Metagenomic next-generation sequencing (mNGS) has greatly improved our understanding of pathogens in infectious diseases such as pulmonary tuberculosis (PTB). However, high human DNA background (>
95%) impedes the detection sensitivity of mNGS in identifying intracellular Mycobacterium tuberculosis (MTB), posing a pressing challenge for MTB diagnosis. Therefore, there is an urgent need to improve MTB diagnosis performance in PTB patients. In this study, we optimized mNGS method for diagnosis of PTB. This led to the development of the host DNA depletion assisted mNGS (HDA-mNGS) technique, which we compared with conventional mNGS and the host DNA depletion-assisted Nanopore sequencing (HDA-Nanopore) in diagnostic performance. We collected 105 bronchoalveolar lavage fluid (BALF) samples from suspected PTB patients across three medical centers to assess the clinical performance of these methods. The results of our study showed that HDA-mNGS had the highest sensitivity (72.0%) and accuracy (74.5%) in PTB detection. This was significantly higher compared to mNGS (51.2%, 58.2%) and HDA-Nanopore (58.5%, 62.2%). Furthermore, HDA-mNGS provided an increased coverage of the MTB genome by up to 16-fold. Antibiotic resistance gene analysis indicated that HDA-mNGS could provide increased depth to the detection of Antimicrobial resistance (AMR) locus more effectively. These findings indicate that HDA-mNGS can significantly improve the clinical performance of PTB diagnosis for BALF samples, offering great potential in managing antibiotic resistance in PTB patients.