Sebastian Berisha, Ph.D.
Assistant Professor
- Milwaukee WI UNITED STATES
- Diercks Hall DH415
- Electrical Engineering and Computer Science
Dr. Sebastian Berisha is a computer science professor at the Milwaukee School of Engineering.
Education, Licensure and Certification
Ph.D.
Computer Science and Informatics
Emory University
2014
M.S.
Computer Science
Wake Forest University
2009
B.S.
Computer Science and Mathematics
Averett University
2003
Biography
Accomplishments
IEEE Transactions on Geoscience and Remote Sensing: top 15 reviewer of the year
2017
Postdoctoral Fellowship, University of Houston
2015 - 2018
Postdoctoral Fellowship, University of Pennsylvania
2014 - 2015
National Library of Medicine Postdoctoral Fellowship
2018
Ph.D. Fellowship, Emory University
2009 - 2014
Selected Publications
Three-dimensional GPU-accelerated active contours for automated localization of cells in large images
PloS oneLotfollahi, M., Berisha, S., Saadatifard, L., Montier, L., Žiburkus, J., Mayerich, D.
2019
Cell segmentation in microscopy is a challenging problem, since cells are often asymmetric and densely packed. Successful cell segmentation algorithms rely identifying seed points, and are highly sensitive to variablility in cell size. In this paper, we present an efficient and highly parallel formulation for symmetric three-dimensional contour evolution that extends previous work on fast two-dimensional snakes. We provide a formulation for optimization on 3D images, as well as a strategy for accelerating computation on consumer graphics hardware. The proposed software takes advantage of Monte-Carlo sampling schemes in order to speed up convergence and reduce thread divergence. Experimental results show that this method provides superior performance for large 2D and 3D cell localization tasks when compared to existing methods on large 3D brain images.
Digital Staining of High-Definition Fourier Transform Infrared (FT-IR) Images Using Deep Learning
Applied SpectroscopyLotfollahi, M., Berisha, S., Daeinejad, D., Mayerich, D
2019
Histological stains, such as hematoxylin and eosin (H&E), are routinely used in clinical diagnosis and research. While these labels offer a high degree of specificity, throughput is limited by the need for multiple samples. Traditional histology stains, such as immunohistochemical labels, also rely only on protein expression and cannot quantify small molecules and metabolites that may aid in diagnosis. Finally, chemical stains and dyes permanently alter the tissue, making downstream analysis impossible. Fourier transform infrared (FT-IR) spectroscopic imaging has shown promise for label-free characterization of important tissue phenotypes and can bypass the need for many chemical labels. Fourier transform infrared classification commonly leverages supervised learning, requiring human annotation that is tedious and prone to errors. One alternative is digital staining, which leverages machine learning to map IR spectra to a corresponding chemical stain. This replaces human annotation with computer-aided alignment. Previous work relies on alignment of adjacent serial tissue sections. Since the tissue samples are not identical at the cellular level, this technique cannot be applied to high-definition FT-IR images. In this paper, we demonstrate that cellular-level mapping can be accomplished using identical samples for both FT-IR and chemical labels. In addition, higher-resolution results can be achieved using a deep convolutional neural network that integrates spatial and spectral features.
Automated osteosclerosis grading of clinical biopsies using infrared spectroscopic imaging
BioRxivMankar, R., Bueso-Ramps, C.E., Yin, C.C., Hidalgo-Lopez, J.E., Berisha, S., Kansiz, M., Mayerich, D.
2019
Osteosclerosis and myefibrosis are complications of myeloproliferative neoplasms. These disorders result in excess growth of trabecular bone and collagen fibers that replace hematopoietic cells, resulting in abnormal bone marrow function. Treatments using imatinib and JAK2 pathway inhibitors can be effective on osteosclerosis and fibrosis, therefore accurate grading is critical for tracking treatment effectiveness. Current grading standards use a four-class system based on analysis of biopsies stained with three histological stains: hematoxylin and eosin (H&E), Masson’s trichrome, and reticulin. However, conventional grading can be subjective and imprecise, impacting the effectiveness of treatment. In this paper, we demonstrate that mid-infrared spectroscopic imaging may serve as a quantitative diagnostic tool for quantitatively tracking disease progression and response to treatment. The proposed approach is label-free and provides automated quantitative analysis of osteosclerosis and collagen fibrosis.
Deep learning for FTIR histology: leveraging spatial and spectral features with convolutional neural networks
AnalystBerisha, S., Lotfollahi, M., Jahanipour, J., Gurcan, I., Walsh, M., Bhargava, R., Van Nguyen, H., Mayerich, D.
2019
Current methods for cancer detection rely on tissue biopsy, chemical labeling/staining, and examination of the tissue by a pathologist. Though these methods continue to remain the gold standard, they are non-quantitative and susceptible to human error. Fourier transform infrared (FTIR) spectroscopic imaging has shown potential as a quantitative alternative to traditional histology. However, identification of histological components requires reliable classification based on molecular spectra, which are susceptible to artifacts introduced by noise and scattering. Several tissue types, particularly in heterogeneous tissue regions, tend to confound traditional classification methods. Convolutional neural networks (CNNs) are the current state-of-the-art in image classification, providing the ability to learn spatial characteristics of images. In this paper, we demonstrate that CNNs with architectures designed to process both spectral and spatial information can significantly improve classifier performance over per-pixel spectral classification. We report classification results after applying CNNs to data from tissue microarrays (TMAs) to identify six major cellular and acellular constituents of tissue, namely adipocytes, blood, collagen, epithelium, necrosis, and myofibroblasts. Experimental results show that the use of spatial information in addition to the spectral information brings significant improvements in the classifier performance and allows classification of cellular subtypes, such as adipocytes, that exhibit minimal chemical information but have distinct spatial characteristics. This work demonstrates the application and efficiency of deep learning algorithms in improving the diagnostic techniques in clinical and research activities related to cancer.
Mitigating fringing in discrete frequency infrared imaging using time-delayed integration
Biomedical Optics ExpressRan, S., Berisha, S., Mankar, R., Shih, W.C., Mayerich, D.
2018
Infrared (IR) spectroscopic microscopes provide the potential for label-free quantitative molecular imaging of biological samples, which can be used to aid in histology, forensics, and pharmaceutical analysis. Most IR imaging systems use broadband illumination combined with a spectrometer to separate the signal into spectral components. This technique is currently too slow for many biomedical applications such as clinical diagnosis, primarily due to the availability of bright mid-infrared sources and sensitive MCT detectors. There has been a recent push to increase throughput using coherent light sources, such as synchrotron radiation and quantum cascade lasers. While these sources provide a significant increase in intensity, the coherence introduces fringing artifacts in the final image. We demonstrate that applying time-delayed integration in one dimension can dramatically reduce fringing artifacts with minimal alterations to the standard infrared imaging pipeline. The proposed technique also offers the potential for less expensive focal plane array detectors, since linear arrays can be more readily incorporated into the proposed framework.
Phase retrieval based deconvolution algorithm in optical systems
2017 IEEE Global Conference on Signal and Information Processing (GlobalSIP)Qin, S., Berisha, S., Mayerich, D., Han, Z.
2017
In an optical imaging system, the retrieved image of an object is blurred by the point spread function (PSF) of the system, and cannot exactly represent the object. Deconvolution is an effective method to recover the object from the blurred image and improve the resolution of the optical system. But in real optical system, the detector only measures the intensity of the light, not the phase. So, the information used in image deconvolution is incomplete and reduces the quality of the final result. Fortunately, phase retrieval theory provides an effective method to recover the complex field from an intensity image. With the phase information, we can get use the complete field information to reconstruct the object more accurately. In this paper, we propose a method based on the phase retrieval to recover the object from the image. Simulation results indicate that our proposed method has improved performance.
BIM-Sim: Interactive Simulation of Broadband Imaging Using Mie Theory
Frontiers in PhysicsBerisha, S., van Dijk, T., Bhargava, R., Carney, P.S., Mayerich, D.
2017
Understanding the structure of a scattered electromagnetic (EM) field is critical to improving the imaging process. Mechanisms such as diffraction, scattering, and interference affect an image, limiting the resolution, and potentially introducing artifacts. Simulation and visualization of scattered fields thus plays an important role in imaging science. However, EM fields are high-dimensional, making them time-consuming to simulate, and difficult to visualize. In this paper, we present a framework for interactively computing and visualizing EM fields scattered by micro and nano-particles. Our software uses graphics hardware for evaluating the field both inside and outside of these particles. We then use Monte-Carlo sampling to reconstruct and visualize the three-dimensional structure of the field, spectral profiles at individual points, the structure of the field at the surface of the particle, and the resulting image produced by an optical system.
SIproc: an open-source biomedical data processing platform for large hyperspectral images
AnalystBerisha, S., Chang, S., Saki, S., Daeinejad, D., He, Z., Mankar, R., Mayerich, D.
2017
There has recently been significant interest within the vibrational spectroscopy community to apply quantitative spectroscopic imaging techniques to histology and clinical diagnosis. However, many of the proposed methods require collecting spectroscopic images that have a similar region size and resolution to the corresponding histological images. Since spectroscopic images contain significantly more spectral samples than traditional histology, the resulting data sets can approach hundreds of gigabytes to terabytes in size. This makes them difficult to store and process, and the tools available to researchers for handling large spectroscopic data sets are limited. Fundamental mathematical tools, such as MATLAB, Octave, and SciPy, are extremely powerful but require that the data be stored in fast memory. This memory limitation becomes impractical for even modestly sized histological images, which can be hundreds of gigabytes in size. In this paper, we propose an open-source toolkit designed to perform out-of-core processing of hyperspectral images. By taking advantage of graphical processing unit (GPU) computing combined with adaptive data streaming, our software alleviates common workstation memory limitations while achieving better performance than existing applications.
Impact of Respiration on LV Volume and Function Using rt-MRI
Journal of Cardiovascular Magnetic ResonanceContijoch, F., Berisha, S., Gorman, J.H., Gorman, R.C., Witschey, W.R., Han, Y.
2016
ECG-gated cardiac MRI acquired during breathholds is the gold standard for volumetric evaluation of patients, and clinically, ejection fraction is used as a surrogate for function. We hypothesized that the breathholds alter hemodynamic measurements by changing the loading conditions of the heart as well as the heart rate. Real-time MRI and semi-automated LV endocardial segmentation can be used to quantify slice volume during respiration. We derive global hemodynamic measurements during breathholds and free respiration to measure changes related to respiration.
Measurement of Myocardial T1ρ with a Motion Corrected, Parametric Mapping Sequence in Humans
PloS oneBerisha, S., Han, J., Shahid, M., Han, Y., Witschey, W.R.
2016
Purpose: To develop a robust T1ρ magnetic resonance imaging (MRI) sequence for assessment of myocardial disease in humans.