京都大学 / 新制・論文博士 / 博士(医学) / 乙第13462号 / 論医博第2249号 / 新制||医||1055(附属図書館) / (主査)教授 萩原 正敏, 教授 松田 秀一, 教授 齊藤 博英 / 学位規則第4条第2項該当 / Doctor of Medical Science / Kyoto University / DFAM

Indiana University-Purdue University Indianapolis (IUPUI) / The Convolutional Neural Network (CNN) have shown a substantial improvement in the field of Machine Learning. But they do come with their own set of drawbacks. Capsule Networks have addressed the limitations of CNNs and have shown a great improvement by calculating the pose and transformation of the image. Deeper networks are more powerful than shallow networks but at the same time, more difficult to train. Residual Networks ease the training and have shown evidence that they can give good accuracy with considerable depth. Putting the best of Capsule Network and Residual Network together, we present Residual Capsule Network and 3-Level Residual Capsule Network, a framework that uses the best of Residual Networks and Capsule Networks. The conventional Convolutional layer in Capsule Network is replaced by skip connections like the Residual Networks to decrease the complexity of the Baseline Capsule Network and seven ensemble Capsule Network. We trained our models on MNIST and CIFAR-10 datasets and have seen a significant decrease in the number of parameters when compared to the Baseline models.

Background: Capsule endoscopy (CE) is a non-invasive imaging modality designed to evaluate various small bowel pathologies. Failure to reach the cecum within the battery lifespan, termed incomplete examination, may result in inadequate testing and possibly delayed therapy. Several studies have attempted to evaluate the association between CE completion and opioid use. However, their results are conflicting. The aim of this meta-analysis is to evaluate the previously published literature on the association between opioid use and CE completion. Methods: We performed a comprehensive literature search in PubMed, PubMed Central, Embase, and ScienceDirect databases from inception through June 1, 2018, to identify all studies that evaluated the association between CE completion and opioid use. We included studies that presented an odds ratio (OR) with a 95% confidence interval (CI) or presented the data sufficient to calculate the OR with a 95% CI. Statistical analysis was performed using the comprehensive meta-analysis (CMA), version 3 software. Results: Five studies with a total of 1,614 patients undergoing CE in the inpatient (IP) and outpatient (OP) setting were included in this study, 349 of which had an incomplete CE (21.6%). The pooled OR for CE completion is 0.50 (95% CI: 0.38-0.66, I2=36.9%) in opioid users compared to non-users. No publication bias was found using Egger's regression test. Conclusions: Our results indicate that patients on opioids are significantly less likely to have a complete CE examination compared to non-users. To our knowledge, this study represents the first meta-analysis to assess this association.

Recent reports suggest that measuring the objective quality is very essential towards the success of colonoscopy. Several quality indicators (i.e. metrics) proposed in recent studies are implemented in software systems that compute real-time quality scores for routine screening colonoscopy. Most quality metrics are derived based on various temporal events occurred during the colonoscopy procedure. The location of the phase boundary between the insertion and the withdrawal phases and the amount of circumferential inspection are two such important temporal events. These two temporal events can be determined by analyzing various camera motions of the colonoscope. This dissertation put forward a novel method to estimate X, Y and Z directional motions of the colonoscope using motion vector templates. Since abnormalities of a WCE or a colonoscopy video can be found in a small number of frames (around 5% out of total frames), it is very helpful if a computer system can decide whether a frame has any mucosal abnormalities. Also, the number of detected abnormal lesions during a procedure is used as a quality indicator. Majority of the existing abnormal detection methods focus on detecting only one type of abnormality or the overall accuracies are somewhat low if the method tries to detect multiple abnormalities. Most abnormalities in endoscopy images have unique textures which are clearly distinguishable from normal textures. In this dissertation a new method is proposed that achieves the objective of detecting multiple abnormalities with a higher accuracy using a multi-texture analysis technique. The multi-texture analysis method is designed by representing WCE and colonoscopy image textures as textons.

The development of atmospheric entry guidance methods is crucial to achieving the requirements for future missions to Mars; however, many missions implement a unique controller which are spacecraft specific. Here we look at the implementation of neural networks as a baseline controller that will work for a variety of different spacecraft. To accomplish this, a simulation is developed and validated with the Apollo controller. A feedforward neural network controller is then analyzed and compared to the Apollo case.

Our lab discovered that approximately 1 in 10,000 Escherichia coli cells in stationary phase remain in suspension after a high g-force centrifuge event. To establish the mechanism behind this curious phenotype, multiple mutant strains of E. coli were independently evolved such that the majority of their populations resisted migration when exposed to high centrifugal forces. Genomic DNA sequencing of the mutants' revealed unique, isolated mutations in genes involved in capsule synthesis and exopolysaccharide (EPS) production. Each mutant exhibits a novel mechanism that allows them to remain in suspension. The mutants were further characterized by determining their growth rates, strengths of resistance to various centrifugal forces, the phenotype's dependence on a carbon source, and timing of the phenotype's presentation. The results revealed: comparable mutant generation times to the wild-type strain, variable resistance to centrifugal force, phenotype dependence on carbon source, and phenotype presentation during early stationary phase. To interrogate the mechanism by which these cells stay in suspension the production of EPS was quantified, and gene knock-outs were performed. Quantification of the EPS revealed approximately a seventeen-fold increase in EPS in the mutants' compared to the wild-type strain. Gene knock-outs revealed the EPS produced can be attached to the outer-membrane or freely secreted into the media by different mechanisms. In addition, this mechanism was further confirmed to be responsible for the centrifuge resistant trait by attaching extracted EPS to polystyrene microspheres. Experimental results show that mutant extracted EPS treated beads caused increased bead retention in suspension compared to wild-type EPS treated beads. These results reveal that E. coli is using a novel mechanism to adapt to a new environmental factor introduced to remove the bacteria. With the discovery of this mechanism and the transferability to inorganic objects industrial applications are now envisioned where particle sedimentation is controllable and mixtures remain homogenized by attaching optically transparent biomolecules.

No description available.

No description available.

Indiana University-Purdue University Indianapolis (IUPUI) / During recent years, the field of computer vision has evolved rapidly. Convolutional Neural Networks (CNNs) have become the chosen default for implementing computer vision tasks. The popularity is based on how the CNNs have successfully performed the well-known computer vision tasks such as image annotation, instance segmentation, and others with promising outcomes. However, CNNs have their caveats and need further research to turn them into reliable machine learning algorithms. The disadvantages of CNNs become more evident as the approach to breaking down an input image becomes apparent. Convolutional neural networks group blobs of pixels to identify objects in a given image. Such a technique makes CNNs incapable of breaking down the input images into sub-parts, which could distinguish the orientation and transformation of objects and their parts. The functions in a CNN are competent at learning only the shift-invariant features of the object in an image. The discussed limitations provides researchers and developers a purpose for further enhancing an effective algorithm for computer vision. The opportunity to improve is explored by several distinct approaches, each tackling a unique set of issues in the convolutional neural network’s architecture. The Capsule Network (CapsNet) which brings an innovative approach to resolve issues pertaining to affine transformations by sharing transformation matrices between the different levels of capsules. While, the Residual Network (ResNet) introduced skip connections which allows deeper networks to be more powerful and solves vanishing gradient problem. The motivation of these fusion of these advantageous ideas of CapsNet and ResNet with Squeeze and Excite (SE) Block from Squeeze and Excite Network, this research work presents SE-Residual Capsule Network (SE-RCN), an efficient neural network model. The proposed model, replaces the traditional convolutional layer of CapsNet with skip connections and SE Block to lower the complexity of the CapsNet. The performance of the model is demonstrated on the well known datasets like MNIST and CIFAR-10 and a substantial reduction in the number of training parameters is observed in comparison to similar neural networks. The proposed SE-RCN produces 6.37 Million parameters with an accuracy of 99.71% on the MNIST dataset and on CIFAR-10 dataset it produces 10.55 Million parameters with 83.86% accuracy.

Surgical procedures conducted without proper guidance and dynamic feedback mechanism could lead to unintended consequences. In-vivo diagnostics and imaging (the Gastro-Intestinal tract) has shown to be inconvenient for the patients using traditional endoscopic instruments and often these conventional methods are limited in terms of their access to various organs (e.g. small intestines). Embedding sensors inside the living body is complex and further the communication with the implanted sensors is challenging using the current RF technology. Additionally, continuous replacement and/or batteries recharging for wireless sensors networks both in-vivo and ex-vivo adds towards the complexity. Advances in diagnostics and prognostics techniques require development at multiple levels through systems approach, guided by the futuristic intelligent decision making environment that reduces the human interference. The demands are not only at the component level, but also at the connectivity of the components such that secure, sustainable, self-reliant, and intelligent environment can be realized. This thesis provides important breakthroughs required to achieve the vision of intelligent healthcare environment. The research contributions of this thesis provide foundation for developing a new architecture for continuous medical diagnostic and monitoring. The chapters in this thesis cover four fundamental technologies covering the in-vivo imaging, ex-vivo imaging, energy for sensors, and acoustic communication. These technologies are: locomotion mechanism for wireless capsule endoscope (WCE), multifunctional image guided surgical (MIGS) platform, shape memory alloy (SMA) thermal energy harvester and thermo-acoustic sonar using carbon nanotube (CNT) sheets. First, two types of locomotion mechanisms were developed, the first one inspired by millipede legged type mechanism and the second one based on the traveling waves that were induced onto the walls of the WCEs through vibration. Both mechanisms utilize piezoelectric actuators and couple their dynamics and actuation capability in order to achieve propulsion. This controlled locomotion will provide WCE advantage in terms of conducting localized diagnostics. Next, in order to conduct ex-vivo surgical procedures using the OCT such as removing the unwanted tissue and tumors short distance beneath the skin, MIGS platform was developed. The MIGS platform is composed of three key elements: optical coherence tomography (OCT) probe, laser scalpel and high precision miniature scanning and positioning stage. The focus in this dissertation was on design and development of the programmable scanning and positioning stage. The combination of in-vivo tool such as WCE and ex-vivo tool such as MIGS will provide opportunity to conduct many non-invasive procedures which will save time and cost. In order to power the feedback sensors that assist in remote operation of surgical procedures and automation of the diagnostic algorithms, an energy harvester technology based on the SMA thermal engine was designed, fabricated, and characterized. A mechano-thermal model for the overall SMA engine was developed and experimentally validated. Finally, the thermo-acoustic sound generation mechanism using CNT sheets was investigated with the goal of developing techniques for acoustic localization of WCE and customized sound generation devices. CNT thermo-acoustic projectors were modeled and experimentally characterized to quantify the dynamics of the system under varying drive conditions. The overall vision of this thesis is to lay down the foundation for intelligent healthcare environment that provides the ability to conduct automated diagnostics, prognostics, and non-invasive surgical procedures. In accomplishing this vision, the thesis has addressed several key fundamental aspects of various technologies that will be required for implementing the automation algorithms. / Ph. D.