| WIRAATMAJA CHRISTOPHER | D, 中間発表 | 大規模システム管理 | 笠原 正治, 松本 健一, 笹部 昌弘（客員教授） |
|
title: Cost-Efficient Anonymous Authentication Scheme based on Set-Membership Zero-Knowledge Proof
abstract: In this paper, we propose zero-knowledge named proof, a replay attack prevention scheme that ensures the user's anonymity against malicious administrators. We begin with adopting the zero-knowledge set-membership proof into an authentication setting in which users would delegate their requests to an agent that obstructs the user's identity from the administrator. This anonymous agent carries the guarantee of authenticity, which the administrator through the set-membership proof can confirm. Next, we prevent replay attacks from other parties by binding the agent's identity to the delegation request verifiable by the administrators. By leveraging these properties, a blockchain-based authentication scheme is then built. We quantitatively evaluate the security, cost-efficiency, and performance of our scheme and provide a third-party authorization scheme from our authentication framework to demonstrate its real-world relevancy. language of the presentation: English | |||
| 山田 浩太 | M, 2回目発表 | 大規模システム管理 | 笠原 正治, 松本 健一, 笹部 昌弘（客員教授）, 原 崇徳 |
|
title: On a Repeated Stochastic Game for Offload Mining in Decentralized Applications *** abstract: With the proliferation of decentralized applications (DApps) based on the blockchain, a computationally intensive mining process is a challenging problem because most DApp users own only devices with the limited resources, e.g., mobile devices. Therefore, the DApp aims at operating securely by offloading the computational task required by the users for mining to the cloud and/or edge, i.e., cloud service provider (CSP). In this paper, we formulate the CSP selection problem for the offload mining in which each user (miner) selects which CSP from the available candidates as a repeated stochastic game and propose an offloading method so as to achieve a coarse correlated equilibrium (CCE) among the DApp users. Through the numerical experiment, we demonstrate the effectiveness of the proposed method. *** language of the presentation: *** Japanese *** | |||
| 玉井 駿哉 | M, 2回目発表 | 大規模システム管理 | 笠原 正治, 松本 健一, 笹部 昌弘（客員教授） |
|
title:
Analysis of Governance Mechanisms in DAO Adopting Quadratic Voting
abstract: With the spread of blockchain technology, decentralized organizational forms known as DAOs (Decentralized Autonomous Organizations) have been attracting attention. DAOs, which make decisions through member voting actions, require governance mechanisms to address issues unique to decentralized organizations that were not present in traditional centralized organizations. This study focuses on two major challenges in DAOs governance: the "whale problem," where power becomes concentrated among a few members, and the "collusion problem," where voting outcomes can be manipulated through fraudulent collusion. We examine the voting system called Quadratic Voting, which is expected to mitigate the "whale problem," and analyze its resistance to the "collusion problem." language of the presentation: Japanese | |||
| 南 椋斗 | M, 2回目発表 | 大規模システム管理 | 笠原 正治, 藤川 和利, 笹部 昌弘（客員教授）, 原 崇徳 |
|
title: Modeling and Forecasting Service Demand in NFV Networks
abstract: In NFV networks, user-desired services can be modeled as service chains, which consist of one or more network functions connected sequentially. These service chains are implemented as service paths within the physical network infrastructure. Service paths extend the traditional end-to-end routes, executing the desired functions sequentially as software components on physical servers along the path from the source to the destination. To efficiently fulfill user requests, it is essential to pre-deploy the required software on physical servers, enabling swift service delivery. This pre-deployment can be likened to function caching. However, given the finite nature of server resources, the accuracy of service demand prediction significantly affects caching efficiency.In this presentation, we analyze publicly available data concerning the usage patterns of individual users and applications, gathered from a diverse range of mobile devices. Furthermore, we examine the results of application demand forecasting for a specific day. language of the presentation: Japanese | |||
| ZHANG WEIQI | M, 2回目発表 | 生体医用画像 | 佐藤 嘉伸, 加藤 博一, 大竹 義人, SOUFI Mazen |
|
title: Generative adversarial network for estimating lumbar spine bone mineral density from x-ray images via digitally reconstructed radiographs and hybrid training
abstract: Osteoporosis is a common disease characterized by low bone mineral density (BMD), leading to an increased risk of fractures. Dual-energy x-ray absorptiometry (DXA) and quantitative computed tomography (QCT) are widely used to measure BMD and osteoporosis diagnosis; however, they are not routinely performed due to higher cost or radiation exposure. The measurement of BMD by an x-ray image would allow an opportunistic screening for early diagnosis of osteoporosis. Previous methods that directly regress BMD from an x-ray image require a large training dataset and complicated preprocessing. Therefore, we proposed a method to estimate BMD using an x-ray image with a generative adversarial network(GAN), which we named BMD-GAN-Spine. Specifically, we concurrently synthesized two types of digitally reconstructed radiographs (DRRs) of four lumbar spines (L1, L2, L3, L4) from an x-ray image through multi-channel image synthesis to estimate the DXA-BMD and CT volume BMD (CT-vBMD). We evaluated our method using 37 patient datasets with pretraining with simulated x-ray images generated from a large-scale CT dataset (>9000 CTs). We achieved a Pearson correlation coefficient of 0.808, 0.753 between the predicted and ground truth DXA-measured BMD on both anteroposterior (AP) and lateral (LAT) views and 0.728, 0.885 between the predicted and ground truth QCT-measured CT-vBMD on AP and LAT, respectively. Our proposed method provides a new approach for estimating CT-vBMD from x-ray images with the capability for simultaneously improving the accuracy of DXA-BMD and CT-vBMD. language of the presentation: English | |||
| ZHAO BOHONG | M, 2回目発表 | 生体医用画像 | 佐藤 嘉伸, 加藤 博一, 大竹 義人, SOUFI Mazen |
|
title: Instance Segmentation of Distal Radius Fractures for Enhanced Surgical Planning and Intraoperative Guidance
abstract: Distal Radius Fractures (DRF) are among the most prevalent orthopedic injuries encountered in clinical practice. Effective management often necessitates surgical intervention, which relies heavily on accurate anatomical delineation of fractured bone fragments. To address this need, we developed a deep learning-based framework employing Mask R-CNN for instance segmentation of bone fragments in DRF. We evaluated our model on a dataset comprising CT scans from 50 patients, encompassing more than 19,000 image slices. Our model achieved an mean average precision (IoU=0.75) of 0.897, enabling the creation of a detailed 3D spatial representation of the fractured site. This segmentation streamlines surgical planning and offers intraoperative utility in 3D registration, thereby empowering surgeons to enhance operational control and potential patient outcomes. language of the presentation: English | |||
| 坂本 龍士郎 | M, 2回目発表 | 生体医用画像 | 佐藤 嘉伸, 金谷 重彦, 大竹 義人, SOUFI Mazen |
|
title: Construction of Age-related Statistical Models for Trunk Musculoskeletal Structures
abstract: It is important for surgical planning and biomechanical analysis to construct a model of the trunk musculoskeletal and to understand the deformation process of the trunk based on aging changes. In this study, we construct a statistical shape model (SSM) of the trunk musculoskeleton for each gender at each age using a large CT database. This study aims to construct a SSM that includes multiple musculoskeletal structures of the torso, whereas in the past SSM construction was limited to partial skeletal structures. We will evaluate the performance of the constructed statistical shape models for each age group, and investigate the age-related changes in musculoskeletal shape. language of the presentation: Japanese 発表題目: 体幹部筋骨格の年齢変化統計モデルの構築 発表概要: 体幹部筋骨格モデルを構築し,加齢変化に基づく，体幹部筋骨格の変形過程を理解することは,手術計画や生体 力学解析に重要である.本研究では,大規模な CT データベースを用いて,連続的な各年齢の男女別の体幹部筋骨格の 統計形状モデル(SSM)を構築する.J-MID データベースと共同研究施設で収集した 4 万症例以上のデータを活用し, 従来は部分的な骨格形状に限定されていた SSM 構築を,本研究では体幹部の複数の筋骨格構造を含めた SSM の構築を 行う.構築された各年齢の統計形状モデルの性能評価を行い,筋骨格形状の加齢変化について調査する. | |||
| 中谷 亮太 | M, 2回目発表 | 生体医用画像 | 佐藤 嘉伸, 金谷 重彦, 大竹 義人, SOUFI Mazen |
|
title: Automatic segmentation of 4DCT images for analysis of subject-specific swallowing motion
abstract: Accidents related to swallowing are not uncommon, and many diseases, such as strokes or neurological disorders, induce swallowing difficulties. Therefore, although it is important to understand the functional principles of swallowing, they have not been elucidated yet. Therefore, we aim to analyze the patient-specific swallowing motion by automatically recognizing the musculoskeletal structures involved in this motion from medical images. Previous studies analyzed the motion using 2D dynamic images and 3D static images of swallowing motion. This study is the first to analyze 4D (spatial 3D + time) CT images, which realistically reflect the swallowing function. 4DCT images were used to develop an automatic segmentation model that takes into account the temporal motion of the 3D shapes. The automatic segmentation of 7 structures, including a bolus, from the images using 3D nnUNet. The mean Dice coefficients were 0.550±0.304, 0.773±0.086, and 0.865±0.105 for the bolus, soft tissues, and bones, respectively. language of the presentation: Japanese 発表題目: 験者個別の嚥下動態解析を目的とした4DCTの自動セグメンテーション 発表概要: 嚥下に関連する事故は後を立たず，脳卒中や神経難病により嚥下障害を誘発する疾病も多い．そのため，嚥下の動作原理の解明が求められているものの， 現在においても解明されていない．そこで，我々は嚥下動作に関係する筋骨格を医用画像から自動認識することで，被験者個別の嚥下動態解析を目指している． 既存の研究では，嚥下における2次元動画像や3次元静止画像を用いたAI解析は行われているが，4次元CT画像を利用したAI解析は行われていない． 本研究では，4次元CTにより撮影された動画像を用いることで，3次元における形状及び時間軸を考慮した自動セグメンテーションモデルを作成する． 3D nnUNetを用いて，CT画像から食塊を含む7つの構造について自動セグメンテーションを行った結果，平均Dice係数は，食塊，軟組織，骨において それぞれ0.550±0.304，0.773±0.086，0.865±0.105であった． | |||