| Item type |
紀要論文 / Departmental Bulletin Paper(1) |
| 公開日 |
2020-11-24 |
| タイトル |
|
|
タイトル |
CNN を用いた自己進化型CAD システムの提案 |
|
言語 |
ja |
| タイトル |
|
|
タイトル |
Proposal for self-evolving CAD system using CNN |
|
言語 |
en |
| 言語 |
|
|
言語 |
jpn |
| キーワード |
|
|
言語 |
en |
|
主題Scheme |
Other |
|
主題 |
Breast Cancer |
| キーワード |
|
|
言語 |
en |
|
主題Scheme |
Other |
|
主題 |
Tumor Mass |
| キーワード |
|
|
言語 |
en |
|
主題Scheme |
Other |
|
主題 |
CNN |
| キーワード |
|
|
言語 |
en |
|
主題Scheme |
Other |
|
主題 |
Self-evolving CAD |
| キーワード |
|
|
言語 |
en |
|
主題Scheme |
Other |
|
主題 |
Relearning |
| 資源タイプ |
|
|
資源タイプ識別子 |
http://purl.org/coar/resource_type/c_6501 |
|
資源タイプ |
departmental bulletin paper |
| ID登録 |
|
|
ID登録 |
10.34411/00001163 |
|
ID登録タイプ |
JaLC |
| 著者 |
安倍, 和弥
武尾, 英哉
永井, 優一
縄野, 繁
|
| 抄録 |
|
|
内容記述タイプ |
Abstract |
|
内容記述 |
In recent years, convolutional neural networks (CNN) have found active application in the field of computer-aided diagnosis (CAD). Typically, general-purpose, high performance detectors are designed by conducting machine learning provided with comprehensive sets of case images having different kinds of variations. The reasoning behind this is that images not included in learning are not expected to be successfully detected. In fact, a comparison of CNN trained using only relatively typical cases and CNN trained using a comprehensive set of cases showed superior performance by the latter. Such is also expected to be the case in the commercial release of CAD. This is because the cases used in development may not encompass the cases at the medical facilities that will actually operate the system. In response to this situation, this paper proposes a self-evolving CAD system that incorporates a relearning function. It is not possible to upgrade the detectors of previous CAD systems following their commercial release. As a remedy, we propose a mechanism for upgrading the detector, accomplished by endowing the CAD system with a misrecognition correction function that uses interpretation report information (i.e., electronic medical records), and performing relearning after certain data is applied. This function provided improved performance compared to when relearning is not performed. Moreover, multiple relearning resulted in gradual improvements, demonstrating a process by which the system evolved into a CAD system matched to the involved medical facility. |
|
言語 |
en |
| 書誌情報 |
神奈川工科大学研究報告.B,理工学編
巻 43,
p. 1-5,
発行日 2019-03-01
|
| 出版者 |
|
|
出版者 |
神奈川工科大学 |
| ISSN |
|
|
収録物識別子タイプ |
PISSN |
|
収録物識別子 |
21882878 |
| 書誌レコードID |
|
|
収録物識別子タイプ |
NCID |
|
収録物識別子 |
AA12669200 |
| フォーマット |
|
|
内容記述タイプ |
Other |
|
内容記述 |
application/pdf |
| 著者版フラグ |
|
|
出版タイプ |
VoR |
|
出版タイプResource |
http://purl.org/coar/version/c_970fb48d4fbd8a85 |
kkb-043-001.pdf (4.4 MB)
