A Machine Learning Model to Predict the Histology of Retroperitoneal Lymph Node Dissection Specimens.

BACKGROUND/AIM: While post-chemotherapy retroperitoneal lymph node dissection (PC-RPLND) benefits patients with teratoma or viable germ cell tumors (GCT), it becomes overtreatment if necrosis is detected in PC-RPLND specimens. Serum microRNA-371a-3p correctly predicts residual viable GCT with 100% sensitivity; however, prediction of residual teratoma in PC-RPLND specimens using current modalities remains difficult. Therefore, we developed a machine learning model using CT imaging and clinical variables to predict the presence of residual teratoma in PC-RPLND specimens. PATIENTS AND METHODS: This study included 58 patients who underwent PC-RPLND between 2005 and 2019 at the University of Tsukuba Hospital. On CT imaging, 155 lymph nodes were identified as regions of interest (ROIs). The ResNet50 algorithm and/or Support Vector Machine (SVM) classification were applied and a nested, 3-fold cross-validation protocol was used to determine classifier accuracy. RESULTS: PC-RPLND specimen analysis revealed 35 patients with necrosis and 23 patients with residual teratoma, while histology of 155 total ROIs showed necrosis in 84 ROIs and teratoma in 71 ROIs. The ResNet50 algorithm, using CT imaging, achieved a diagnostic accuracy of 80.0%, corresponding to a sensitivity of 67.3%, a specificity of 90.5%, and an AUC of 0.84, whereas SVM classification using clinical variables achieved a diagnostic accuracy of 74.8%, corresponding to a sensitivity of 59.0%, a specificity of 88.1%, and an AUC of 0.84. CONCLUSION: Our machine learning models reliably distinguish between necrosis and residual teratoma in clinical PC-RPLND specimens.

A new systematic collection and classification of odour words by using a product review dataset.

The odours encountered on a daily basis are dependent on an individual's society and culture. Therefore, when conducting olfactory tests, the odour stimuli utilized must be appropriate for the individual's environment. In this study, we gathered and classified the odours experienced by Japanese individuals in their daily lives through a large dataset of product reviews encompassing food and household items. Specifically, we performed morphological analysis on product review sentences in Japanese that contained odour descriptions, and we compiled the nouns used to describe odours. A total of 617,208 sentences that reviewed odour experiences and their corresponding nouns were collected. The top 100 frequently used odour nouns were classified into 15 clusters according to the context in which they were used. The methodology employed in collecting and classifying odour nouns as presented in this study can be utilized in other situations. It can assist in selecting appropriate odour stimuli for the olfactory test based on the society, culture, and time period.

Viewing zone expansion of a dual-viewer autostereoscopic display with inclined interleaved linear Fresnel lens arrays and a time-division quadruplexing directional backlight.

A novel autostereoscopic display with a directional backlight unit using a pair of inclined interleaved linear Fresnel lens arrays is designed and evaluated. Different high-resolution stereoscopic image pairs are provided to each of the two viewers simultaneously by applying time-division quadruplexing. The viewing zone is enlarged in the horizontal direction by inclining the lens array, which enables two viewers to have specific views corresponding to the positions of their own eyes without interfering with each other's view. Thus, two viewers with no special goggles on can share the same 3D world, which enables direct-manipulation style interaction and collaboration maintaining eye contact with each other.

Backlight system using an interleaved Fresnel lens array that attains a uniform luminance and two-dimensional directional light control.

A novel (to the best of our knowledge) Fresnel lens array is proposed to realize a uniform directional backlight with two-dimensional directionality. Autostereoscopic display with the proposed lens array improves image output quality and relieves the viewer's posture restriction without the need for any additional eye aid. In the proposed lens array, tiny prisms composing two adjacent linear Fresnel lenses are interleaved so that the two lenses may be virtually overlapped and work independently. The widths of the elemental prisms vary depending on the distance from the center of each lens. Thus, the light passing through the two lenses is mixed, which results in higher and more uniform luminance intensity. A prototype of an autostereoscopic display based on the time-multiplexed directional backlight method and made with the proposed lens array attained uniform luminance as well as low cross-talk between left-eye and right-eye images.

Autostereoscopic display for multiviewers positioned at different distances using time-multiplexed layered directional backlight.

In this paper, an autostereoscopic display system based on a time-multiplexed directional backlight using a large aperture Fresnel lens is proposed. High-resolution stereoscopy for multiple viewers positioned at different distances from the screen is achieved in the proposed system by layering polymer dispersed liquid crystal screens behind the Fresnel lens. The screens with segmented electrodes are electrically controlled to change the position of light diffusion, while the time-multiplexed backlight is projected by a digital mirror device projector at a high refresh rate. The light is diffused at the conjugate focal points of the observers' eyes to deliver directional light to each eye. The right-eye image and the left-eye image are alternated on the LCD panel in front of the lens to synchronize with the backlight.

Time-division and color multiplexing light-field display using liquid-crystal display panels to induce focal accommodation.

A super-multiview light-field display with horizontal and vertical parallax is realized by time-division and color multiplexing to deliver full-color images to each viewpoint. In the conventional study, an image of a different color is delivered to each viewpoint to induce focal accommodation. In the proposed method, we deliver images of different colors sequentially to generate a full-color image by an after-image effect. Though the number of time-divisions increases in the proposed method, perceived flicker is suppressed by showing different colors at different timings. We compare the observed images given by the proposed method with those given by the conventional method to find out that the former reproduces a natural blur effect when the image is defocused. We also confirm with a psychophysical experiment using a refractometer that the proposed method induces a stronger focal accommodation than other super-multiview methods with a smaller number of time-divisions or with a stronger flicker. The proposed optics is applicable to a near-eye display with a natural focal effect.

Reduction of image discontinuity in coarse integral volumetric imaging.

The authors describe two methods to reduce the discontinuity of the image in coarse integral volumetric imaging. The discontinuity of elemental images from neighboring elemental lenses is smoothed by introducing Fresnel elemental lenses with interleaved grooves. The discontinuity of volumetric layers with long intervals is made indistinct by introducing a smoothing filter and an edge filter to connect the layered images smoothly. The effectiveness of the proposed methods is confirmed by the prototype systems.

Realization of an aerial 3D image that occludes the background scenery.

In this paper we describe an aerial 3D image that occludes far background scenery based on coarse integral volumetric imaging (CIVI) technology. There have been many volumetric display devices that present floating 3D images, most of which have not reproduced the visual occlusion. CIVI is a kind of multilayered integral imaging and realizes an aerial volumetric image with visual occlusion by combining multiview and volumetric display technologies. The conventional CIVI, however, cannot show a deep space, for the number of layered panels is limited because of the low transmittance of each panel. To overcome this problem, we propose a novel optical design to attain an aerial 3D image that occludes far background scenery. In the proposed system, a translucent display panel with 120 Hz refresh rate is located between the CIVI system and the aerial 3D image. The system modulates between the aerial image mode and the background image mode. In the aerial image mode, the elemental images are shown on the CIVI display and the inserted translucent display is uniformly translucent. In the background image mode, the black shadows of the elemental images in a white background are shown on the CIVI display and the background scenery is displayed on the inserted translucent panel. By alternation of these two modes at 120 Hz, an aerial 3D image that visually occludes the far background scenery is perceived by the viewer.

Reduction of moiré for coarse integral volumetric imaging.

This paper proposes two moiré reduction methods for coarse integral volumetric displays composed of multilayer display panels. The quality of the image presented by the conventional multilayer integral imaging stays relatively low because of the moiré pattern, which appears due to the layering structure of display panels with limited pixel aperture. The authors propose two methods to solve this problem. The first method is to add another layer of convex lens array with short focal length on the front side of the layered display panels. The second method is to enlarge the pixel pitch of the back-layer panel while fixing that of the front-layer panel.

Energy-efficient integral imaging with suppression of pseudo images.

The authors present an energy-efficient integral imaging system with suppression of pseudo images, which are caused by a leakage of rays from each elemental image to the adjacent elemental lenses. The leakage of rays can be decreased when the lens is paired with a segmented backlight with limited aperture, which consumes less electricity. To remove the leakage of rays completely without damaging the image to be presented, three modifications are implemented. First, a lens with a larger focal distance is used to decrease aberration. Second, a lens array composed of thick elemental lenses is used so that the leakage of rays can be blocked by the rough surface on the side of the thick lens. Third, the aperture of the segmented backlight is expanded to avoid loss of light in the right image.

Integral volumetric imaging using decentered elemental lenses.

This paper proposes a high resolution integral imaging system using a lens array composed of non-uniform decentered elemental lenses. One of the problems of integral imaging is the trade-off relationship between the resolution and the number of views. When the number of views is small, motion parallax becomes strongly discrete to maintain the viewing angle. In order to overcome this trade-off, the proposed method uses the elemental lenses whose size is smaller than that of the elemental images. To keep the images generated by the elemental lenses at constant depth, the lens array is designed so that the optical centers of elemental lenses may be located in the centers of elemental images, not in the centers of elemental lenses. To compensate optical distortion, new image rendering algorithm is developed so that undistorted 3D image may be presented with a non-uniform lens array. The proposed design of lens array can be applied to integral volumetric imaging, where display panels are layered to show volumetric images in the scheme of integral imaging.

Integral volumetric imaging with dual layer fly-eye lenses.

This paper proposes an integral volumetric imaging system that uses a coarse fly-eye lens and a fine fly-eye lens to show smooth and deep 3D image. Conventional integral volumetric imaging displays using a coarse fly-eye lens have suffered from low image quality due to distinct seam of lenses and moiré pattern caused by layered panel structure. To solve these problems the proposed system uses a fine fly-eye lens whose elemental lens has a long focal distance. By placing a fine fly-eye lens near the layered real image, the seam and the moiré are removed while the degradation of the presented 3D image is kept small.

Realization of undistorted volumetric multiview image with multilayered integral imaging.

This paper presents a 3D display based on the coarse integral volumetric imaging (CIVI) technique. Though expression of focal effect and specular light is enabled by combining volumetric and multiview solutions, the image qualities of conventional systems have stayed low. In this paper high quality 3D image is attained with the CIVI technology, which compensates distortion and discontinuity of image based on the optical calculations. In addition, compact system design by layering color and monochrome panels is proposed.

A Geometrical Analysis of Associative Memory.

A geometrical method is proposed for analyzing the properties of an autocorrelation associative memory model. From the present geometrical viewpoint, the state transition of the model is expressed as dynamics on a sphere. The method shows that there is a critical memory ratio corresponding to the memory capacity at which the characteristic of the dynamics on the sphere changes and that the stored vectors are distributed around the stored band which is the intersection between the sphere and deformed sphere defined with the square of the inner states of neurons. The stored band seperates the sphere into the upper side, where initial state vectors in recalling processes are distributed, and the lower side, where spurious stored vectors are distributed. The method gives a geometrical picture of the dynamical behaviour in recalling processes: the state vector starts from the upper side and falls into the stored vector around the stored band or into some spurious stored vector on the lower side. Based on this picture, it is explained that Morita's partial reverse method of enhancing the memory capacity makes state vectors in the lower side unstable and stored vectors around the stored band stable.

Eigenspace Separation of Autocorrelation Memory Matrices for Capacity Expansion.

New autocorrelation memory matrices which realize eigenspace separation are proposed from the geometrical viewpoint over association dynamics. These memorization methods realize narrower distributions of eigenvalues and remove self-connections in a natural way. As a result, stable memory patterns and large basins of attraction are achieved without iterative learning. Modification of recalling dynamics for the new memory matrices is discussed based on the existing modified dynamics to attain even more memory capacity. Copyright 1997 Elsevier Science Ltd.