A Novel Multiple-Instance Learning-Based Approach to Computer-Aided Detection of Tuberculosis on Chest X-Rays

A Novel Multiple-Instance Learning-Based Approach to Computer-Aided Detection of Tuberculosis on Chest X-Rays

Abstract—To reach performance levels comparable to human experts, computeraided detection < CAD > systems are typically optimized following a supervised learning approach that relies onlarge training databases comprising manually annotated lesions. However, manuall y outlining those lesions constitutes a difficult and time-consuming process that renders detailedly annotated data difficult to obtain. We investigate an alternative approach, namely multiple-instance learning (MIL), that does not require detailed information for optimization. We have applied MIL to a CAD system for tuberculosis detection. Only the case condition (normal or abnormal) was required during training. Based upon the well-known miSVM technique, we propose an improved algorithm that overcomes miSVM’s drawbacks related to positive instance underestimation and costly iteration.