An efficient deep learning method to detect lesions on 3D medical images via information of anatomical landmarks
Project/Area Number |
17K17680
|
Research Category |
Grant-in-Aid for Young Scientists (B)
|
Allocation Type | Multi-year Fund |
Research Field |
Perceptual information processing
Intelligent informatics
|
Research Institution | Kindai University |
Principal Investigator |
|
Project Period (FY) |
2017-04-01 – 2020-03-31
|
Project Status |
Completed (Fiscal Year 2019)
|
Budget Amount *help |
¥2,990,000 (Direct Cost: ¥2,300,000、Indirect Cost: ¥690,000)
Fiscal Year 2019: ¥910,000 (Direct Cost: ¥700,000、Indirect Cost: ¥210,000)
Fiscal Year 2018: ¥910,000 (Direct Cost: ¥700,000、Indirect Cost: ¥210,000)
Fiscal Year 2017: ¥1,170,000 (Direct Cost: ¥900,000、Indirect Cost: ¥270,000)
|
Keywords | 転移学習 / 深層学習 / コンピュータ検出支援 / 医用画像診断 / 解剖学的ランドマーク / 深層畳み込みニューラルネットワーク / プレトレーニングモデル / 畳込みニューラルネットワーク / Fine Tuning / 機械学習 / 医用画像認識 / 診断支援 |
Outline of Final Research Achievements |
The deep transfer learning is efficient deep learning to construct AI image recognition processes, such as a deep convolutional neural network. In the learning, a pre-training is performed with a different image dataset from a target object at first. The target pattern recognizer is constructed based on the pre-trained network. In this study, we evaluated the pre-training by medical image patches, including anatomical landmarks (LMs) for lesion pattern classification. The LM is a unique anatomical structure in the human body. Collecting the LM dataset is more accessible than the lesion data collection because those LMs exist not only in healthy cases but also in malignant cases. Our experiments with the clinical dataset showed that the pre-training with the LM dataset brought effective image features for the lesion classification.
|
Academic Significance and Societal Importance of the Research Achievements |
一般的に,最新のAI画像認識処理の構築には大量の学習用データが必要となる。一方で,医用画像上の病変パターンデータの大量収集は容易ではない。転移学習は,AI画像認識処理を効率的かつ少数データで構築するための2段階のAI学習法として知られている。 本研究は,医用画像上の病変パターン認識AIをより効率的に構築するための方法論に関するものである。この研究の発展により,多くの病変パターン認識AIの開発が促進され,さらには医療サービスの質向上につながるものと考える。
|
Report
(4 results)
Research Products
(7 results)