Feasibility of Dedicated Breast Positron Emission Tomography Image Denoising Using a Residual Neural Network.

Objectives: This study aimed to create a deep learning (DL)-based denoising model using a residual neural network (Res-Net) trained to reduce noise in ring-type dedicated breast positron emission tomography (dbPET) images acquired in about half the emission time, and to evaluate the feasibility and the effectiveness of the model in terms of its noise reduction performance and preservation of quantitative values compared to conventional post-image filtering techniques. Methods: Low-count (LC) and full-count (FC) PET images with acquisition durations of 3 and 7 minutes, respectively, were reconstructed. A Res-Net was trained to create a noise reduction model using fifteen patients' data. The inputs to the network were LC images and its outputs were denoised PET (LC + DL) images, which should resemble FC images. To evaluate the LC + DL images, Gaussian and non-local mean (NLM) filters were applied to the LC images (LC + Gaussian and LC + NLM, respectively). To create reference images, a Gaussian filter was applied to the FC images (FC + Gaussian). The usefulness of our denoising model was objectively and visually evaluated using test data set of thirteen patients. The coefficient of variation (CV) of background fibroglandular tissue or fat tissue were measured to evaluate the performance of the noise reduction. The SUVmax and SUVpeak of lesions were also measured. The agreement of the SUV measurements was evaluated by Bland-Altman plots. Results: The CV of background fibroglandular tissue in the LC + DL images was significantly lower (9.10±2.76) than the CVs in the LC (13.60± 3.66) and LC + Gaussian images (11.51± 3.56). No significant difference was observed in both SUVmax and SUVpeak of lesions between LC + DL and reference images. For the visual assessment, the smoothness rating for the LC + DL images was significantly better than that for the other images except for the reference images. Conclusion: Our model reduced the noise in dbPET images acquired in about half the emission time while preserving quantitative values of lesions. This study demonstrates that machine learning is feasible and potentially performs better than conventional post-image filtering in dbPET denoising.

Effect of Tumor-Pixel Positioning on the Variability of SUV Measurements in PET Images.

Objectives: The aim of this study was to investigate the effect on standardized uptake value (SUV) measurement variability of the positional relationship between objects of different sizes and the pixel of a positron emission tomography (PET) image. Methods: We used a NEMA IEC body phantom comprising six spheres with diameters of 10, 13, 17, 22, 28, and 37 mm. The phantom was filled with 18F solution and contained target-to-background ratios (TBRs) of 2, 4, and 8. The PET data were acquired for 30 min using a SIGNA PET/MR scanner. The PET images were reconstructed with the ordered subsets expectation maximization (OSEM) algorithm with and without point-spread function (PSF) correction (OSEM + PSF + Filter and OSEM + Filter, respectively). A Gaussian filter of 4 mm full width at half maximum was applied in all reconstructions, except for one model (OSEM + PSF + no Filter). The matrix sizes were 128×128, 192×192, 256×256 and 384×384. Reconstruction was performed by shifting the reconstruction center position by 1 mm in the range 0 to 3 mm in the upward or rightward direction for each parameter. For all reconstructed images, the SUVmax of each hot sphere was measured. To investigate the resulting variation in the SUVmax, the coefficient of variation (CV) of each SUVmax was calculated. Results: The CV of the SUVmax increased as the matrix size and the diameter of the hot sphere decreased in all reconstruction settings. With PSF correction, the CV of SUVmax increased as the TBR increased except when the TBR was 2. The CV of the SUVmax measured in the OSEM + PSF + no Filter images were larger than those measured in the OSEM + PSF + Filter images. The amount of this increase was higher for smaller spheres and larger matrix sizes and was independent of TBR. Conclusions: Shifting the reconstruction center position of the PET image causes variability in SUVmax measurements. To reduce the variability of SUV measurements, it is necessary to use sufficient matrix sizes to satisfy sampling criterion and appropriate filters.

Evaluation of motion artifacts in brain magnetic resonance images using convolutional neural network-based prediction of full-reference image quality assessment metrics.

Purpose: Motion artifacts in magnetic resonance (MR) images mostly undergo subjective evaluation, which is poorly reproducible, time consuming, and costly. Recently, full-reference image quality assessment (FR-IQA) metrics, such as structural similarity (SSIM), have been used, but they require a reference image and hence cannot be used to evaluate clinical images. We developed a convolutional neural network (CNN) model to quantify motion artifacts without using reference images. Approach: The brain MR images were obtained from an open dataset. The motion-corrupted images were generated retrospectively, and the peak signal-to-noise ratio, cross-correlation coefficient, and SSIM were calculated. The CNN was trained using these images and their FR-IQA metrics to predict the FR-IQA metrics without reference images. Receiver operating characteristic (ROC) curves were created for binary classification, with artifact scores < 4 indicating the need for rescanning. ROC curve analysis was performed on the binary classification of the real motion images. Results: The predicted FR-IQA metric having the highest correlation with the subjective evaluation was SSIM, which was able to classify images requiring rescanning with a sensitivity of 89.5%, specificity of 78.2%, and area under the ROC curve (AUC) of 0.930. The real motion artifacts were classified with the AUC of 0.928. Conclusions: Our CNN model predicts FR-IQA metrics with high accuracy, which enables quantitative assessment of motion artifacts in MR images without reference images. It enables classification of images requiring rescanning with a high AUC, which can improve the workflow of MR imaging examinations.

Synthesis and Structural Analysis of Palladium(II) Complexes Containing Neutral or Anionic C 2-Symmetric Bis(oxazoline) Ligands: Effects of Substituents in the 5-Position.

A series of neutral and cationic palladium(II) complexes containing C 2-symmetric bis(oxazoline) (BOX) ligands, (BOX)PdCl2 (2a-d), (BOX)Pd(Me)Cl (3a-d), and [(BOX)PdMe(2,6-Me2C5H3N)]+PF6 - (4a-d) [BOX: 2,2'-(2-propylidene)bis{(4R)-4-phenyl-5,5-dimethyl-2-oxazoline}, 2,2'-methylenebis{(4R)-4-phenyl-5,5-dimethyl-2-oxazoline}, 2,2'-methylenebis{(4R)-4,5,5-triphenyl-2-oxazoline}, and 2,2'-methylenebis{(4R,5S)-4,5-diphenyl-2-oxazoline}], were prepared, and their structures were determined by X-ray crystallography. It was found that substituents at the 5-position (Ph, Me) in addition to substituents on the bridgehead carbon directly affect the structure around palladium, especially the BOX bite angle and the dihedral angles between the phenyl rings at the 4-position and the N2Pd plane. Treatment of the bridged methylene proton in the BOX ligand (1b-d) with KH afforded the anionic BOX ligand; also, the neutral Pd complexes, (BOX)PdMe(2,6-Me2C5H3N) (5b-d), could thus be prepared by reaction with Pd(Me)Cl(cod) (cod = 1,5-cyclooctadiene); 5b-d showed strong coordination to Pd, as demonstrated by X-ray crystallographic analysis.

Effect of aryloxo substituents in ethylene polymerisation by tris(pyrazolyl)borate titanium(IV) complexes containing aryloxo ligands of type, TpTiCl2(OAr).

Tris(pyrazolyl)borate titanium(IV) complexes containing aryloxo ligand of type, TpTiCl(2)(OAr) [Tp = HB(pyrazolyl)(3); Ar = 2-(i)PrC(6)H(4), C(6)F(5), 2,6-Ph(2)C(6)H(3)], have been prepared, identified, and the structures of TpTiCl(2)(OC(6)F(5)), TpTiCl(2)(O-2,6-Ph(2)C(6)H(3)) have been determined by X-ray crystallography. The 2-(i)PrC(6)H(4) analogue exhibits especially high catalytic activities for ethylene polymerisation in the presence of methylaluminoxane (MAO), and the activity by TpTiCl(2)(OAr) increased in the order: Ar = 2-(i)PrC(6)H(4) > Ph > C(6)F(5) >> 2,6-Ph(2)C(6)H(3) > 2,6-Me(2)C(6)H(3), 2,6-(i)Pr(2)C(6)H(3).

Design and physicochemical properties of new fluorescent ligands of protein kinase C isozymes focused on CH/pi interaction.

Phorbol ester-type tumor promoters such as indolactam-V (IL-V, 1) bind to the C1 domains of protein kinase C (PKC) isozymes. A more convenient method to investigate the interaction between each tumor promoter and PKC C1 domain is needed. Focusing on our recent finding that the indole ring of IL-V is involved in the CH/pi interaction with Pro-11 of the PKCdelta-C1B domain, we developed new fluorescent probes (2-4) from IL-V by forming a pyrroloindazole ring. Compound 2 without a substituent at the pyrroloindazole ring bound most strongly to PKC C1 domains with a potency similar to IL-V, but its fluorescent intensity was the weakest of any of the probes. Although the binding affinity of 3 with a methyl group was significantly weaker than that of IL-V, 4 with a trifluoromethyl group showed moderate affinity and the most potent fluorescence intensity. The fluorescence intensity and emission maxima of 4 changed significantly when bound to the PKCdelta-C1B peptide in both the presence and absence of phosphatidylserine. These results suggest that 4 could be a useful probe for analyzing the interaction of tumor promoters with PKC C1 domains.