Induced Sputum Versus Bronchoscopy in Smear Negative Pulmonary Tuberculosis

This is a randomised study that compares different diagnostic approaches for diagnosing pulmonary tuberculosis in patients suspected of pulmonary tuberculosis in whom the three classic (non-induced) sputum samples didn't show tuberculous bacillus on direct examination. The investigators compare the sensibility of induced sputum technique with an endoscopic approach (CT-scan followed by BAL and fluoroscopy-guided transbronchial biopsies and eventually sputum collection immediately after the bronchoscopy). People in high risk population for tuberculosis undergoing screening by chest X-ray or symptomatic patients will be admitted to the hospital if their chest X-ray shows a suspicion of active tuberculosis. According good clinical practice: (non-induced) sputum samples will be taken at admission and every following morning. If direct examination and PCR of the first three classic sputum samples are negative: patients will be randomised in two groups with a different diagnostic approach (induced sputum versus endoscopic approach) The aim of our study is to proof that a thoroughgoing endoscopic approach has a higher sensibility than an induced sputum in the diagnosis of pulmonary tuberculosis in patients with a high suspicion of active tuberculosis on the chest X-ray but with a negative direct examination and/or PCR on three consecutive normal sputum samples. The investigators will include 154 patients (based on a statistical analysis for a hypothesis that the endoscopic approach has a sensibility that's twice the sensibility of the induced sputum). * first arm: 2 consecutive induced sputum using an ultrasonic nebulizer. * second arm: CT thorax to evaluate the exact anatomic localisation of the disease followed by fluoroscopy-guided bronchoscopy for BAL (bronchoalveolar lavage) and transbronchial biopsies. A sputum sample immediately after the endoscopy will be collected if possible.