Study Group Leader: Deniz Balcı
Early Allograft Dysfunction (EAD) refers to the insufficient functioning of a graft in the days
following liver transplantation. EAD is characterized by a range of clinical symptoms that threaten
the life of both the graft and the patient, including hyperbilirubinemia, coagulopathy, encephalopathy,
and/or ascites. During Living Donor Liver Transplantation (LDLT), if the graft is small relative to the
recipient’s needs, a condition known as Small Graft Syndrome (SGS) may occur. SGS develops due
to excessive portal venous flow and resulting high sinusoidal pressure in a small graft volume.
Preventing the development of this syndrome is critical. Clinicians have developed a treatment
approach called "portal flow modulation," which regulates the flow to the graft. The underlying
mechanism of SGS is thought to be the excessive portal flow leading to increased pressure within the
graft after reperfusion, resulting in sinusoidal damage. In response to excessive portal flow, the liver
undergoes adaptive changes, including vasoconstriction of the hepatic artery. While high portal flow
causes damage through excessive elevation in the graft's portal and sinusoidal systems, endothelial
activation, and sinusoidal shear stress, the resultant arterial vasoconstriction leads to secondary
ischemia through oxidative stress. It is well known that oxidative stress contributes to liver tissue
damage by targeting cell membranes, lipids, and proteins via free radicals, leading to various clinical
conditions ranging from cirrhosis to hepatocellular carcinoma (HCC). It plays a significant role in the
pathophysiology of various liver diseases, such as Alcohol-Related Liver Disease (ALD), Metabolic
Dysfunction-Associated Liver Disease (MALD), and Hepatitis C. Endothelial mitochondrial
dysfunction and oxidative stress are closely related to ischemia-reperfusion injury during organ
transplantation, as well as various liver diseases like cholestasis, steatosis, viral hepatitis, and drug-
induced liver injury. Therefore, a better understanding of mitochondrial dysfunction and endothelial
oxidative stress is thought to aid in developing new mitochondrial-targeted therapies or interventions.
In our study, we plan to evaluate the viability, specifically mitochondrial activity, of the liver graft
exposed to ischemic processes during transplantation surgery. Subsequently, we aim to determine the
redox status based on the Nicotinamide Adenine Dinucleotide/Flavin Adenine Dinucleotide
(NAD/FADH) ratio using an algorithm based on known spectroscopic principles. Mitochondria, the
organelle where energy metabolism primarily occurs, contain the components of the electron
transport chain (ETC) in their inner membrane. In this chain, oxygen acts as an electron carrier at
sites where enzymatic processes occur; any oxygen irregularity halts the progression of the chain,
leading to disruptions in ATP production and resulting in NAD consumption and FADH
accumulation, creating an imbalance in the NAD/FADH ratio. This ratio is assessed as the
mitochondrial redox status and offers a new perspective for evaluating oxidative processes and tissue
health.
The ischemia/reperfusion process can be predicted during transplant operations based on existing
blood flow. However, recent advances in microcirculation assessment have highlighted that
evaluating tissue oxygenation solely through flow or oxygen diffusion is inadequate. Even if regional
blood flow and oxygen diffusion are ensured, it remains uncertain whether tissue cells can utilize this
oxygen. Therefore, evaluating endothelial function has become essential for accurate assessments of
oxygenation and tissue health status. Evaluating mitochondrial activity, which is responsible for
energy production and thus vital functions, has become a current and reliable approach for assessing
Page 3 of 1
endothelial function. Early detection of EAD, characterized by hyperbilirubinemia and coagulation
dysfunction, in cases where the damage mechanisms mentioned above occur postoperatively, may
offer a life-saving treatment approach and provide clinicians with a crucial tool for guiding treatment
based on graft performance and postoperative development.
In conclusion, our study aims to predict the development of SGS and EAD that may arise after living
donor liver transplantation by utilizing intraoperative hemodynamic and spectrophotometric methods.
We believe that assessing endothelial damage resulting from microcirculatory imbalances due to flow
changes at the mitochondrial level and its relationship with the clinically significant EAD and SGS,
using new parameters, will provide important scientific insights and innovations.
Analysis and Modeling Approach of Imaging Data in Organ Transplantation
Imaging procedures performed after organ transplantation are of great importance for evaluating the
success of the transplant and for the early detection of possible complications. In this study, it is
aimed to determine whether the organ transplantation has been performed successfully by using two
numerical parameters obtained through image processing tools: the heterogeneity index and the
mitochondrial redox status. The objective is to develop a model that predicts the success of organ
transplantation mainly based on these a few parameters (EAD criteria). This constitutes a binary
classification problem involving the classification of the organ as a successful or an unsuccessful
transplant.
In the modeling process, it is planned to use three different machine learning methods: logistic
regression, support vector machines (SVM), and feedforward neural networks (FFNN). The
reason for utilizing logistic regression and SVM is that these models are simple and can provide
effective results with a small number of inputs. Particularly when working with two parameters, the
logistic regression model allows for easy interpretation of the results and can be trained quickly.
SVM, on the other hand, can achieve high accuracy rates even with limited datasets and can provide
better generalization by reducing the risk of overfitting. FFNN has the potential to capture non-linear
relationships between inputs. However, when working with only two parameters, the complexity of
the model and the risk of overfitting should be considered. Therefore, it is aimed to determine the
most suitable model by comparing the performance of FFNN with logistic regression and SVM.
All three models will be trained on the same dataset, and their performances will be compared.
During the training of the models, the data will be divided into training and test sets, and
hyperparameter adjustments will be made. The performance of the models will be evaluated on the
test data using metrics such as accuracy, precision, recall, and F1 score. Additionally, the
generalization capabilities of the models will be measured using the k-fold cross-validation method.
In this process, as the first step, the heterogeneity index and mitochondrial redox status will be
calculated using image processing tools, and the data will be cleaned and prepared in an appropriate
format for training of the machine learning methods. Subsequently, the models will be trained with
this data, and performance of each machine learning method will be compared.
