2023-05-09 06:00:36
What relationship is there between the study of satellites and kidney transplant rejection diagnoses? Data science. Daniel Yoo, busy as a datascientist doing machine learning on satellite positioning for the South Korean Air Force until 2018, is the first signatory ofa study published on May 4 in Nature Medicine on an automated computer assistant that corrects diagnoses of graft rejection after kidney transplantation.
This tool, tested on 4,409 biopsies from 3,054 transplant patients followed in twenty reference centers for transplantation in Europe and North America, showed that nearly 45% of rejection diagnoses made by doctors were incorrect. “Imperfect diagnoses with regard to the criteria of the international classification”prefers to say the professor of nephrology and epidemiology Alexandre Loupy (Inserm, Paris Cité University, AP-HP), whom Daniel Yoo joined within the team which worked for five years on this project.
The objective assigned by this doctor from the Necker Hospital to the experts in transplantation, nephrology, anatomopathology, data science, epidemiology and artificial intelligence brought together in an international consortium was to find a tool making it possible to secure rejection diagnoses. Since 1991, there has been the Banff international classification, named after the Canadian town where it was established at a conference of doctors. This reference makes it possible, from a biopsy of the kidney and other analysis results, to establish a diagnosis on the rejection of the graft. And to adapt the patient’s immunosuppressive (anti-rejection) treatment.
“Over the past twenty years, the progression of knowledge of the disease, the rejection of the graft, has made us discover that it [était] much more complex than we thought”, explains Alexandre Loupy. The causes of rejection are varied, even if they are based on the general principle of an immune system focused on the elimination of the foreign body. As the classification improved by adding criteria to refine the diagnosis, the tool became more complex. “However, the quality of diagnosis, the cornerstone of our profession, allows us to have the right treatment, at the right time, for the right person”adds the nephrologist.
Develop a companion tool
The first four years of work were devoted to feeding the algorithm with data and developing a companion tool. The fifth allowed it to be tested in a prospective study in which doctors did not know that their diagnosis would be compared to that of a machine. The rate of inaccurate diagnoses still surprised the team. “It’s all over the place, observes Alexandre Loupy, with scenarios of patients wrongly diagnosed as “rejection” who undergo heavy immunosuppressive treatments for nothing, with the consequences that we know on infectious diseases and cancers, or of patients diagnosed as “non-rejection” when they were authentic rejections and that they did not benefit from a treatment ensuring a good survival of their graft. »
You have 43.45% of this article left to read. The following is for subscribers only.
#algorithm #improves #diagnosis #kidney #transplant #rejection