Flowchart of medical image analysis concepts from RIMI Lab. At the top level, 'Evidential Inference' branches into three components: 'Evidential Learning,' 'Automated Multi-objective Learning (AutoMO),' and 'Reliable Rule Base.' The second level titled 'Radiomics/Delta Radiomics/Radiogenomics/Multi-Omics' leads to four outcomes: 'Treatment Outcome Prediction,' 'Clinical Diagnostic Support,' 'Medical Image Processing,' and 'Quantitative Imaging Biomarker Discovery.' The flow indicates a progression from general evidence-based frameworks to specific applications in medical imaging and diagnostics.

Evidential Inference
Evidential Learning Automated Multi-objective Learning (AutoMO) Reliable Rule Base
Radiomics/Delta Radiomics/Radiogenomics/Multi-Omics
Treatment Outcome Prediction Clinical Diagnostic Support Medical Image Processing Quantitative Imaging Biomarker Discovery

The Reliable Intelligence and Medical Innovation Laboratory (RIMI Lab) is devoted to developing reliable artificial intelligence (RAI) theory to achieve balance, credibility, adaption as well as interpretation, and developing RAI based models or methods for clinical problems, particular in cancer.

The fundamental theory is evidential inference, a general probabilistic inference engine. Three major methods are developed, they are: evidential learning, automated multi-objective learning and reliable rule base. Based on these methods, radiomics, delta radiomics, radiogenomics, and multi-omics are investigated.

Several clinical research are also conducted, including treatment outcome prediction, clinical diagnostic support, medical image processing, quantitative imaging biomarker discovery, etc. In cancer research, we are working on treatment follow-up prediction, immunotherapy response prediction, neoadjuvant therapy response prediction, malignancy prediction, tumor segmentation, etc.