- Статьи
- Science and technology
- Calculations have been made: dangerous lung diseases have been distinguished for the first time without a biopsy
Calculations have been made: dangerous lung diseases have been distinguished for the first time without a biopsy
Scientists have developed a way to distinguish between tuberculosis and sarcoidosis, diseases that are often referred to as "twin diseases." On X-rays and microscopic examination, their granulomas practically do not differ, so the probability of a diagnostic error remains high. According to experts, doctors have to deal with uncertainty in 40-60% of cases. Researchers have patented a method that makes it possible to make an accurate diagnosis based on a routine blood test. The development is based on a mathematical model that reveals subtle differences in the ratio of immune cells. Due to this, instead of an invasive biopsy, it is enough to examine a blood sample: the algorithm determines with about 90% accuracy which of the two diseases develops in a particular patient. At the same time, experts interviewed by Izvestia emphasize that the new technology is an auxiliary tool. In complex diagnostic cases, biopsy is still the gold standard for confirming the diagnosis.
How to distinguish tuberculosis from sarcoidosis
Specialists from St. Petersburg State University, together with colleagues from the Almazov National Research Medical Center, have developed and patented a new method for diagnosing lung diseases that allows for high accuracy in distinguishing tuberculosis from sarcoidosis by blood analysis without surgical intervention. The research was carried out within the framework of the megagrant program of the Ministry of Education and Science in the laboratory "Probabilistic Methods in Analysis" of St. Petersburg State University.
The differential diagnosis of tuberculosis and sarcoidosis is one of the most difficult tasks of pulmonology and phthisiology, the scientists said. Due to the similarity of clinical, radiological and morphological signs, the probability of diagnostic errors reaches 40-60%, which is critical: sarcoidosis is not contagious and requires a fundamentally different treatment regimen.
Traditional diagnostic methods either do not reliably distinguish between these diseases, or require obtaining biopsy material, a painful procedure associated with certain risks for the patient. The new method is based on a mathematical analysis of the concentration of certain cell types in peripheral blood and completely eliminates the need for invasive intervention. A routine venous blood test is sufficient for the study, which greatly simplifies the diagnosis and makes it less traumatic for the patient.
— We have proved a method in which the diagnosis is based not just on comparing indicators, but on the application of two mathematical inequalities obtained using symbolic regression and direct multicriteria optimization of sensitivity and specificity. The idea of the invention is as follows: the patient donates blood, and our program for the concentrations of natural regulatory T cells and the balance of memory B cells determines the most likely diagnosis with high accuracy," explained Per Jan Hokan Hedenmalm, head of the Probabilistic Methods in Analysis Laboratory at St. Petersburg State University.
To create a diagnostic model, the researchers used a combination of several machine learning methods, including the aforementioned machine learning methods. As a result, a system has been developed that not only gives the result "tuberculosis" or "sarcoidosis", but also identifies a group of patients for whom the mathematical model does not allow an unambiguous conclusion. Such cases are classified as "doubtful" and require additional examinations. According to the developers, this allows you to avoid false confidence in situations where the clinical picture is really ambiguous, and promptly refer the patient for an in-depth diagnosis.
Head of the Laboratory "Probabilistic Methods in Analysis" of St. Petersburg State University Per Jan Hokan Hedenmalm
Tests have shown the high effectiveness of the method: the sensitivity in detecting sarcoidosis was 90.5%, and in diagnosing tuberculosis — 88.5%. Another advantage of the development is its full automation. It is enough for the doctor to enter the results of the immunological examination into the program, after which an algorithm based on a patented mathematical model will automatically perform calculations and determine which risk group the patient belongs to.
Accurate diagnosis by blood test
We are talking about two diseases that look almost the same on X—ray examination and under a microscope: granulomas form in the lungs - small foci of inflammatory tissue, but the nature of these processes is fundamentally different, explained molecular biologist Arina Kholkina. Tuberculosis is an infectious disease that requires antibacterial therapy and isolation of the patient. Sarcoidosis, on the contrary, is a non-contagious autoimmune condition and is treated according to a completely different scheme, often with the use of hormonal drugs.
According to her, a diagnostic error can have serious consequences: with tuberculosis, it increases the risk of infection of others, and with sarcoidosis, it leads to ineffective and potentially harmful treatment.
— St. Petersburg State University scientists have found a way out: they have trained a model to distinguish these diseases by the composition of immune cells in a routine blood test from a vein. Moreover, the algorithm honestly reports when there is not enough data for a confident answer, and marks the case as "doubtful". This is insurance against false confidence where the clinical picture is really ambiguous. The method is fully automated and does not require a lung biopsy," she noted.
The differential diagnosis of tuberculosis and sarcoidosis really remains one of the most difficult problems in pulmonology and phthisiology, the expert emphasized. According to her, the erroneous prescription of anti-tuberculosis therapy to a patient with sarcoidosis not only does not benefit, but can also cause serious harm to health.
— But I would not say that mathematical analysis becomes more accurate than histological examination in an absolute sense — a biopsy remains the gold standard where the picture is really unclear. But the proposed method has another advantage: it is non—invasive, reproducible, and devoid of the subjectivity inherent in human interpretation of histological preparations," she said.
The use of mathematical modeling methods, machine learning, creation of diagnostic algorithms and analysis of medical images is one of the key global trends in healthcare, said Yulia Ermolaeva, an infectious disease specialist at the Institute of Medicine and Medical Technologies at NSU, Deputy director of the Novosibirsk Tuberculosis Research Institute of the Russian Ministry of Health.
— Common symptoms such as a lingering cough, weakness, fever, night sweats, as well as similar changes on X-rays make it difficult to make a diagnosis. Most patients with sarcoidosis receive TB therapy due to an initial misdiagnosis. Therefore, the use of mathematical analysis for screening will reduce the number of invasive procedures and detect these diseases in a timely manner," she said.
The trend towards the use of matanalysis and machine learning methods in diagnostics will only intensify in the future, said Marina Chumakova, a leading market expert at NTI Helsnet. At the same time, according to her, such technologies should still be considered as an auxiliary tool. They help the doctor to narrow the range of diagnostic hypotheses, improve the accuracy of assessing the patient's condition and reduce the number of invasive procedures, but they are not able to completely replace traditional diagnostic methods.
Переведено сервисом «Яндекс Переводчик»
