This paper presents a new approach of transfer learning-based medical imageclassification to mitigate insufficient labeled data problem in medical domain.Instead of direct transfer learning from source to small number of labeledtarget data, we propose a modality-bridge transfer learning which employs thebridge database in the same medical imaging acquisition modality as targetdatabase. By learning the projection function from source to bridge and frombridge to target, the domain difference between source (e.g., natural images)and target (e.g., X-ray images) can be mitigated. Experimental results showthat the proposed method can achieve a high classification performance even fora small number of labeled target medical images, compared to various transferlearning approaches.
展开▼
