TB kills thousands in Africa

The sensors use artificial intelligence to identify bacteria in TB cases, as well as other respiratory diseases.

The technique analyses sputum - saliva and mucus - converted into gas form.

Infected sputum contains various acids and other products of infected lung tissue.

Professor Anthony Turner, one of the project leaders, said: "Each of these chemicals is capable of generating unique 'signature odours' when analysed under appropriate conditions."

Health rewards

If the method is developed further it could offer hope and support to patients and health workers worldwide, he said.

"The rewards both socially and financially are immeasurable."

Scientists have set up an experimental "nose"

Current laboratory methods of analysing sputum can only detect the infectious mycobacteria if it is present in large quantities.

It cannot distinguish between different strains of the bacteria.

To do that can take up to six weeks, with scientists having to grow cultures in a laboratory.

The scientists have proved the system works in laboratory tests.

Now they need funding for field trials and to manufacture equipment.

Growing toll

Clearance has been obtained to collect sputum samples from patients at Gloucestershire Royal Hospital in the UK and the State Centre of Tuberculosis in Latvia.

Research by the British Thoracic Society showed TB cases rose by nearly 20% between 1987 and 1998.

London presently has more cases of the disease than any other large European city, with more than 4,000 diagnoses a year.