We explore the potential of deep convolutional neural network (CNN) models for differential diagnosis of gout from musculoskeletal ultrasound (MSKUS). Our exhaustive study of state-of-the-art (SOTA) CNN image classification models for this problem reveals that they often fail to learn the gouty MSKUS features, including the double contour sign, tophus, and snowstorm, which are essential for sonographers' decisions. In this study, we establish a human-centered adjusting framework to make CNN models diagnosis gout in the way of thinking like sonographers. This framework consists of three components: (1) Where to adjust: Modeling sonographers' attention map to emphasize the region that needs adjust
(2) What to adjust: Identifying MSKUS instances that mislead the model to predict according to unreasonable/biased features
(3) How to adjust: Developing a human-centered training fine-tuning mechanism that adjusts CNN models to focus on the desired MSKUS feature regions. This mechanism balances the gout prediction accuracy and attention reasonability by introducing the sonographers-like attention constraints. The experimental results on clinical MSKUS datasets demonstrate that our framework significantly outperforms the existing SOTA CNN models.