Pixel-Wise Modulated Dice Loss for Medical Image Segmentation

Class imbalance and the difficulty imbalance are the two types of data imbalance that affect the performance of neural networks in medical segmentation tasks. In class imbalance the loss is dominated by the majority classes and in difficulty imbalance the loss is dominated by easy to classify pixels. This leads to an ineffective training. Dice loss, which is based on a geometrical metric, is very effective in addressing the class imbalance compared to the cross entropy (CE) loss, which is adopted directly from classification tasks. To address the difficulty imbalance, the common approach is employing a re-weighted CE loss or a modified Dice loss. In this study we propose a simple modification to the Dice loss which is more effective in focusing the loss on difficult to classify areas compared to existing methods. With a pixel-level modulating term the impact of difficult pixels on the overall loss is increased. The modulating term is changed dynamically based on the predictions of the network, keeping the loss focused on difficult areas during training. Results of three commonly used medical segmentation tasks show that the proposed Pixel-wise Modulated Dice loss (PM Dice loss) improves the performance of the Dice loss and outperforms other more complicated methods, which are designed to tackle the difficulty imbalance problem.