Multiple Instance Learning with Auxiliary Task Weighting for Multiple Myeloma Classification

Whole body magnetic resonance imaging (WB-MRI) is the recommended modality for diagnosis of multiple myeloma (MM). WB-MRI used to detect sites disease across entire skeletal system, but it requires significant expertise and time-consuming report due great number images. To aid radiological reading, we propose an auxiliary task-based instance learning approach (ATMIL) MM classification with ability localize disease. This appealing as only patient-level annotations where attention mechanism identify local regions active We borrow ideas from multi-task define task adaptive reweighting support improve efficiency in presence data scarcity. validate our on both synthetic real multi-center clinical data. show that MIL module provides a bone while considerably improves performance.