Boosting Medical Image-based Cancer Detection via Text-guided Supervision from Reports

arXiv:2405.14230v1 Announce Type: new
Abstract: The absence of adequately sufficient expert-level tumor annotations hinders the effectiveness of supervised learning based opportunistic cancer screening on medical imaging. Clinical reports (that are rich in descriptive textual details) can offer a "free lunch'' supervision information and provide tumor location as a type of weak label to cope with screening tasks, thus saving human labeling workloads, if properly leveraged. However, predicting cancer only using such weak labels can be very changeling since tumors are …

abstract annotations arxiv boosting cancer cancer detection cancer screening clinical cs.ai cs.cl cs.cv detection expert free image imaging information location medical medical imaging reports screening supervised learning supervision text textual type via