.
RESULTS: There was no significant difference (P > 0.05) between 120 and 100 kVp measures for image noise for FBP vs ASIR 60% (16.6 ± 3.8 vs 16.7 ± 4.8), SNR of ASIR 40% vs ASIR 60% (27.3 ± 5.4 vs 26.4 ± 4.8), and CNR of FBP vs ASIR 40% (31.3 ± 3.9 vs 30.1 ± 4.3), respectively. Based on the Modulation Transfer Function (MTF) analysis, there was a minimal change of image quality for each tube voltage but increases when higher strengths of ASIR were used. The best measure of low contrast detectability was observed at ASIR 60% at 120 kVp.
CONCLUSIONS: Changing the IR strength has yielded different image quality noise characteristics. In this study, the use of 100 kVp and ASIR 60% yielded comparable image quality noise characteristics to the standard CCTA protocols using 120 kVp of ASIR 40%. A combination of 3D-printed and Catphan® 500 phantoms could be used to perform CT dose optimization protocols.
METHODS: Twenty volumes of interests consisting of six anterior and fourteen posterior edentulous regions were obtained from human mandibular cadavers. A CBCT system with a resolution of 80 µm (3D Accuitomo 170, J. Morita, Kyoto, Japan) and a µCT system with a resolution of 35 µm (SkyScan 1173, Kontich, Belgium) were used to scan the mandibles. Three structural parameters namely, trabecular number (Tb.N), trabecular thickness (Tb.Th), and trabecular separation (Tb.Sp) were analysed using CTAn software (v 1.11, SkyScan, Kontich, Belgium). For each system, the measurements obtained from anterior and posterior regions were tested using independent sample t-test. Subsequently, all measurements between systems were tested using paired t-test.
RESULTS: In CBCT, all parameters of the anterior and posterior mandible showed no significant differences (p > 0.05). However, µCT showed a significant different of Tb.Th (p = 0.023) between anterior and posterior region. Regardless of regions, the measurements obtained using both imaging systems were significantly different (p ≤ 0.021) for Tb.Th and Tb.N.
CONCLUSIONS: The current study demonstrated that only the variation of Tb.Th between anterior and posterior edentulous region of mandible can be detected using µCT. In addition, CBCT is less feasible than µCT in assessing trabecular bone microstructures at both regions.
METHODS: In this paper, we analyze four wide-spread deep learning models designed for the segmentation of three retinal fluids outputting dense predictions in the RETOUCH challenge data. We aim to demonstrate how a patch-based approach could push the performance for each method. Besides, we also evaluate the methods using the OPTIMA challenge dataset for generalizing network performance. The analysis is driven into two sections: the comparison between the four approaches and the significance of patching the images.
RESULTS: The performance of networks trained on the RETOUCH dataset is higher than human performance. The analysis further generalized the performance of the best network obtained by fine-tuning it and achieved a mean Dice similarity coefficient (DSC) of 0.85. Out of the three types of fluids, intraretinal fluid (IRF) is more recognized, and the highest DSC value of 0.922 is achieved using Spectralis dataset. Additionally, the highest average DSC score is 0.84, which is achieved by PaDeeplabv3+ model using Cirrus dataset.
CONCLUSIONS: The proposed method segments the three fluids in the retina with high DSC value. Fine-tuning the networks trained on the RETOUCH dataset makes the network perform better and faster than training from scratch. Enriching the networks with inputting a variety of shapes by extracting patches helped to segment the fluids better than using a full image.
METHODS: Sixteen computed tomography scan of SC patients (8 months-6 years old) were imported to Materialise Interactive Medical Image Control System (MIMICS) and Materialise 3-matics software. Three-dimensional (3D) OC models were fabricated, and linear measurements were obtained. Mathematical formulas were used for calculation of OC volume and surface area from the 3D model. The same measurements were obtained from the software and used as ground truth. Data normality was investigated before statistical analyses were performed. Wilcoxon test was used to validate differences of OC volume and surface area between 3D model and software.
RESULTS: The mean values for OC surface area for 3D model and MIMICS software were 103.19 mm2 and 31.27 mm2, respectively, whereas the mean for OC volume for 3D model and MIMICS software were 184.37 mm2 and 147.07 mm2, respectively. Significant difference was found between OC volume (P = 0.0681) and surface area (P = 0.0002) between 3D model and software.
CONCLUSION: Optic canal in SC is not a perfect conical frustum thus making 3D model measurement and mathematical formula for surface area and volume estimation not ideal. Computer software remains the best modality to gauge dimensional parameter and is useful to elucidates the relationship of OC and eye function as well as aiding intervention in SC patients.
METHODS: This was a prospective observational study performed at the Glaucoma Research Centre, Montchoisi Clinic, Lausanne. In total 40 eyes with open-angle glaucoma were included. OCT-A scans were performed before glaucoma surgery, and at 1-month, 3-month, 6-month, and 12-month post-operatively. AngioVue AngioAnalytic (Optovue Inc, Fremont, CA) software was used to analyse the RNFL, FAZ, peripapillary and macular VD. Changes were analysed using analysis of variance (ANOVA) models.
RESULTS: Mean IOP dropped from 19.4 (±7.0) mmHg pre-surgery and stabilized at 13.0 (±3.1) mmHg at 12 months (p