BioE 1330 Review Other PDF

BioE 1330 - Review Other 9/27/2018 Instructions: On the Answer Sheet, enter your 2-digit ID number (with a leading 0 if needed) in the boxes of the ID section. Fill in the corresponding numbered circles. Answer each of the numbered questions by filling in the corresponding circles in...

BioE 1330 - Review Other 9/27/2018 Instructions: On the Answer Sheet, enter your 2-digit ID number (with a leading 0 if needed) in the boxes of the ID section. Fill in the corresponding numbered circles. Answer each of the numbered questions by filling in the corresponding circles in the numbered question section. Print your name in the space at the bottom of the answer sheet. Sign here stating that you have neither given nor received help. your signature 1. Which of the following statements is false about this matrix (or all are true):   1 0 2 0 1 3 0 0 1 A. It performs a geometric (rigid body) transformation. B. It performs pure translation (no rotation or scaling). C. All are true. D. It transforms coordinates for 2-dimensional locations. E. It operates in homogeneous coordinates. 2. The following are true about segmentation in image analysis, except A. Manual segmentation suffers from inter-observer variation, in which different people will perform different seg- mentations on a given image. B. Manual segmentation suffers from intra-observer variation, in which the same person will perform different segmentations at different times on a given image. C. It involves the labeling of individual pixels (or voxels) according to some model of the expected anatomical target. D. Automated segmentation suffers from a complete lack of any ability to incorporate prior information about the target anatomical structure. E. One form of segmentation is deformable contours (snakes), in which an initial contour is placed on the image, and then the contour is changed to better fit an underlying anatomical structure in the image. 3. The following are true about registration in image analysis, except A. It involves a transform to remap one image onto the coordinate system of another image in such a way that a given anatomical location corresponds to the same coordinates in both images. B. It requires a metric to determine the fitness value of the registration, which is optimized to determine the best transform to remap one image onto the coordinate system of the other. C. It may be performed on 2D, but not 3D, image data. D. Examples of registration transforms include, geometric (rigid body), similarity, affine, projection, and deformation (warping). E. Similarity metrics for registration include root mean square (RMS), sum absolute different (SAD), covariance, correlation, and mutual information. 1 4. The following are true about optical microscopy, except A. The resolution of a light microscope is limited by the particle nature of photons, which models electromagnetic radiation as individual “bullets.” B. If a screen is placed in the plane of a real image in a compound microscope, the image will be visible on the screen. C. The virtual image seen through a typical compound microscope appears to be located where there is no actual source of photons, that location being determined by tracing the rays that emerge from the eyepiece lens. D. Confocal microscopy uses pin-holes in addition to lenses to sharpen the focal plane. E. 3D microscopic image data may be obtained using confocal microscopy or optical coherence tomography (OCT). 5. The following are true about mathematical morphology, except A. Its operations depend upon the definiton of a structuring element, whose shape can vary depending on the desired effect. B. It includes two secondary operations, opening and closing. C. It operates only on binary images, consisting of 1’s and 0’s. D. It is particularly useful for eliminating small objects and filling in holes in objects. E. It is based on two fundamental operations, erosion and dilation. 6. The following are true about quaternions, except (or all are true) A. They are the 3D equivalent of phasors in 2D. B. All are true. C. They can be multiplied by each other, effecting rotation from any of the x, y, or z axes to any of the other axes. D. They are used to represent orientation in 3D. E. They consist of one real and 3 mutually orthogonal imaginary axes. 7. The following are true about lenses in optical imaging, except A. They create standing waves according to Huygen’s principle. B. Although a prism is capable of breaking white light into its constituent colors through the phenomenon of dispersion, lenses do not exhibit this phenomenon, and focus all colors identically. C. The resolution of what can be seen through a lens is limited by the wave nature of light. D. They slow down the speed of light along a particular ray, introducing a delay proportional to the thickness of the glass along that particular ray. E. When looking through a convex lens (such as a magnifying glass) at an object, a virtual image may be observed, from which photons appear to be emanating. 2 8. The following are true about depth of focus (range resolution) in conventional optical microscopy, except (or all are true) A. Optical Coherence Tomography has far better range resolution than optical microscopy. B. As in ultrasound, range resolution in conventional optical microscopy is determined by time of flight and is therefore independent on the field pattern. C. Depth of focus can be reduced (range resolution improved) by using confocal microscopy, which incorporates pinhole apertures in addition to lenses. D. All are true. E. In the focal region, resolution in depth is generally less than resolution within the image plane (transverse to the optical axis). 9. The following are true about real and virtual images, except (or all are true) A. The virtual image of an object placed in front of a mirror is located behind the mirror. B. If a screen is placed in the plane of a virtual image, the image will become visible on the screen. C. All are true. D. A virtual image is an image that only appears to be at a particular location, determined by tracing the rays that emerge from an optical device. E. A real image is located in the plane of convergence for the light rays that originate from a given object. 10. The following are true about segmentation in medical imaging, except (or all are true) A. It can be performed in 2D or in 3D. B. All are true. C. If a 3D object is segmented using a set of 2D slices, the orientation of the slices can produce a situation where the object appears in two separated regions within a given slice. D. It is limited by the precision with which labels have been applied to anatomical structures. E. It is an attempt to label pixels (or voxels) according to the underlying physical structures. 11. Manual segmentation of medical images suffers from the following limitations, except A. It is generally performed in 2D rather than in 3D. B. Inter-observer variation can be significant. C. It can be time-consuming, especially when done on multiple slices in 3D data sets. D. Human vision does not have sufficient dynamic range to cover the dynamic range of a typical display on which medical images are shown. E. Intra-observer variation can be significant. 3 12. The following are true about registration in medical imaging, except (or all are true) A. It can only be done properly if the two images are of the same imaging modality (e.g., CT, MR, etc.). B. It uses global transforms (such as geometric, similarity, affine), as well as possibly deformation to move one of the images into the other images physical coordinate system, followed by resampling onto the other image’s lattice. C. It consists of properly superimposing the coordinate systems of two images such that the underlying anatomy corresponds by pixel location. D. All are true. E. It requires a metric of the accuracy of the registration to optimize. 13. Which of the following 3D transforms can produce non-parallel lines from parallel lines (or none can)? A. Geometric. B. Projection. C. Similarity. D. None of the others. E. Affine. 14. The following are true about the spatial gradient of image intensity in medical imaging, except (or all are true) A. It is frequently computed using convolution of the image with a set of edge detection kernels, one for each dimension in the image yielding that component of the gradient. B. At the boundary between one homogeneous region and another with differing intensity, it is a vector approximately normal to the boundary (approximately because of spatial sampling in the image). C. It can exist in 2D and 3D images D. The gradient magnitude is a scalar measure of boundary strength, independent of boundary orientation. E. All are true. 15. The following are true about rendering opaque surfaces in medical imaging, except (or all are true) A. All are true. B. It is often accomplished by tessellation, in which individual small planar elements (usually triangles) are rendered using a lighting model based on orientation and location. C. A Z-buffer may be used to guarantee that opaque objects properly occlude either other based on distance from the viewer, independent of the rendering order. D. Marching cubes is a commonly used algorithm to quickly build tessellated iso-surfaces between two intensity levels. E. Texture mapping is a method of interpolating 2D or 3D image data onto a tessellated surface to be rendered graphically in 3D space. 4 16. The transform matrix below performs which operations? A. Projection and Translation B. Scaling and Rotation C. Scaling and Translation. D. Rotation and Translation E. Projection and Scaling 17. The following are true about convolving and image with a Gaussian, except A. All are true. B. It removes noise within otherwise homogeneous objects. C. Convolving an image with a Gaussian can be done in 2D but not in 3D. D. Convolving an image with the derivative of a Gaussian finds edges and removes noise. E. It blurs edges between objects with different intensities. 18. The following are true about machine learning, except A. It requires training data from which patterns are extracted. B. It is often based on layers of neural networks, with the number of layers becoming large in the case of deep learning. C. It is a form of connectionist rather than symbolic artificial intelligence. D. It may function in ways not understood by humans. E. It cannot be used to analyze images, but is only effective for problems based on words. 19. The following are examples of Point Operators on intensity, that is, each pixel in the output image is a function of only the equivalent pixel in the input image, except (or all are examples). A. Mapping the intensity values in an image to a new set of values such that the histogram of intensity is uniform (histogram equalization). B. All are examples. C. Adjusting the brightness and contrast of an image. D. Convolution with an edge detection kernel to find boundaries. E. Thresholding the intensity of an image to identify pixels above a given threshold. For official use only permutation number = 1249 5 BioE 1330 - Review Other 9/27/2018 Answer Sheet - Correct answer is A for all questions 1. Which of the following statements is false about this matrix (or all are true):   1 0 2 0 1 3 0 0 1 A. All are true. B. It performs pure translation (no rotation or scaling). C. It operates in homogeneous coordinates. D. It transforms coordinates for 2-dimensional locations. E. It performs a geometric (rigid body) transformation. Explanation: The matrix translates a 2D location in the form of a homogeneous vector (which is scaled to keep its last element equal to 1) by +2 in the x direction and +3 in the y direction. [ imaging0473.mcq ] 2. The following are true about segmentation in image analysis, except A. Automated segmentation suffers from a complete lack of any ability to incorporate prior information about the target anatomical structure. B. It involves the labeling of individual pixels (or voxels) according to some model of the expected anatomical target. C. One form of segmentation is deformable contours (snakes), in which an initial contour is placed on the image, and then the contour is changed to better fit an underlying anatomical structure in the image. D. Manual segmentation suffers from inter-observer variation, in which different people will perform different seg- mentations on a given image. E. Manual segmentation suffers from intra-observer variation, in which the same person will perform different seg- mentations at different times on a given image. Explanation: Answer A is false: Prior information can be incorporated into automated segmentation in a number of ways, concerning expected intensity (e.g. thresholds) or shape (e.g. initial shape and placement of snakes, or principle component analysis of examples of the expected target). [ imaging0503.mcq ] 1 3. The following are true about registration in image analysis, except A. It may be performed on 2D, but not 3D, image data. B. It involves a transform to remap one image onto the coordinate system of another image in such a way that a given anatomical location corresponds to the same coordinates in both images. C. It requires a metric to determine the fitness value of the registration, which is optimized to determine the best transform to remap one image onto the coordinate system of the other. D. Examples of registration transforms include, geometric (rigid body), similarity, affine, projection, and deformation (warping). E. Similarity metrics for registration include root mean square (RMS), sum absolute different (SAD), covariance, correlation, and mutual information. Explanation: Registration may be performed on 2D or 3D images pairs, or even between a 2D image and a 3D image. [ imaging0504.mcq ] 4. The following are true about optical microscopy, except A. The resolution of a light microscope is limited by the particle nature of photons, which models electromagnetic radiation as individual “bullets.” B. 3D microscopic image data may be obtained using confocal microscopy or optical coherence tomography (OCT). C. Confocal microscopy uses pin-holes in addition to lenses to sharpen the focal plane. D. The virtual image seen through a typical compound microscope appears to be located where there is no actual source of photons, that location being determined by tracing the rays that emerge from the eyepiece lens. E. If a screen is placed in the plane of a real image in a compound microscope, the image will be visible on the screen. Explanation: The resolution of a light microscope is limited by the wave nature of photons, which models the microscope using field patterns, standing waves of constructive and destructive interference. [ imaging0505.mcq ] 5. The following are true about mathematical morphology, except A. It operates only on binary images, consisting of 1’s and 0’s. B. It is based on two fundamental operations, erosion and dilation. C. It includes two secondary operations, opening and closing. D. Its operations depend upon the definiton of a structuring element, whose shape can vary depending on the desired effect. E. It is particularly useful for eliminating small objects and filling in holes in objects. Explanation: It can operate either on binary or grayscale images. [ imaging0524.mcq ] 2 6. The following are true about quaternions, except (or all are true) A. All are true. B. They are the 3D equivalent of phasors in 2D. C. They consist of one real and 3 mutually orthogonal imaginary axes. D. They are used to represent orientation in 3D. E. They can be multiplied by each other, effecting rotation from any of the x, y, or z axes to any of the other axes. Explanation: All are true. [ imaging0525.mcq ] 7. The following are true about lenses in optical imaging, except A. Although a prism is capable of breaking white light into its constituent colors through the phenomenon of dispersion, lenses do not exhibit this phenomenon, and focus all colors identically. B. They create standing waves according to Huygen’s principle. C. They slow down the speed of light along a particular ray, introducing a delay proportional to the thickness of the glass along that particular ray. D. When looking through a convex lens (such as a magnifying glass) at an object, a virtual image may be observed, from which photons appear to be emanating. E. The resolution of what can be seen through a lens is limited by the wave nature of light. Explanation: Lenses do exhibit dispersion, which is one the optical aberrations that all lenses suffer from. [ imaging0526.mcq ] 8. The following are true about depth of focus (range resolution) in conventional optical microscopy, except (or all are true) A. As in ultrasound, range resolution in conventional optical microscopy is determined by time of flight and is therefore independent on the field pattern. B. Depth of focus can be reduced (range resolution improved) by using confocal microscopy, which incorporates pinhole apertures in addition to lenses. C. All are true. D. In the focal region, resolution in depth is generally less than resolution within the image plane (transverse to the optical axis). E. Optical Coherence Tomography has far better range resolution than optical microscopy. Explanation: Range resolution in conventional optical microscopy is determined by the field pattern, and it not determined by time of flight. [ imaging0533.mcq ] 3 9. The following are true about real and virtual images, except (or all are true) A. If a screen is placed in the plane of a virtual image, the image will become visible on the screen. B. A real image is located in the plane of convergence for the light rays that originate from a given object. C. All are true. D. A virtual image is an image that only appears to be at a particular location, determined by tracing the rays that emerge from an optical device. E. The virtual image of an object placed in front of a mirror is located behind the mirror. Explanation: Statement A is true for real images, not virtual images. [ imaging0534.mcq ] 10. The following are true about segmentation in medical imaging, except (or all are true) A. All are true. B. It is an attempt to label pixels (or voxels) according to the underlying physical structures. C. It can be performed in 2D or in 3D. D. If a 3D object is segmented using a set of 2D slices, the orientation of the slices can produce a situation where the object appears in two separated regions within a given slice. E. It is limited by the precision with which labels have been applied to anatomical structures. Explanation: All are true. [ imaging0535.mcq ] 11. Manual segmentation of medical images suffers from the following limitations, except A. Human vision does not have sufficient dynamic range to cover the dynamic range of a typical display on which medical images are shown. B. Inter-observer variation can be significant. C. It is generally performed in 2D rather than in 3D. D. Intra-observer variation can be significant. E. It can be time-consuming, especially when done on multiple slices in 3D data sets. Explanation: The dynamic range of human vision is far greater than that of a typical display. [ imaging0536.mcq ] 12. The following are true about registration in medical imaging, except (or all are true) A. It can only be done properly if the two images are of the same imaging modality (e.g., CT, MR, etc.). B. All are true. C. It consists of properly superimposing the coordinate systems of two images such that the underlying anatomy corresponds by pixel location. D. It uses global transforms (such as geometric, similarity, affine), as well as possibly deformation to move one of the images into the other images physical coordinate system, followed by resampling onto the other image’s lattice. E. It requires a metric of the accuracy of the registration to optimize. Explanation: Registration of two images from different imaging modalities is possible, especially using a metric such as mutual information. [ imaging0537.mcq ] 4 13. Which of the following 3D transforms can produce non-parallel lines from parallel lines (or none can)? A. Projection. B. Geometric. C. Similarity. D. Affine. E. None of the others. Explanation: Only the Projection Transform can take the parallel railroad tracks and have them meet at a point at infinite distance, by shrinking the x-y coordinates as z gets bigger. [ imaging0538.mcq ] 14. The following are true about the spatial gradient of image intensity in medical imaging, except (or all are true) A. All are true. B. At the boundary between one homogeneous region and another with differing intensity, it is a vector approximately normal to the boundary (approximately because of spatial sampling in the image). C. It can exist in 2D and 3D images D. The gradient magnitude is a scalar measure of boundary strength, independent of boundary orientation. E. It is frequently computed using convolution of the image with a set of edge detection kernels, one for each dimension in the image yielding that component of the gradient. Explanation: All are true. [ imaging0539.mcq ] 15. The following are true about rendering opaque surfaces in medical imaging, except (or all are true) A. All are true. B. It is often accomplished by tessellation, in which individual small planar elements (usually triangles) are rendered using a lighting model based on orientation and location. C. A Z-buffer may be used to guarantee that opaque objects properly occlude either other based on distance from the viewer, independent of the rendering order. D. Marching cubes is a commonly used algorithm to quickly build tessellated iso-surfaces between two intensity levels. E. Texture mapping is a method of interpolating 2D or 3D image data onto a tessellated surface to be rendered graphically in 3D space. Explanation: All are true. [ imaging0540.mcq ] 5 16. The transform matrix below performs which operations? A. Scaling and Translation. B. Rotation and Translation C. Scaling and Rotation D. Projection and Translation E. Projection and Scaling Explanation: The upper left 3x3 elements with only the diagonal non-zero does anisotropic scaling. The middle of the top 3 elements in the right-hand row does translation in y. [ imaging0541.mcq ] 17. The following are true about convolving and image with a Gaussian, except A. Convolving an image with a Gaussian can be done in 2D but not in 3D. B. It removes noise within otherwise homogeneous objects. C. It blurs edges between objects with different intensities. D. Convolving an image with the derivative of a Gaussian finds edges and removes noise. E. All are true. Explanation: One can convolve a 3D Gaussian with a 3D image, blurring in 3D. [ imaging0542.mcq ] 18. The following are true about machine learning, except A. It cannot be used to analyze images, but is only effective for problems based on words. B. It may function in ways not understood by humans. C. It requires training data from which patterns are extracted. D. It is a form of connectionist rather than symbolic artificial intelligence. E. It is often based on layers of neural networks, with the number of layers becoming large in the case of deep learning. Explanation: Machine learning has proven effecting in analyzing images, often with the use of convolutional neural networks, as well as with problems based on words. [ imaging0543.mcq ] 6 19. The following are examples of Point Operators on intensity, that is, each pixel in the output image is a function of only the equivalent pixel in the input image, except (or all are examples). A. Convolution with an edge detection kernel to find boundaries. B. Thresholding the intensity of an image to identify pixels above a given threshold. C. Adjusting the brightness and contrast of an image. D. Mapping the intensity values in an image to a new set of values such that the histogram of intensity is uniform (histogram equalization). E. All are examples. Explanation: Convolution is inherently not a Point Operator, but rather takes the region around each pixel into account. [ imaging0544.mcq ] 7