End-stage renal disease is the permanent cessation of the nephrons’ (kidney cells) capability to eliminate unwarranted body fluids and noxious wastes from the body. Whereas dialysis bids a nonnatural medium to sustain kidney-like function, by only 10% to say the least, kidney transplant has effectively passed the former therapeutic modality, resulting in improved health outcomes and longevity for kidney transplant recipients. Glomerular filtration rate and kidney biopsy are the mainstay methods for evaluation of kidney allograft function after transplant. However, these widely used diagnostic methods are devalued by marginal sensitivity, delayed disease indication, and clinical complications. Recently, the role of deep learning in early diagnosis of acute kidney allograft rejection has been investigated. Because deep learning can capture delicate underlying hierarchical data patterns, the ability to enhance computer-aided diagnostic models for the early projection of acute kidney allograft rejection could bring a real paradigm shift in the tactic that kidney transplant physicians utilize, allowing for precise and early identification of acute kidney allograft rejection.
Key words : Early diagnosis, End-stage renal disease, Renal transplantation
End-stage renal disease is the permanent cessation of the nephrons’ (kidney cells) capability to eliminate unwarranted body fluids and noxious wastes from the body. Although dialysis bids a nonnatural medium to sustain kidney-like function, by only 10% to say the least, kidney transplant has effectively passed this former therapeutic modality, resulting in improved health outcomes and longevity for kidney transplant recipients. Nevertheless, organ procurement and transplant, be it kidney or any other organ, are extremely convoluted processes. In fact, kidney allograft rejection by the immune system is 15% imminent in the first 5 years after transplant. Therefore, rescuing the kidney allograft from immune-mediated rejection is of utmost significance in transplantation.
Acute rejection is the predominant source of kidney allograft dysfunction. According to the National Kidney Foundation, glomerular filtration rate is the mainstay method for evaluation of kidney allograft function after transplant. However, this widely used diagnostic method is devalued by its marginal sensitivity and delayed disease indication. Kidney microstructure investigations (biopsies) are perhaps the gold standard for assessment of allograft function and for ascertaining the origin of underlying dysfunction in the kidney allograft. However, biopsies are high-cost and invasive procedures and carry their own clinical perils, including kidney trauma and bleeding. Furthermore, a kidney biopsy may not reflect the gravity of the allograft rejection as the biopsy sample only embodies a small segment of the kidney cells.1
Apart from glomerular filtration rate and histopathological exploration, kidney imaging modalities have significantly improved investigations of acute kidney allograft rejection dynamics noninvasively, for instance, by radionuclide imaging, helical computed tomography, and ultrasonographic imaging. However, the utilization of these imaging options is limited by their low sensitivity, questionable validity, and adverse health ramifications such as contrast-induced nephrotoxicity.1 Of late, magnetic resonance imaging (MRI) studies, diffusion-weighted (DW) MRI in particular, have been eyed for the anatomical and functional assessment of the kidney allograft. This option is relatively superior because it is a contrast-free imaging modality and has been used to detect and classify tumor tissue and for kidney imaging.2
A handful of studies have probed the implications of DW-MRI in identifying appropriate kidney allograft function through formation of quantitative maps, also called apparent diffusion coefficients, at various magnetic field strengths and durations (b values).3-5 The apparent diffusion coefficient is the factor of discrimination between biopsy-guided normal kidney allografts and pathological kidney allografts, including those with acute rejection. However, previous findings were deemed inadequate because of absence of integration of clinical imaging and biomarkers and advanced machine learning, such as deep learning, to distinguish between normal and pathological kidney allografts.
Because deep learning can capture delicate underlying hierarchical data patterns, it is ideal to enhance computer-aided diagnostic (CAD) models to allow for the early projection of acute kidney allograft rejection. The DW-MRI, a benchmark diagnostic modality for acute rejection, appears appropriate for development of state-of-the-art deep learning algorithms for good predictive models of acute kidney allograft rejection. In 2019, Abdeltawab and colleagues proposed a unique deep learning-based CAD model for early identification of kidney allograft dysfunction.1 Their deep learning-based CAD model showed an overall diagnostic accuracy of 92.9%, sensitivity of 93.3%, and specificity of 92.3% in differentiating between kidney allografts with and without rejection, irrespective of the geographical dissimilarities and/or kidney imaging practices.
In conclusion, development of deep learning-based diagnostic methods could certainly mine subtle predictive data patterns from DW-MRI and bring a real paradigm shift in the tactic that kidney transplant physicians can utilize for precise and early identification of acute kidney allograft rejection.
Volume : 19
Issue : 2
Pages : 176 - 177
DOI : 10.6002/ect.2020.0238
From the 1Prince Mohammed bin Abdulaziz Hospital, Riyadh, Saudi Arabia; the
2Institute of Nursing, Dow University of Health Sciences, Karachi, Pakistan; and
the 3School of Public Health, Dow University of Health Sciences, Karachi,
Pakistan and Department of Medicine, Aga Khan University Hospital, Karachi,
Acknowledgements: The authors have not received any funding or grants in support of the presented research or for the preparation of this work and have no potential declarations of interest.
Corresponding author: Akbar Shoukat Ali, Department of Medicine, Aga Khan University Hospital, Karachi, Pakistan and School of Public Health, Dow University of Health Sciences, Karachi, Pakistan
Phone: +92 312 8624328