The cytology of the cerebrospinal fluid is traditionally performed by an operator (physician, biologist) by means of a conventional light microscope. The operator visually counts the leukocytes (white blood cells) present in a sample of cerebrospinal fluid (10 μl). It is a tedious job and the result is operator-dependent. Here in order to circumvent the limitations of manual counting, we approach the question of numeration of erythrocytes and leukocytes for the cytological diagnosis of meningitis by means of lens-free microscopy. In a first step, a prospective counts of leukocytes was performed by five different operators using conventional optical microscopy. The visual counting yielded an overall 16.7% misclassification of 72 cerebrospinal fluid specimens in meningitis/non-meningitis categories using a 10 leukocyte/μL cut-off. In a second step, the lens-free microscopy algorithm was adapted step-by-step for counting cerebrospinal fluid cells and discriminating leukocytes from erythrocytes. The optimization of the automatic lens-free counting was based on the prospective analysis of 215 cerebrospinal fluid specimens. The optimized algorithm yielded a 100% sensitivity and a 86% specificity compared to confirmed diagnostics. In a third step, a blind lens-free microscopic analysis of 116 cerebrospinal fluid specimens, including six cases of microbiology confirmed infectious meningitis, yielded a 100% sensitivity and a 79% specificity. Adapted lens-free microscopy is thus emerging as an operator-independent technique for the rapid numeration of leukocytes and erythrocytes in cerebrospinal fluid. In particular, this technique is well suited to the rapid diagnosis of meningitis at point-of-care laboratories.