Automated Computer Assistant for Kidney Transplant Rejection Diagnostics: Interview with Study Authors

Automated Computer Assistant for Kidney Transplant Rejection Diagnostics: Interview with Study Authors
Automated Computer Assistant for Kidney Transplant Rejection Diagnostics: Interview with Study Authors


A recent study in Nature Medicineentitled “An automated histological classification system for precision diagnostics of kidney allografts,” has showcased the efforts of a group of researchers who have developed an automated system that can diagnose kidney transplant rejection.

A variety of disparate factors can affect the chances that a transplant will be rejected. At present, clinicians have to manually consider these complex data when making decisions about transplant patients, which can lead to a high level of misdiagnosis and patient morbidity.

This new system incorporates an algorithm that can synthesize these complex clinical data into a reliable answer for busy clinicians. So far, the researchers tested the system with over 4,000 kidney transplant patients in Europe and the United States, and have shown that the technology can avoid 40% of human misdiagnoses of allograft rejection.

Schematic diagram illustrating the decoding, encoding, and rectifying processes used to construct the Banff Automation System. During the development process, the multidisciplinary consortium (pathologists, transplant physicians, data scientists, and developers) worked closely together and improved the application numerous times. The outputs of the application are a decision tree for better visualization of the process that generates the diagnosis and automated reports in either PDF or Excel format. Abbreviations: C4d, complement component C4d staining; DSA, donor-specific antibody.

Medgadget had the opportunity to speak with some of the authors of the study. These are Alexandre Loupy, senior author, nephrologist with Department of Kidney Transplantation Necker Hospital (APHP) and data science expert at INSERM and director of the Paris Institute for Transplantation and Organ Regeneration (Université Paris Cité), Valentin Goutaudier (co-first author), nephrologist and epidemiologist, Paris Institute for Transplantation and Organ Regeneration (Université Paris Cité) and Department of Kidney Transplantation at Necker Hospital, and Daniel Yoo (co-first author), data scientist, Paris Institute for Transplantation and Organ Regeneration (Université Paris Cité).

Conn Hastings, Medgadget: Please give us a brief overview of the factors affecting kidney transplant rejection and failure.

Alexandre Loupy: Rejection is the main cause of graft failure after kidney transplantation and is a major public health problem given the current global organ shortage.

Many factors can affect kidney transplant rejection: medication non-compliance, underimmunosuppression by physicians, HLA incompatibility and other immunological incompatibilities (blood group, other non-HLA antigens), the presence of donor-specific antibodies (i.eantibodies produced by the recipient’s immune system that specifically target and react against the antigens present on the cells of the transplanted organ), infections that can activate the immune system (e.g., viral infections such as cytomegalovirus), reperfusion injury, age, underlying medical conditions, quality of the transplanted organ, etc. Managing these factors through proper medication and follow-up care can help prevent kidney transplant rejection and increase the likelihood of long-term transplant success.

Another important factor of kidney transplant failure is simply the misdiagnosing of rejection. The diagnosis of rejection relies on an international classification, called the Banff classification, which has become considerably more complex over the past 30 years due to the use of modern precision medicine applied to this multifactorial disease. It is now necessary for physicians to analyze and integrate complex and extremely diverse data – histological, immunological, and transcriptomic data – to make a correct diagnosis that will guide the therapeutic management of patients. However, if physicians misdiagnose rejection, they also can not choose the best treatment for their patients, thus increasing the risk of allograft failure.

Medgadget: What consequences does transplant rejection have for patients and healthcare providers?

Valentin Goutaudier: Transplant rejection can have significant consequences for both patients and healthcare providers.

  • Consequences for patients:
    • Diminished organ function: this can lead to graft failure, and then the need for dialysis or repeat transplantation.Health complications related to the treatment of rejection episodes (i.eimmunosuppressive drugs at high dose): infections, cardiovascular events, etc.Increased risk of mortality.Emotional and psychological impact.
  • Healthcare providers:
    • Treatment complexity: managing transplant rejection requires close monitoring, frequent laboratory tests, and adjustments of immunosuppressive medications. This task can be complex and very time-consuming.Resource utilization: treating transplant rejection may need hospitalization, additional diagnostic tests, etc. This can increase healthcare costs.Emotional strain, given the negative impact on patient’s health, which could potentially lead to burnout and compassion fatigue.
    • Transplant programs: rejection can have an impact on the success rates of transplantation programs and impact the reputation of healthcare providers and transplant centers.

Medgadget: How has the risk of kidney transplant rejection been determined previously? What are the limitations of this approach?

Valentin Goutaudier: For 30 years, the gold standard to define rejection is the international Banff classification. It requires the integration of data from a kidney transplant biopsy (i.ehistological lesions scores), as well as clinical, immunological, and transcriptomic data. These multimodal data and rules are necessary to provide a precise diagnostic, which considers all the biological operating processes, grades, and types of rejection.

The main limitation of this approach is that it is now very difficult for pathologists and physicians to interpret all these complex data and make a correct diagnosis. The consequence is that a lot of misdiagnoses of rejection are made in routine practice and in clinical trials, leading to deleterious therapeutic decisions for patients.

Medgadget: How did this new tool come about? What inspired its creation?

Alexandre Loupy: This complexity in the diagnosis of rejection, initially necessary to better understand and define its type and severity, has become a daily challenge for physicians faced with situations where it can be difficult to make a correct diagnosis. In the face of the growing number of diagnostic errors continually documented in the scientific literature, international transplantation societies have called on researchers from around the world to react and find a solution to simplify the diagnosis of rejection. Our hypothesis was that an automation of the classification could solve this issue. Hence, we aimed to develop a computer-based diagnostic support tool that is reliable, robust, accurate, widely validated, and demonstrate a real and measurable benefit for patients.

Medgadget: Please give us an overview of the tool and how it works to assess the likelihood of a transplant rejection.

Daniel Yoo: The tool is the result of the work of an international consortium of experts in rejection and health data. In a first phase, we conducted a systematic review of the scientific literature to collect and decode all the diagnostic rules of the rejection classification published over the past 30 years. Then we translated these diagnostic rules into a computer algorithm covering all possible rejection scenarios and created an easy-to-use automated computer assistant available online, which instantly interprets the complex medical data entered by physicians using the algorithm and provides a diagnosis that strictly considers the rules of the classification. With just a few clicks, the computer assistant outputs an analysis report with the correct diagnosis and a decision tree that explains the algorithm’s reasoning to avoid any “black box” effect.

Medgadget: How does the tool compare in its accuracy in predicting transplant rejection compared with conventional approaches?

Daniel Yoo: This answer is very simple. As the algorithm strictly follows the Banff diagnostic rules, its accuracy rate to predict rejection using the available data is… 100%!

Medgadget: How do you see these types of technology progressing in the future?

Alexandre Loupy: Our study is a significant step towards the development of precision medicine accompanied by automated computer systems. In fact, we are the first, across all medical specialties, to demonstrate that a computer assistant can help doctors to make better diagnoses. Moreover, our tool, given its intrinsic biotechnology, is reliable, robust, accurate, and provides a real and measurable benefit for patients.

Transplantation is not the only medical specialty facing increasingly complex data. I have no doubt that doctors in other specialties – such as oncology and immunology, where diverse and complex data are increasingly used – will embark on the adventure of automating disease classifications to improve the management of their patients.

Study in Nature Medicine: “An automated histological classification system for precision diagnostics of kidney allografts

(function(d, s, id)
var js, fjs = d.getElementsByTagName(s)[0];
if (d.getElementById(id)) return;
js = d.createElement(s); js.id = id;
js.src = “//connect.facebook.net/en_US/sdk.js#xfbml=1&version=v2.4&appId=2624537092”;
fjs.parentNode.insertBefore(js, fjs);
(document, ‘script’, ‘facebook-jssdk’));



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *