Improving the Quality of Synthetic FLAIR Images with Deep Learning Using a Conditional Generative Adversarial Network for Pixel-by-Pixel Image Translation.

BACKGROUND AND PURPOSE: Synthetic FLAIR images are of lower quality than conventional FLAIR images. Here, we aimed to improve the synthetic FLAIR image quality using deep learning with pixel-by-pixel translation through conditional generative adversarial network training. MATERIALS AND METHODS: Forty patients with MS were prospectively included and scanned (3T) to acquire synthetic MR imaging and conventional FLAIR images. Synthetic FLAIR images were created with the SyMRI software. Acquired data were divided into 30 training and 10 test datasets. A conditional generative adversarial network was trained to generate improved FLAIR images from raw synthetic MR imaging data using conventional FLAIR images as targets. The peak signal-to-noise ratio, normalized root mean square error, and the Dice index of MS lesion maps were calculated for synthetic and deep learning FLAIR images against conventional FLAIR images, respectively. Lesion conspicuity and the existence of artifacts were visually assessed. RESULTS: The peak signal-to-noise ratio and normalized root mean square error were significantly higher and lower, respectively, in generated-versus-synthetic FLAIR images in aggregate intracranial tissues and all tissue segments (all P < .001). The Dice index of lesion maps and visual lesion conspicuity were comparable between generated and synthetic FLAIR images (P = 1 and .59, respectively). Generated FLAIR images showed fewer granular artifacts (P = .003) and swelling artifacts (in all cases) than synthetic FLAIR images. CONCLUSIONS: Using deep learning, we improved the synthetic FLAIR image quality by generating FLAIR images that have contrast closer to that of conventional FLAIR images and fewer granular and swelling artifacts, while preserving the lesion contrast.

Effect of Gadolinium on the Estimation of Myelin and Brain Tissue Volumes Based on Quantitative Synthetic MRI.

BACKGROUND AND PURPOSE: The effect of gadolinium on the estimation of myelin has not been reported. The aim of the current study was to investigate the effects of gadolinium on automatic myelin and brain tissue volumetry via quantitative synthetic MR imaging. MATERIALS AND METHODS: The study included 36 patients who were referred for brain metastases screening, and quantitative synthetic MR imaging data before and after gadolinium-based contrast agent administration were analyzed retrospectively. Brain metastases were detected in 17 patients. WM volume, GM volume, CSF volume, non-WM/GM/CSF volume, myelin volume, brain parenchymal volume, myelin fraction (myelin volume/brain parenchymal volume), and intracranial volume were estimated. T1 and T2 relaxation times, proton density, and myelin partial volume per voxel averaged across the brain parenchyma were also analyzed. RESULTS: In patients with and without metastases after gadolinium-based contrast agent administration, measurements of WM and myelin volumes, and myelin fraction were significantly increased (+26.65 and +29.42 mL, +10.14 and +12.46 mL, +0.88% and +1.09%, respectively), whereas measurements of GM, CSF, brain parenchymal, and intracranial volumes were significantly decreased (-36.23 and -34.49 mL, -20.77 and -18.94 mL, -6.76 and -2.84 mL, -27.41 and -21.84 mL, respectively). Non-WM/GM/CSF volume did not show a significant change. T1, T2, and proton density were significantly decreased (-51.34 and -46.84 ms, -2.67 and -4.70 ms, -1.05%, and -1.28%, respectively) after gadolinium-based contrast agent administration, whereas measurements of myelin partial volume were significantly increased (+0.78% and +0.75%, respectively). CONCLUSIONS: Gadolinium had a significant effect on the automatic calculation of myelin and brain tissue volumes using quantitative synthetic MR imaging, which can be explained by decreases in T1, T2, and proton density.

Automated brain tissue and myelin volumetry based on quantitative MR imaging with various in-plane resolutions.

BACKGROUND AND PURPOSE: Segmented brain tissue and myelin volumes can now be automatically calculated using dedicated software (SyMRI), which is based on quantification of R1 and R2 relaxation rates and proton density. The aim of this study was to determine the validity of SyMRI brain tissue and myelin volumetry using various in-plane resolutions. METHODS: We scanned 10 healthy subjects on a 1.5T MR scanner with in-plane resolutions of 0.8, 2.0 and 3.0mm. Two scans were performed for each resolution. The acquisition time was 7-min and 24-sec for 0.8mm, 3-min and 9-sec for 2.0mm and 1-min and 56-sec for 3.0mm resolutions. The volumes of white matter (WM), gray matter (GM), cerebrospinal fluid (CSF), non-WM/GM/CSF (NoN), brain parenchymal volume (BPV), intracranial volume (ICV) and myelin were compared between in-plane resolutions. Repeatability for each resolution was then analyzed. RESULTS: No significant differences in volumes measured were found between the different in-plane resolutions, except for NoN between 0.8mm and 2.0mm and between 2.0mm and 3.0mm. The repeatability error value for the WM, GM, CSF, NoN, BPV and myelin volumes relative to ICV was 0.97%, 1.01%, 0.65%, 0.86%, 1.06% and 0.25% in 0.8mm; 1.22%, 1.36%, 0.73%, 0.37%, 1.18% and 0.35% in 2.0mm and 1.18%, 1.02%, 0.96%, 0.45%, 1.36%, and 0.28% in 3.0mm resolutions. CONCLUSION: SyMRI brain tissue and myelin volumetry with low in-plane resolution and short acquisition times is robust and has a good repeatability so could be useful for follow-up studies.

Increase in thermostability of recombinant barley beta-amylase by random mutagenesis.

Random mutations were introduced into recombinant barley beta-amylase by modified PCR to increase its thermostability. Two clones were obtained. One was found to have a change of Ser-351 to Pro and another, a change of Ala-376 to Ser, and 2.3 degrees C and 1.0 degrees C increases, respectively, in the thermostabilities compared with that of native recombinant beta-amylase.

Role of the C-terminal region of beta-amylase from barley.

To investigate the role of the C-terminal region of barley beta-amylase, plasmid delta 54 was constructed with an expression vector (pBETA92) of barley beta-amylase by site-directed mutagenesis. Escherichia coli JM109 harboring plasmid delta 54 was expected to express delta 54 beta-amylase in which 54 amino acid residues were deleted from the C-terminus. The enzyme production started in the logarithmic phase, increased linearly, and reached a maximum after 12 h. delta 54 beta-amylase gave a single activity band on isoelectric focusing (pI 6.85). delta 54 beta-amylase was purified from the cells by consecutive alpha-cyclodextrin/Sepharose 6B column chromatography. A comparison of the properties of the mutant enzyme with those of the original recombinant beta-amylase [Biosci. Biotech. Biochem. (1994) 58, 1080-1086] revealed two major differences. First, the original recombinant beta-amylase showed heterogeneity on isoelectric focusing, but delta 54 beta-amylase gave a single main band of protein (pI 6.85). Therefore, the isoelectrophoretic heterogeneity of the original recombinant beta-amylase was apparently due to its C-terminal region. Secondly, delta 54 beta-amylase lacked thermostability. Therefore, it was concluded that the C-terminal region was significantly involved in the thermostability of beta-amylase.

Expression in Escherichia coli of cDNA encoding barley beta-amylase and properties of recombinant beta-amylase.

To express the cloned beta-amylase cDNA in Escherichia coli under control of the tac promoter, a plasmid pBETA92 was constructed. The plasmid consisted of 6312 bp. An extract of E. coli JM109 harboring pBETA92 had beta-amylase activity that produced beta-maltose from soluble starch. The enzyme production started in the logarithmic phase, increased linearly, and reached a maximum after 12 h. The recombinant barley beta-amylase gave two major (pI 5.43 and 5.63) and four minor (pI 5.20, 5.36, 5.80, and 6.13) activity bands on isoelectric focusing, and their pIs didn't change throughout the incubation. But Western blot analysis found that one beta-amylase having a molecular weight of about 56,000 was synthesized. The recombinant beta-amylase was purified from the cells by consecutive column chromatography. The purified enzyme gave a single band of protein on SDS-PAGE but showed heterogeneity on isoelectric focusing. The N-terminal amino acid sequence showed that the recombinant beta-amylase lacked four amino acids at positions 2-5 (Glu-Val-Asn-Val) when compared with the presumed amino acid sequence of barley beta-amylase. Therefore, the recombinant beta-amylase consisted of 531 amino acids, and its molecular weight was calculated to be 59,169. The N-terminal amino acid sequence of the recombinant beta-amylase and the nucleotide sequence of the junction position in plasmid pBETA92 indicated that GTG (Val-5 in the case of barley beta-amylase) at positions 27-29 from the SD sequence (AGGA) was the translation initiation codon. The properties of the recombinant beta-amylase were almost the same as those of barley beta-amylase except for the pI and the Km values for maltohexaose and maltoheptaose.(ABSTRACT TRUNCATED AT 250 WORDS)

Molecular cloning and analysis of the yeast flocculation gene FLO1.

The DNA sequence of the flocculation gene FLO1 of Saccharomyces cerevisiae, which is located on chromosome I (Watari et al., 1989) was determined. The sequence contains a large open reading frame (ORF) of 2586 bp and codes for a protein of 862 amino acids. However, further study (genomic Southern and polymerase chain reaction analyses) indicated that the gene we cloned was not the intact FLO1 gene but a form with an approximately 2 kb deletion in the ORF region. The intact FLO1 gene was then cloned and its nucleotide sequence determined. The sequence revealed that the ORF of the intact gene is composed of 4611 bp which code for a protein of 1537 amino acids. A remarkable feature of the putative Flo1 protein is that it contains four families of repeated sequences composed of 18, 2, 3 and 3 repeats and that it has a large number of serines and threonines. In the deleted FLO1 form, a large part of these repeated sequences was missing. The N- and C-terminal regions are hydrophobic and both contain a potential membrane-spanning region, suggesting that the Flo1 protein is an integral membrane protein and a cell wall component.

PCR cloning and sequencing of the beta-amylase cDNA from barley.

Polymerase chain reaction (PCR) amplification of mRNA from developing barley (cultivar Haruna two-rows) endosperm was used to clone and sequence full-length cDNA encoding beta-amylase. The beta-amylase cDNA was 1,775 bp in length. The beta-amylase was deduced to be composed of 535 amino acid residues and its molecular weight was calculated to be 59,610. Kreis et al. reported that the beta-amylase cDNA from barley (cultivar Hiproly) was 1,754 bp in length and coded for a polypeptide of 535 amino acids [Eur. J. Biochem. (1987) 169, 517-525]. A comparison of the beta-amylase sequences from Haruna two-rows and Hiproly barleys revealed nine differences in the nucleotide sequence which resulted in three changes in the amino acid residues and 21 additional nucleotides at its 3'-end in the cultivar Haruna two-rows. The three changes were as follows: Ala-233, Ser-347, Met-527 (Haruna two-rows) and Val-233, Met-347, Ile-527 (Hiproly). Lundgard and Svensson pointed out that 23 amino acid residues of the peptide fragment derived from the COOH-terminal region of barley (cultivar Gula) beta-amylase were in agreement with the deduced amino acid sequence reported by Kreis et al., with the exception of a single position (Met-527 compared to Ile) [Carlsberg Res. Commun. (1986) 51, 487-491]. Our findings described above showed Met-527 is reasonable. In the cases of beta-amylases from soybean and sweet potato, the positions that corresponded to those at 233 and 347 in the amino acid sequence of beta-amylase from barley were Ala and Ser, respectively.(ABSTRACT TRUNCATED AT 250 WORDS)