I was just curious, did anyone try utilizing the multiple images available per patient? Even a simple ensemble can lead to a large score difference.