Browsing by Subject "Image Processing"

Now showing 1 - 4 of 4
  • Thumbnail Image
    Item
    Atomization of viscous fluids using counterflow nozzle
    (2020-08) Rangarajan, Roshan
    In the present work, we study the enhanced atomization of viscous liquids by using a novel twin-fluid atomizer. A two-phase mixing region is developed within the nozzle using counterflow configuration by supplying air and liquid streams in opposite directions. Detailed qualitative and quantitative measurements for droplet size were conducted using shadowgraph technique. Near-field spray images from the nozzle exit suggest that the spray emerges out as a fine droplets with little scope for further atomization. The performance of this nozzle is compared to that of ‘flow-blurring’ nozzle. Three test liquids (Water, Propylene Glycol & Glycerol 85% soln) are used to vary the liquid viscosity in the range from 1 to 133.5 mPa.s. The counterflow nozzle produces a spray whose characteristics are relatively insensitive to fluid viscosity over the range of gas-liquid mass flow ratios between 0.25 and 1.
  • Thumbnail Image
    Item
    Bridging Mri Reconstruction Across Eras: From Novel Optimization Of Traditional Methods To Efficient Deep Learning Strategies
    (2024-03) Gu, Hongyi
    Magnetic Resonance Imaging (MRI) has been extensively used as a non-invasive modality for imaging the human body. Despite substantial advances over the past decades, scan duration remains as a principal issue for MRI scanning, requiring novel techniques to accelerate data acquisition. Such techniques are poised to improve clinical patient throughput, reduce motion artifacts, enhance subject comfort, and allow higher resolution imaging in many applications. Several methods have been proposed to accelerate MRI scans. In parallel imaging (PI), k-space data was acquired at a sub-Nyquist rate with with multiple receiver coils, and the redundancy among these coils were used for image reconstruction. Following the clinical impact and success of PI methods, compressed sensing (CS) techniques were developed to reconstruct images by using compressibility of images in a pre-specified linear transform domain. Transform learning (TL) was another line of work that learned the linear transforms from data, while enforcing sparsity as in CS. Recently, deep learning (DL) has shown great promise for MRI reconstruction, especially at high acceleration rates where other traditional methods would fail. Specially, physics-guided DL (PG-DL) unrolls a traditional optimization algorithm for solving regularized least squares for a fixed number of iterations, and uses neural networks to implicitly perform regularization. These unrolled networks are trained end-to-end with large databases, using well-designed loss functions and advanced optimizers, usually using a reference fully-sampled image for supervised learning. Several approaches have noted the difficulty or impossibility of acquiring fully-sampled data in various MRI applications. Among these, self-supervised learning with data undersmapling (SSDU) was developed to allow training without fully-sampled data, and multi-mask SSDU was subsequently proposed for better reconstruction quality at high acceleration rates. Although PG-DL generally shows strong ability for excellent reconstruction performance, there are concerns for generalizabilty, interpretability and stability issues. In this thesis, we aimed to bridge the gap between traditional and DL methods, while also extending the utility of DL methods for non-Cartesian imaging. We first revisited l1-wavelet CS reconstruction for accelerated MRI by using modern data science tools similar to those used in DL for optimized performance. We showed that our proposed optimization approach improved traditional CS, and further performance boost was observed by incorporating wavelet subband processing and reweighted l1 minimization. The final version reached a performance similar to state-of-the-art PG-DL, while preserving better interpretability by solving a convex optimization problem in inference time. Second, we combined ideas from CS, TL and DL to enable the learning of deep linear convolutional transforms in a format similar to PG-DL. Our proposed method performed better than CS and TL, and gave similar performance as state-of-the-art PG-DL. It used a linear representation of image as regularization at inference time, and enabled convex sparse image reconstruction that may have better robustness, stability and generalizability properties. Third, we adapted a self-supervised PG-DL technique to non-Cartesian trajectories and showed its potential for reconstructing 10-fold accelerated spiral fMRI multi-echo acquisitions. Our proposed approach gave substantial improvements in reconstructed image quality over conventional methods. Furthermore, the blood oxygenation level dependent (BOLD) signal analysis of our proposed method provided meaningful sensitivities, with similar activation patterns and extent to the expected baselines.
  • Thumbnail Image
    Item
    Combating Fusarium Head Blight Resistance in Wheat with Genomic Selection and Computer Vision Technology
    (2022-01) Adeyemo, Emmanuel
    Fusarium head blight (FHB), primarily caused by Fusarium graminarum, Schwabe, is a devastating fungal disease that limits wheat production globally and can significantly reduce yield and grain quality. At the University of Minnesota, screening for FHB begins at the F5 stage and continues annually until the line is released as a cultivar. Before implementing genomic selection at the F5 stage in 2016, we evaluated ~ 3,000 F5 lines annually in field nurseries. The use of genomic selection allowed the prediction of untested lines with a training population of 500 lines selected by pedigree information. The first study showed that a set of 200 lines selected by genomic relationship led to predictive abilities of up to 0.49, whereas a larger, randomly selected subset of 500 F5 lines had a maximum predictive ability of 0.34. While the addition of parents also led to increased predictive abilities, the increments were not significant in most cases. The second study examined the merit of incorporating available germplasm into our existing genomic selection pipeline. We observed that training populations that contained historical data were less useful while those that included a subset of 200 F5 lines selected by genomic relationship, were more effective for predicting FHB. The third study demonstrated the use of computer vision to estimate the percentage of kernels damaged by Fusarium. We utilized 85 samples containing five check cultivars with varying levels of FHB susceptibility and maturity and achieved an accuracy of 90%. Additional studies should be done to assess the utilization of this technology among our experimental lines.
  • Thumbnail Image
    Item
    Studies on apple peel color regulation.
    (2009-05) Rabinovich, Adriana Telias
    One of the most important factors determining apple [Malus pumila P. Mill.] market acceptance is peel color. Most apple cultivars (e.g. `Royal Gala') produce fruit with a defined fruit pigment pattern, but in the case of `Honeycrisp' apple, trees can produce fruits of two different kinds: striped and blushed. The causes of this phenomenon are unknown. We compared 'Honeycrisp' fruit from trees that were propagated from buds occurring on branches carrying only blushed or only striped fruit and concluded that blushed trees tend to produce a higher percentage of blushed fruit than striped trees, indicating a mechanism conserved through cell division. The percentage of blushed fruit on any given tree changed from year to year. Blushed and striped fruit occurred together on the same branch, and even on the same spur, with fruits located in the outer canopy being more likely to be striped. Higher crop loads were associated with a lower percentage of blushed fruit on the tree. Blushed and striped fruit do not consistently differ in their maximum pigment accumulation before ripening. The comparison of average hue angle for the whole peel at harvest indicates that blushed fruit are redder on average. We have also shown that striped areas of `Honeycrisp' and `Royal Gala' are due to sectorial increases in anthocyanin concentration. Transcript levels of the major biosynthetic genes and MdMYB10, a transcription factor that upregulates apple anthocyanin production, correlated with increased anthocyanin concentration in stripes. However, changes in the promoter and coding sequence of MdMYB10 do not correlate with skin pattern in `Honeycrisp' and other cultivars differing in peel pigmentation patterns. A survey of methylation levels throughout the coding region of MdMYB10 and a 2.5 kb region 5' of the ATG translation start site indicated that an area 900 bp long, starting 1400 bp upstream of the translation start site, is highly methylated. Comparisons of methylation levels of red and green stripes indicated that the degree of methylation of the MdMYB10 promoter is likely to be associated with the presence of stripes in these cultivars, with red stripes having lower methylation levels. Methylation may be associated with the presence of a TRIM retrotransposon within the promoter region, but the presence of the TRIM element alone cannot explain the phenotypic variability observed in `Honeycrisp'. We suggest that methylation in the MdMYB10 promoter is more variable in `Honeycrisp' than in `Royal Gala', leading to more variable color patterns in the peel of this cultivar.