Although it can appear in cities, Chagas disease is more common in rural areas and mainly affects vulnerable populations, such as riverside dwellers, rural workers, indigenous people and quilombolas. Today, a blood test indicates the presence of the parasite Trypanosoma cruzi. The method, however, requires a professional microscope coupled to a high resolution machine, as well as a specialist operator. This tends to be expensive and unaffordable for those most affected by the disease. Researchers at the University of São Paulo (USP) develop an algorithm that could facilitate this diagnosis. With it, blood sample images recorded by a simple smartphone will be enough to accuse the infection.
The system, called random forest, is trained to detect the presence of T. Cruzi in 12-megapixel images taken by cell phones and later analyzed under an optical microscope. “The algorithm will extract characteristics that represent the parasite in the photo, such as color, texture, brightness intensity, shape, curvature and size”, explains Mauro César Cafundó, postdoctoral fellow at the Computational Systems Biology Laboratory (CSBL), at USP, and lead author of the study, recently published in the journal PeerJ.
According to the Brazilian scientist, a statistical method was used to teach the algorithm to find patterns that identify the parasite. After being trained and exposed to a baseline, the system was subjected to experiments that evaluated how much it learned. “It is one thing to recognize the parasite, another is to teach the algorithm”, points out Cafundó. Therefore, artificial intelligence was developed with the collaboration of researchers in mathematics, statistics, parasitology and computing.
In tests, the algorithm showed promising performances in terms of accuracy (87.6%) and sensitivity (90.5%). Cafundó says that the perspective regarding the system is to be able to assess how far it operates: “For now, what we have is analytical evaluation. Now we need clinical validation, leaving the controlled environment of the laboratory”, he says. “This is the time to understand what needs to be adapted for this breakthrough.” The base code of the solution is available on the Internet and can be incorporated into the devices: “It is open to those who know how to program”, says Cafundó, reiterating that the system still needs improvement.
The team also intends to extend the software to detect other parasites, such as the one that causes leishmaniasis. “We achieved good results in this machine learning initiative. The algorithm works well for Chagas and can be adapted for other purposes that depend on images, such as the analysis of stool samples, skin and colposcopy”, indicates, in a note, Helder Nakaya, leader of the study, funded by the Fundação de Amparo à Pesquisa do Estado de São Paulo (Fapesp).
clinical impact
For the infectious disease doctor at the Center Specialized in Infectious Diseases Joana D’arc Gonçalves, solutions such as the one presented by the USP team help in dealing with a disease that is still very neglected in Brazil. “The various forms of presentation and the difficulties of early diagnosis make health professionals miss the time of adequate treatment”, she justifies.
In the expert’s assessment, this type of technological innovation can prevent the chronic evolution of the disease. “Early diagnosis of Chagas is essential to reduce sequelae and mortality associated with the disease. An infected person who is diagnosed in the acute phase and receives adequate treatment has approximately a 100% chance of cure”, she says.
*Intern under the supervision of Carmen Souza
Alert for new variants
A collaboration by researchers at the Karolinska Institute in Sweden found that Trypanosoma cruzi can form new variants as a result of combinations of different strains. These mutations are often more efficient at evading the immune system, which can make the parasite more dangerous and compromise the effectiveness of diagnostic tests.
By mapping the genome of parental strains of T. Cruzi and its descendants over time, the researchers obtained detailed images of how these hybrids are formed. The results show that these variants initially contain all the DNA of the parents, but that the amount of genetic material is gradually reduced until it forms an independent organism.
The study is based on parasite strains that spontaneously formed hybrids in the laboratory. DNA from the parental and offspring parasites were isolated and the genomes were mapped using large-scale DNA sequencing. Now, the team works towards an application of the results outside the laboratory. “We will study material from nature and from patients to map in more detail how the parasite varies its genes. We are also working to improve the diagnosis of Chagas disease in Bolivia,” says Björn Andersson, the study’s principal investigator, in a note.