| 眞田 将希 | M, 2回目発表 | 情報基盤システム学 | 藤川 和利, 安本 慶一, 新井 イスマイル |
|
title: Proposal of indoor positioning method using fingerprinting with acoustic features
abstract: Apps that acquire and use people's location information are becoming popular. An example of such an application is one to track the location of people in danger. In the case of the cleaning plant in this study, there is a demand to know the location of workers because of the high temperatures inside the plant. While there are well established outdoor positioning methods, there are few accurate indoor positioning methods. Therefore, it is necessary to propose a cost-effective positioning method for indoor use. So far, a positioning method using magnetism has been proposed for the clean-up area. However, it has been pointed out that it is difficult to use only magnetic positioning due to the vastness of the target space. There are also studies that take into account differences in atmospheric pressure, and these studies have improved the accuracy of positioning in different levels of space. However, there is still room for improvement in positioning accuracy. In this study, we show the effectiveness of sound features for positioning in a cleaning factory, and incorporate them as fingerprints to improve the accuracy of positioning. In order to use the fingerprints, it is desirable to have no fluctuation in the values. However, as a result of this verification, we found that the sound fluctuates with time. On the other hand, the acoustic characteristics themselves were unique to each room, and we believe that they can be used sufficiently for positioning. In the future, we aim to improve the positioning error by incorporating sound as a fingerprint in addition to magnetic and atmospheric pressure differences. language of the presentation: Japanese 発表題目: 音の特徴を加えたフィンガープリントによる屋内測位手法の提案 発表概要: 人の位置情報を取得し, 利用したアプリが普及している. その一例として, 危険な人の位置を把握するといったものがある.今回対象としている清掃工場でも, 屋内が高温になりやすいことから作業員の位置を把握したいといった需要がある. 屋外における測位手法は確立されている一方で, 屋内においては精度を有した手法があまり確立されていない. そのため清掃工場内において, コストのかからない測位手法を提案する必要がある. これまで清掃工場では磁気を使用した測位手法が提案されている. しかし, 対象の空間が広大であることから磁気のみでの測位は難しいことが挙げられている. また気圧差を考慮した研究も存在しており, こちらは階層の異なる空間において測位精度が向上している. しかしながら, 測位精度はまだまだ改善の余地がある. 本研究では清掃工場内での測位において, 音の特徴量の有効性を示した上で実際にフィンガープリントとして組み込み, 測位の精度向上に取り組む. フィンガープリントとして使用する上で, 値の変動はない方が望ましい. しかし今回の検証結果として, 音が時間により変動している部分が見られた. 一方で音響特性自体は部屋ごとに固有の値が得られたため, 十分測位に利用可能であると考えている. 今後は磁気や気圧差に加えて, 音もフィンガープリントとして組み込み, 測位誤差の改善を目指す. | |||
| 桝田 修慎 | M, 2回目発表 | 生体医用画像 | 佐藤 嘉伸, 池田 和司, 大竹 義人, SOUFI Mazen |
|
title:Automatic hip osteoarthritis grading with uncertainty estimation from computed tomography using digitally-reconstructed radiographs
abstract: Purpose:Progression of hip osteoarthritis (hip OA) leads to pain and disability, likely leading to surgical treatment such as hip arthroplasty at the terminal stage. The severity of hip OA is often classified using the Crowe and Kellgren-Lawrence (KL) classifications. However, as the classification is subjective, we aimed to develop an automated approach to classify the disease severity based on a combination of the two grades using digitally-reconstructed radiographs (DRRs) from CT images. Methods:Automatic grading of the hip OA severity was performed using deep learning-based models, including the recently developed VisionTransformer (ViT) model. The models were trained to learn a new grading scheme combining Crowe and KL grades, which represented the disease progression of hip OA. The models were trained in classification and regression settings. In addition, the model uncertainty was estimated and validated as a predictor of classification accuracy. The models were validated on a database of 197 hip OA patients. The model accuracy was evaluated using exact class accuracy (ECA) and one-neighbor class accuracy (ONCA). Results:ViT produced the highest accuracy of 0.656 ± 0.015 and 0.962±0.10 for ECA and ONCA, respectively, obtained in the regression setting. The standard error of regression (SE) was lowest in the ViT model. The model uncertainty was significantly larger in cases with large classification errors (P\textless0.01). Conclusion:In this study, an automatic approach for grading hip OA severity from CT images was developed. ViT model has shown the highest accuracy among the other models. Classification accuracy was correlated with the model uncertainty, which would allow for the prediction of classification errors. language of the presentation: Japanese | |||