Mwajuma Abdalla Ngema went to the clinic in Dar es Salaam with a persistent cough and intense pain around her ribs. The mother of two was tested for tuberculosis (TB) but the result came back negative. "I was given some medication to manage the cough but the pain was unbearable," says Ngema, who went home feeling frustrated at the lack of clarity over her health.
A few days later, Ngema was called back to Mbagala Kizuiani clinic - she had subsequently tested positive. She was relieved to finally have a diagnosis: "I am currently on medication and on the road to recovery."
Robert Ndoto, 55, tells a similar story. Living with HIV, he is particularly susceptible to tuberculosis but his tests were also negative. A few days later he was told he did have TB. The diagnosis meant he could start medication for the bacterial infection, which he was able to access through Mkuta, a local health organisation that targets TB and HIV.
"I was in bad shape when I met Salma Matamika [a volunteer with Mkuta]. I had listened to her give a talk about TB. She became the link between me and TB treatment," he says. "If you have HIV and TB, you can easily die."
Unusually, Ngema and Ndoto's revised TB diagnoses came after specially trained giant rats detected the disease in samples of the patients' sputum, or mucus from the lungs and airways.
The screening programme, run by a charity called Apopo, is only used in Tanzania and Ethiopia. They are among 30 countries on the World Health Organization's list of countries with a high TB burden.
Samples assessed as negative using smears or molecular tests in clinics are sent to Apopo's laboratory for secondary screening by the rats. Research this year shows that 52% of initially negative tests are reassessed as positive by the rats. Samples are always clinically tested to confirm the rats' verdict - and research shows the rats are correct in at least 82% of cases compared with standard clinical tests.
Apopo began in 1995 as a Belgian project to use rats' acute sense of smell for mine detection. In 2007 Tanzania became the first country to use rats for screening TB; Apopo now works with 81 clinics across the African state.
Next year the organisation will open a second test laboratory in Moshi, in the northern Kilimanjaro region. It is also in talks with health officials in Sierra Leone and Angola about introducing the programme there.
At the Tanzania Veterinary Laboratory Agency in Dar es Salaam, about a dozen rats screen up to 400 sputum samples a week. Since it started, Apopo has analysed more than 900,000 samples and detected more than 30,000 TB cases that may have been missed, averting at least 300,000 potential infections.
"One untreated person can infect 10-15 people, multiply that by 24,000 people correctly treated, who had been missed through regular tests. These were not just samples, these were lives saved," says Dr Joseph Soka, a manager at Apopo's laboratory.
The rats are trained to detect the presence of Mycobacterium tuberculosis (MTB), the pathogen that causes the disease. "The highly developed olfactory receptor cells in these rats can detect the volatile organic compounds responsible for MTB," says Soka. "The rats can isolate the specific strain even at low levels of these compounds."
In the laboratory a rat will be placed in a glass enclosure with 10 holes, each one containing a sample. Trained to associate positive samples with food, the rats hover over the positives and move straight on from negatives. It typically takes about three seconds to sniff out a positive sample and less than two seconds for a negative sample. The process must be repeated by at least five rats before technicians record a positive result.
Using the rats saves time, as they can screen scores of samples within minutes, and is far cheaper than a clinical test. A rat costs 2,600 shillings (90p) for one sample, compared with Sh4,700-7,000 for smear microscopy and Sh42,000 for molecular tests.
Apopo uses African giant pouched rats - one of the world's largest rodent species, which can grow up to 90cm (including the tail) - and breeds them at its base in Morogoro, 120 miles (200km) west of Dar es Salaam.
"We train them from four weeks old," says Fidelis John, a coordinator with Apopo. It takes nine months to a year to train a rat to detect MTB. But before it can become part of the team, he says, "it must 'sit' for the [TB-detection] exam and score 100%."
Before training begins, the rats are socialised to create a bond between them and their human handlers. "There are trust issues at first," says John, as he holds Louis, a two-year-old trainee.
"This rat has to trust that I am not a threat, while I have to be sure the rat will not bite me." In captivity most rats will live for eight years, some for longer. At about seven years old, rats retire to the centre in Morogoro, where "he will just eat, sleep, waiting to die", John says.
Related: Light at the end of the tunnel: TB screening for Pakistan's miners comes out of the clinic and on to the coalface
A recently published paper shows that rats are particularly sensitive to low levels of MTB; in children, who are harder to diagnose, for example. It found that "children, especially those with a lower bacterial levels, were more likely to be diagnosed with TB by rats" and suggested "rat-based technology could be used to improve paediatric TB-case detection in high-burden countries such as Tanzania."
There have been cases where rats have detected TB but follow-up clinical tests have been negative; in some of these cases, patients later developed TB, suggesting that rats detect the disease before a clinical test can. Apopo is monitoring these patients to better understand rats' sensitivity to low levels of bacteria.
The 40% reduction in TB incidence in Tanzania since 2015 can be attributed in part to clinics' public screenings (one programme in October found 49 cases). Those testing negative for TB during screenings are told they could be contacted again after further sample analysis.
Dr Sultan Lusambi, a district TB and leprosy coordinator, says: "Those screened are not surprised that their samples turned positive after further analysis since we had prepared them for such an eventuality."
These outreach programmes largely depend on the efforts of volunteers trained to alert communities to the dangers of living with untreated TB.
Albertina Kisonjo, a volunteer with Mkuta who has had TB herself, describes how she assisted one family: "One of my neighbours had the symptoms. I brought her to a clinic, where she was confirmed positive for TB. Follow-up visits found four other TB cases among her family members.
"Had we lost this family, the community would have said it was due to witchcraft. We also cut off community infections by putting the family under treatment."
The decline in TB is also down to Apopo's work. Last year, its laboratory in Dar es Salaam identified more than 2,000 cases that otherwise would have gone undetected and helped avert at least 24,000 new infections.
"I am happy whenever I hear the sound of a motorcycle bringing in samples," says Soka. "I see lives that can be saved - lives that could be lost if conventional testing methods are not conclusive."