The Hindu Kush-Karakoram-Himalaya (HKH) region is highly susceptible to glacial lake outburst floods due to climate change, and the recent status and changes of glacial lakes in this region remain unclear due to limitations in data quality and traditional satellite mapping methods. In this study, we proposed a comprehensive analysis of glacial lake changes in the HKH between 2000 and 2022 using a deep-learning-based automated mapping framework. By employing a trained U-Net model and integrating cloud, shadow, snow, and topographic shadow masks, we generated high-quality glacial lake datasets with a minimum mapping unit of 0.02 km2. Compared with previous public glacial lake datasets, our dataset detects more glacial lakes, showing more comprehensive changes. Our findings reveal an increase of 325 glacial lakes and an expansion of 72.58 +/- 6.20 km2 in the lake area over the 22 years, with significant growth occurring in the eastern and central Himalayas. Ice-contact glacial lakes exhibited the most substantial area increase, driven by glacier retreat. The spatial heterogeneity in glacial lake distribution and changes reflects a response to increasing temperature, declining precipitation, and elevational gradients. Our study highlights the potential of deep learning techniques for region-scale glacial lake monitoring, providing valuable data for understanding climate impacts on high-mountain geomorphology and advancing glacial lake outburst flood risk assessments.