HASA: Hybrid architecture search with aggregation strategy for echinococcosis classification and ovary segmentation in ultrasound images

作者：

Highlights：

• A hybrid NAS framework that also utilizes expert knowledge from existing CNNs.

• Progressive growing to stabilize the search process.

• Cell re-aggregation to adapt to different tasks.

• Extensive comparisons in two large US image datasets demonstrate the superiority.

摘要

•A hybrid NAS framework that also utilizes expert knowledge from existing CNNs.•Progressive growing to stabilize the search process.•Cell re-aggregation to adapt to different tasks.•Extensive comparisons in two large US image datasets demonstrate the superiority.

论文关键词：Deep learning,Network architecture search,Ultrasound image,Classification,Segmentation

论文评审过程：Received 25 October 2021, Revised 11 January 2022, Accepted 11 April 2022, Available online 27 April 2022, Version of Record 7 May 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.117242