Presentation On Research Paperarial Font

The document discusses using an explainable AI model called SCI-XAI to enable early diagnosis of chronic kidney disease. SCI-XAI allows obtaining a balanced prediction model in terms of classification accuracy and explainability. It analyzes features like hemoglobin and hypertension and achieves 99.2% accuracy, providing a promising solution for early CKD diagnosis especially in developing countries.

Uploaded by

Rejoana Islam Bonna

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

22 views9 pages

Presentation On Research Paperarial Font

Uploaded by

Rejoana Islam Bonna

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 9

Evaluation of an Explainable AI Model

Early Diagnosis of Chronic

Kidney Disease using
eXplainable AI
How XAI can improve medical diagnosis
Table of Contents
01. XAI
Explainable AI is a model that
balances classification accuracy
and explainability.

02. CKD
A fatal diseases, which is
expensive for both treatment and
early diagnosis

03. Feature scaling

Feature selection strategies have
proven to be influential in
achieving the most accurate
classifier.
Why we should
use XAI in CKD
• CKD(Chronic kidney disease) is a fatal
disease, in which the affected patient can’t
recover if diagnosed later.
• Here, The XAI helps to early diagnosis.
• It has an accuracy of 99.2% .
• It gives results in a reduced cost of the
early diagnosis of CKD implying a
promising solution for developing countries
where medical access is more difficult .
The Model
• This model is developed using an
automated optimization, framework, named
SCI-XAI (feature Selection and
Classification for Improving eXplainable AI)

• SCI-XAI, implemented with the Python

scikit-learn package, allows obtaining a
balanced prediction model in terms of
classification performance (accuracy) and
explainability
The SCI-XAI framework tackles the
preprocessing of the data in three phases

Missing data handling Data Encoding Feature Selection

• Mean value imputation • A minimum-maximum • Feature selection
is used for numerical scaling process is procedures can remove
features. applied to numerical features with non-
• while the mode is used features, while the relevant information
for ordinal and nominal categories of the ordinal from the classification.
features. and nominal features • Enhances models
are encoded into explainability
numbers
EXPLAINABILITY ANALYSIS OF THE
PREDICTION MODEL

• XGBoost is the most balanced model in

terms of explainability and accuracy.
• Hemoglobin, hypertension, and specific
gravity features is analyzed using different
post-hoc explainability techniques to
demonstrate their influence on the model’s
output.
• Hemoglobin feature has the greatest
attribution to the CKD probability.
• The presence of hypertension increases the
CKD probability
Results

Feature selection CLASSIFICATION EXPLAINABILITY

result PERFORMANCE METRICS RESULTS
RESULTS
• XGBoost is seen as the • All classifiers considered • In this case XGBoost
best classifiers to select. obtain an accuracy higher shows the best result, 88%
than 97.5%. interpretability
Things that could have
been improved for
better solution
• public open dataset from UCI-ML
repository is used here which lacks an
external validation to support objective
experimentation
• more CKD data would be needed from a
different clinical setting from the original
Unrecognized Points