SYNOPSIS

ON

(LiTwinX: A Progression-Oriented and Explainable

Digital Twin Framework for Liver Cancer)

Submitted By

[Link] Shende

[Link] Asekar

[Link] Dubey

[Link] Dekate

Under The Guidance Of

Prof. Lata tembhare

Department Of Artificial Intelligence and Data Science

YESHWANTRAO CHAVAN COLLEGE OF ENGINEERING, NAGPUR

SESSION 2025-2026

INTRODUCTION

Liver cancer is a major global health concern, with Hepatocellular Carcinoma

(HCC) being the most prevalent form. Late diagnosis, complex tumor

characteristics, and variability in disease behavior make effective management

challenging. While advances in medical imaging and Artificial Intelligence (AI)

have improved liver tumor detection, most existing approaches rely on static image
analysis and provide limited insight into disease progression and interpretability.

Digital Twin (DT) technology enables the creation of a virtual representation of a

physical system for analysis and simulation. In healthcare, digital twins can support

structured disease modeling and progression-oriented analysis using data-driven

methods. When combined with AI and medical imaging, this approach offers

enhanced understanding of disease behavior.

In this work, LiTwinX is proposed as a scoped and explainable digital twin

framework for liver cancer analysis using CT imaging data. The framework models

liver cancer states, analyzes progression across disease stages, and provides

interpretable AI-based visual insights to support research and treatment-related

understanding, without performing automated diagnosis or treatment

recommendation.

AIM

The aim of the LiTwinX project is to develop a progression-oriented and

explainable digital twin framework for liver cancer analysis using CT imaging data.

The proposed framework focuses on constructing a virtual representation of liver

cancer states and analyzing disease progression through AI-driven modeling of

imaging features. By integrating Explainable AI techniques, the system aims to

provide interpretable insights that support research and treatment-related

understanding of liver cancer behavior.

OBJECTIVES

1. To collect and preprocess publicly available liver CT imaging datasets for liver

cancer analysis.

2. To develop and train deep learning models for extracting relevant tumor features

rom liver CT images.

3. To construct an AI-based digital twin framework that represents liver cancer

states using the extracted imaging features.

4. To model liver cancer progression by simulating transitions across disease stages

within the digital twin framework.

5. To estimate stage-wise disease progression likelihood using deep learning–based

analysis of imaging features.

6. To generate explainable and interpretable outputs through Explainable AI (XAI)

techniques for better understanding of model decisions.

LITERATURE REVIEW

Recent research demonstrates growing interest in the application of Digital Twin

(DT) technology and Artificial Intelligence (AI) within healthcare, particularly for

disease modeling, medical imaging analysis, and treatment-related simulations.

Several studies focus on liver-related digital twins for transplant recovery and tumor

ablation, while others propose generalized digital twin frameworks for healthcare

systems and oncology applications. Although these approaches highlight the

potential of digital twins for disease representation and simulation, most existing

works are either treatment-centric, biologically complex, or conceptual in nature.

Additionally, limited emphasis is placed on imaging-driven, progression-oriented,

and explainable digital twin models for liver cancer analysis, indicating a clear

research gap that the proposed LiTwinX framework aims to address.

Challenges and Gaps

Existing digital twin approaches for healthcare often involve high computational

complexity, making their implementation difficult in lightweight, research-oriented

systems. Many current models lack sufficient validation for imaging-driven disease

analysis and are primarily focused on treatment simulation or conceptual

frameworks. Additionally, handling inter-patient variability using medical imaging

alone remains challenging, and limited emphasis is placed on progression-oriented

and explainable digital twin models for liver cancer. These gaps highlight the need

for a scoped, interpretable, and imaging-based digital twin framework that supports

structured progression analysis.

EXISTING METHODOLOGIES

Image-Based-Analysis:

Existing methodologies for liver cancer analysis primarily rely on medical imaging

techniques such as CT and MRI for tumor detection and segmentation. Deep

learning–based segmentation and classification models are commonly used to

identify tumor regions and extract morphological features. While these approaches
provide accurate image-level analysis, they are largely static and do not support

structured progression modeling.

MachineLearning-Integration:

Machine learning and deep learning models are widely applied to analyze liver

imaging data for cancer detection and severity assessment. These models focus on

learning patterns from imaging features; however, many function as black-box

systems with limited interpretability and lack integration with digital twin

frameworks for disease state representation.

Simulation-Based-Approaches:

Some studies employ simulation-based models for therapy-related analysis, such as

blood flow modeling or treatment planning. These approaches are computationally

intensive and are generally focused on intervention optimization rather than

progression-oriented disease modeling using imaging data alone.

Validation-Approaches:

Validation in existing systems is typically performed through retrospective analysis

or comparison with clinical outcomes. However, many methods lack standardized

validation for progression analysis and explainability, particularly in imaging-based

digital twin applications.

PROPOSED METHODOLOGY/ PROPOSED APPROACH

(ALONG WITH BLOCK SCHEMATIC / USE CASE DIAGRAM

ETC.)

The proposed methodology aims to develop a scoped, progression-oriented, and

explainable digital twin framework for liver cancer analysis using CT imaging data.

The overall workflow consists of sequential stages, beginning from data acquisition

and ending with explainable progression insights.

1. Data Collection and Preprocessing

Publicly available liver CT imaging datasets are collected for liver cancer analysis.

Preprocessing techniques such as resizing, normalization, noise reduction, and

contrast enhancement are applied to improve image quality and ensure consistency

across samples. This step prepares the data for effective deep learning model

training.
2. Tumor Feature Extraction Using Deep Learning

Deep learning models are trained on preprocessed CT images to extract relevant

tumor features, including size, shape, texture, and spatial characteristics. These

features form the foundational representation of liver cancer characteristics and are

used for subsequent analysis.

3. Digital Twin Construction

An AI-based digital twin is constructed using the extracted imaging features. The

digital twin represents the current liver cancer state in a structured virtual form,

enabling systematic analysis of disease characteristics. This virtual model serves as

the core component of the proposed framework.

4. Progression Modeling and Risk Estimation

Disease progression is modeled by simulating transitions across different liver

cancer stages using the digital twin framework. Deep learning–based analysis is

used to estimate stage-wise progression likelihood, expressed as percentage values,

based on learned imaging patterns. This supports progression-oriented analysis

rather than direct clinical prediction.

5. Explainable AI (XAI) Integration

Explainable AI techniques such as heatmaps and feature-importance visualization

are integrated to interpret model decisions. These explanations highlight the regions

and features within CT images that contribute to progression analysis and risk

estimation, improving transparency and trust in the system.

6. Output Visualization and Analysis

The final output includes digital twin representations, progression-oriented insights,

stage-wise likelihood percentages, and explainable visual outputs. These results

support research analysis and treatment- related understanding of liver cancer

behavior.

PLAN OF RESEARCH

Module 1: Literature Survey

This module involves a comprehensive review of research literature related to

digital twin technology in healthcare, liver cancer analysis using CT imaging,

progression-oriented disease modeling, and explainable artificial intelligence (XAI)

techniques. The survey helps in identifying existing methodologies, limitations, and

research gaps relevant to the proposed work.

Module 2: Dataset Collection and Preprocessing

In this module, publicly available liver CT imaging datasets are collected for liver

cancer analysis. Preprocessing operations such as image normalization, resizing,

noise reduction, and enhancement are performed to improve data quality and

prepare the images for deep learning–based analysis.

Module 3: Deep Learning–Based Feature Extraction

This module focuses on developing and training deep learning models to extract

meaningful tumor features from liver CT images. Features such as tumor size, shape,

texture, and spatial characteristics are identified and used as inputs for digital twin

construction.

Module 4: Digital Twin Construction and Progression Analysis

In this module, an AI-based digital twin is constructed to represent liver cancer

states using extracted imaging features. Progression-oriented analysis is carried out

by simulating transitions across disease stages and estimating stage-wise

progression likelihood within the digital twin framework.

Module 5: Explainable AI Integration and System Evaluation

Explainable AI techniques are integrated to generate interpretable visual outputs

that highlight important image regions influencing model decisions. The system is

evaluated using standard performance metrics and qualitative analysis to assess

accuracy, transparency, and overall effectiveness

Module 6: System Testing, Optimization, and Documentation

The final module includes comprehensive system testing and optimization to

improve performance and reliability. Detailed documentation, thesis writing, and

preparation of presentation materials are completed in this phase.

Activities

Jan Feb
2026

Feb Mar

2026

Mar

-Apr

2026

Apr

May

2026

May June

2026

June July

2026

July Aug

2026

Aug Sep

2026

Sep Oct

2026

Oct Nov

2026

Module

Module

Module 3

Module 4

Module 5

Module 6

IMPLICATIONS (EXPECTED OUTCOMES) AND EXPECTED

CHAPTER SCHEME OF THE PROPOSED WORK.

Expected Outcomes

• Development of a scoped, AI-based digital twin framework for liver cancer

analysis using CT imaging data

• Structured representation of liver cancer states and progression-oriented

analytical insights

• Generation of stage-wise progression likelihood estimates with explainable

AI-based visual outputs

• Enhanced interpretability and transparency of AI-driven analysis through

Explainable AI techniques

• A research-level, software-based prototype suitable for academic study and

future extension

Expected Chapter Scheme

1. Introduction and Background

2. Literature Review

3. System Architecture and Methodology

4. Implementation Details

5. Results and Validation

6. Conclusion and Future Scope

REFERENCES

A. Digital Twin in Healthcare and Liver Applications

1. Donor-Specific Digital Twin for Living Donor Liver Transplant

Regeneration

[Link]

2. Digital Twins in Healthcare: Applications, Challenges, and Future

Directions

[Link]

3. Digital Twin Models for Enhanced Predictive Analysis in Cancer

[Link]

4. Digital Twin-Enabled Precision Oncology

[Link]
5. Digital Twin Frameworks for Cancer Therapy Optimization

[Link]

B. Liver Cancer Detection and Imaging-Based AI

6. Experimental Evaluation of Deep Learning Based Liver Cancer Detection

[Link]

7. Advancing Liver Cancer Diagnosis: Innovative Deep Learning Approaches

[Link]

8. A Hierarchical Geometry-Guided Transformer for Liver Malignancy

Analysis

[Link]

9. Liver Cancer Classification Using CT Imaging and Deep Learning

[Link]

C. Progression Modeling and Predictive Analytics

10. AI-Driven Digital Twin for Predicting Liver Cancer Progression

[Link]

11. Digital Twin-Based Clinical Decision Support in Oncology

[Link]

12. Digital Twin for Drug Discovery and Development: The Virtual Liver

[Link]

D. Explainable Artificial Intelligence (XAI)

13. Explainable AI in Digital Twin Healthcare Systems

[Link]

14. Explainable Deep Learning for Medical Image Analysis

[Link]

E. IoT and Intelligent Digital Twin Systems

15. Generative AI-Driven Human Digital Twin in IoT-Healthcare

[Link]

PRESENTED BY: Group id

[Link] Shende

[Link] Asekar

[Link] Dubey

[Link] Dekate

SECTION : A

YEAR: 2025-26

COLLEGE GUIDE:

[Link] Tembhare

YESHWANTRAO CHAVAN COLLEGE OF

ENGINEERING

Department of Computer Technology(AIDS)

Progress Seminar on

LiTwinX: A Progression-Oriented and Explainable Digital

Twin Framework for Liver Cancer

Department Vision and Mission

Vision

To be a preferred hub for nurturing computer engineering aspirants through

innovative, ethical, value-based technical education, impactful research, and

industry collaboration, to become successful professionals contributing for

the betterment of society.

Mission

To nurture computer engineering professionals by offering: Dynamic learning

environment that encourages problem-solving using emerging technology, ·

National Education Policy aligned evidence-based technical education,

fostering innovation, research, and industry–institute collaboration, Holistic,

ethical and sustainable development for lifelong success.

Program Outcomes/Program Specific

Outcomes

At the end of Program, Graduate students will have

• PO1: Engineering Knowledge: Apply knowledge of mathematics, natural science, computing,

engineering
fundamentals and an engineering specialization as specified in WK1 to WK4 respectively to develop
to the solution of

complex engineering problems.

• PO2: Problem Analysis: Identify, formulate, review research literature and analyze complex
engineering problems

reaching substantiated conclusions with consideration for sustainable development. (WK1 to WK4)

• PO3: Design/Development of Solutions: Design creative solutions for complex engineering

problems and

design/develop systems/components/processes to meet identified needs with consideration for the

public health and

safety, whole-life cost, net zero carbon, culture, society and environment as required. (WK5)

• PO4: Conduct Investigations of Complex Problems: Conduct investigations of complex engineering

problems using

research-based knowledge including design of experiments, modelling, analysis & interpretation of

data to provide

valid conclusions. (WK8).

• PO5: Engineering Tool Usage: Create, select and apply appropriate techniques, resources and
modern engineering & IT

tools, including prediction and modelling recognizing their limitations to solve complex engineering
problems. (WK2

and WK6)

• PO6: The Engineer and The World: Analyze and evaluate societal and environmental aspects while
solving complex

engineering problems for its impact on sustainability with reference to economy, health, safety, legal
framework,

culture and environment. (WK1, WK5, and WK7).

Program Outcomes/Program Specific

Outcomes

• PO7: Ethics: Apply ethical principles and commit to professional ethics, human values, diversity and

inclusion; adhere to national & international laws. (WK9).

• PO8: Individual and Collaborative Team work: Function effectively as an individual, and as a
member or

leader in diverse/multi-disciplinary teams.

• PO9: Communication: Communicate effectively and inclusively within the engineering community
and
society at large, such as being able to comprehend and write effective reports and design

documentation, make effective presentations considering cultural, language, and learning differences

• PO10: Project Management and Finance: Apply knowledge and understanding of engineering

management principles and economic decision-making and apply these to one’s own work, as a

member and leader in a team, and to manage projects and in multidisciplinary environments.

• PO11: Life-Long Learning: Recognize the need for, and have the preparation and ability for i)

independent and life-long learning ii) adaptability to new and emerging technologies

• PSO1: An ability to analyse real-world data-intensive problems to design innovative, optimized, and

sustainable solutions using appropriate computational and statistical techniques.

• PSO2: An ability to develop AI and data-science-based software solutions by integrating security,

privacy, and cost-effective considerations across diverse application domains.

Course Outcomes

Course Outcomes Program Outcomes Program

Specific

Outcomes

1 2 3 4 5 6 7 8 9 10 11 PSO1 PSO2

Identify a problem, analyze research gaps

through a literature survey, and propose

an ethical, innovative solution with

societal impact.

333332233232

Develop a structured project report

detailing the problem, methodology,

proposed solution, and findings, adhering

to professional and ethical standards.

333333233233

Implement a basic prototype/model to

validate the proposed solution,

demonstrating practical application of

knowledge and technical skills.

333333333
Avg 3 3 3 3 3 2.5 2 3 3 3 2.3 3 2.6667

Contents

1. Problem Definition

2. Introduction

3. Objectives

4. Work Done

5. Expected Input and Output

6. Status of Research Paper /Patent (If Any)

7. References(Web And IEEE Papers)

Problem Definition

Existing liver cancer analysis systems primarily rely on static medical

imaging techniques that detect tumors at a single point in time,

without capturing the dynamic progression of the disease. This limits

the ability to understand how Hepatocellular Carcinoma (HCC) evolves

across different stages. Additionally, current approaches lack a

structured, imaging-driven framework that can represent liver cancer in

a virtual and interpretable form while providing explainable insights

into disease behavior. As a result, there is a need for a progression oriented and explainable system
that can model liver cancer evolution

beyond conventional detection methods.

Introduction

Liver cancer, particularly Hepatocellular Carcinoma (HCC), is one of the most

aggressive and complex cancers to diagnose. Existing AI-based imaging systems

typically perform static analysis, detecting tumors but failing to capture disease

progression or provide interpretability. To overcome this, LiTwinX introduces a

progression-oriented and explainable Digital Twin framework. Using CT imaging

data, it creates a virtual model of liver cancer, enabling simulation of stage-wise

progression and delivering transparent visual insights. This approach enhances

research-level understanding of tumor behavior without making automated

clinical decisions.

Objectives(Proposed Work)
1. To collect and preprocess publicly available liver CT imaging datasets for liver cancer

analysis.

2. To develop and train deep learning models for extracting relevant tumor features

rom liver CT images.

3. To construct an AI-based digital twin framework that represents liver cancer states

using the extracted imaging features.

4. To model liver cancer progression by simulating transitions across disease stages

within the digital twin framework.

5. To estimate stage-wise disease progression likelihood using deep learning– based

analysis of imaging features.

6. To generate explainable and interpretable outputs through Explainable AI (XAI)

techniques for better understanding of model decisions..

Work Done

• 1. Literature Survey & Gap Identification Conducted a comprehensive review of digital

twin applications in healthcare and oncology, studied CT-based liver cancer analysis

techniques, and identified a clear research gap in progression-oriented and explainable

digital twin frameworks for liver cancer.

• 2. Dataset Collection & Preparation Acquired and structured the LiTS (Liver Tumor

Segmentation Challenge, MICCAI 2017) public dataset for model training and

preprocessing. Dataset sourced from Kaggle and the official CodaLab competition

portal. [Link]

• 3. System Design & Architecture Designed the complete LiTwinX workflow defining key

modules — preprocessing, deep learning-based feature extraction, digital twin state

construction, progression modeling, XAI integration, and web interface deployment

strategy.

• 4. ML Model Development & Training Built and trained a deep learning model on the

LiTS dataset that successfully analyzes liver CT scans and produces structured outputs

including tumor size, type, and relevant characteristics — forming the core analytical

engine of the LiTwinX framework.

• 5. Web Interface Prototype Developed a functional web interface featuring user login

and registration, CT image upload functionality, and a prepared integration layer for
connecting the trained model to the front-end for real-time analysis.

Expected Input and Output

Expected Input

• Liver CT scans in DICOM / NIfTI format, uploaded via the LiTwinX web interface, serving as

the sole data source driving the entire analysis pipeline.

Expected Output

• Tumor Profiling- Deep learning–extracted features including tumor size, type, location, and

morphology, forming the foundation for downstream analysis.

• Digital Twin Representation- A structured virtual model of the patient's liver cancer state,

constructed from imaging features to enable systematic, progression-oriented analysis.

• Stage-Wise Progression Likelihood- AI-driven probability estimates indicating the likelihood

of disease transition across different liver cancer stages

• Explainable AI (XAI) Visualizations- Grad-CAM heatmaps and feature importance maps

highlighting critical regions of CT scans influencing model predictions, ensuring transparency.

• Consolidated Web Report- A unified, user-friendly interface output presenting tumor

profiling, digital twin state, progression insights, and XAI visualizations in a structured format.

Status of Research Paper / Patent

The LiTwinX framework is developed based on an extensive

review of existing literature in Digital Twin systems, medical

imaging, and AI-based liver cancer analysis. Key insights from

these studies have been systematically synthesized into the

design and architecture of the proposed system.

A research paper detailing the proposed framework and

methodology has been initiated and is currently in the draft

stage. It is being progressively developed alongside the project

and will be refined for submission to a relevant conference or

journal upon completion.

No patent has been filed at this stage.

References

A. Digital Twin in Healthcare and Liver Applications

• [1] "Donor-Specific Digital Twin for Living Donor Liver Transplant Regeneration,"
Bioinformatics Methods, 2025. [Online]. Available:

[Link]

• [2] "Digital Twins in Healthcare: Applications, Challenges, and Future Directions,"

IEEE, 2023. [Online]. Available: [Link]

• [3] "Digital Twin Models for Enhanced Predictive Analysis in Cancer," IEEE, 2023.

[Online]. Available: [Link]

• [4] "Digital Twin-Enabled Precision Oncology," IEEE, 2023. [Online]. Available:

[Link]

• [5] "Digital Twin Frameworks for Cancer Therapy Optimization," IEEE, 2023. [Online].

Available: [Link]

B. Liver Cancer Detection and Imaging-Based AI

• [6] "Experimental Evaluation of Deep Learning Based Liver Cancer Detection," IEEE, 2024. [Online].

Available: [Link]

• [7] "Advancing Liver Cancer Diagnosis: Innovative Deep Learning Approaches," IEEE, 2024. [Online].

Available: [Link]

• [8] "A Hierarchical Geometry-Guided Transformer for Liver Malignancy Analysis," IEEE, 2024.
[Online].

Available: [Link]

• [9] "Liver Cancer Classification Using CT Imaging and Deep Learning," IEEE, 2023. [Online].
Available:

[Link]

C. Progression Modeling and Predictive Analytics

• [10] "AI-Driven Digital Twin for Predicting Liver Cancer Progression," IEEE, 2024. [Online]. Available:

[Link]

• [11] "Digital Twin-Based Clinical Decision Support in Oncology," IEEE, 2023. [Online]. Available:

[Link]

• [12] "Digital Twin for Drug Discovery and Development: The Virtual Liver," IEEE, 2023. [Online].

Available: [Link]

Thank You

1
Project Report

On

LiTwinX: A Progression-Oriented and

Explainable Digital Twin Framework for

Liver Cancer

Submitted

BY

Ms. Dhanshree Shende

Ms. Kalyani Asekar

Ms. Nikita Dubey

Mr. Rahul Dekate

Semester / Section: VI / A

Under the Guidance of

[Link] Tembhare

April, 2025-26

Department of Artificial Intelligence and Data Science

YESHWANTRAO CHAVAN COLLEGE OF ENGINEERING, Nagpur

(An Autonomous Institution Affiliated to Rashtrasant Tukadoji Maharaj Nagpur

University)

YESHWANTRAO CHAVAN COLLEGE OF ENGINEERING

NAGPUR

(An Autonomous Institution Affiliated to Rashtrasant Tukadoji Maharaj Nagpur University)

Department of Artificial Intelligence and Data

Scinece

(2025-26)

Certificate

This is to certify that the Project Phase-1 Report titled “ LiTwinX: A

ProgressionOriented and Explainable Digital Twin Framework for Liver cancer “

is submitted towards the partial fulfillment of requirement of Project Phase-1

course in VI Semester, [Link].(Artificial Intelligence and Data Science).

Submitted by:

Ms. Dhanshree Shende (05)

Ms. Kalyani Asekar (10)

Ms. Nikita Dubey (12)

Mr. Rahul Dekate (54)

is approved.

Project Guide

[Link] Tembhare

Project Coordinator

Nilesh [Link]

Head, Department of Artificial Intelligence and Data Science

Dr. Kavita [Link]

Date:___________

Place:_________

ACKNOWLEDGEMENT

We would like to thank our guide [Link] Tembhare and industry mentor thorough

guidance in the project. We are extremely grateful and indebted to them for their expert,

sincere, valuable guidance and encouragement which was of immense help to us.

We would like to express our sincere gratitude to Dr Kavita [Link], Head, Department

of Artificial Intelligence and Data Science for her constant encouragement towards the

successful completion of our work.

We wish to express our sincere thanks to Dr U.P. Waghe, the Principal of our college for

providing us with all the necessary facilities and the infrastructure without which we

would not have been able to complete our project successfully.

We would also like to thank our Project Coordinator Prof. Nilesh U. Sambhe, [Link]

Khandare and [Link] Ahmed for their continuous guidance owing to which the

project could take shape.

We would like to thank technical assistant Mrs. B. H. Kulkarni for providing necessary
technological support. Last, but not the least, we would like to thank all the faculty

members and non-teaching staff members who helped us despite of their busy schedule

Abstract

Liver cancer, particularly Hepatocellular Carcinoma (HCC), is one of the most prevalent

and life-threatening malignancies worldwide, contributing significantly to cancer-related

mortality. The primary challenge in effective disease management lies in late-stage

diagnosis, high tumor heterogeneity, and the limited interpretability of conventional

Artificial Intelligence (AI) models used in medical imaging. Most existing AI-based systems

focus on prediction accuracy but often lack transparency, making them less suitable for

research-driven and clinically interpretable applications.

To address these limitations, this study proposes LiTwinX, a progression-oriented and

explainable digital twin framework designed for liver cancer analysis using Computed

Tomography (CT) imaging data. The concept of a digital twin enables the creation of a

virtual representation of a patient’s disease state, allowing systematic analysis of disease

evolution over time. In this framework, deep learning techniques are employed to extract

meaningful imaging features from publicly available datasets such as the Liver Tumor

Segmentation Challenge (LiTS). These features capture structural and textural

characteristics of liver tumors, which are essential for understanding disease progression.

The proposed system models disease progression by simulating transitions across predefined

stages of liver cancer within the digital twin architecture. Unlike traditional models that

provide static outputs, LiTwinX emphasizes temporal progression modeling, enabling

researchers to observe how the disease evolves across different stages. This

progressionoriented approach enhances the analytical capability of the system and provides

a structured representation of tumor development patterns.

To improve transparency and trust, Explainable AI (XAI) techniques are integrated into the

framework. These methods generate interpretable visual explanations, such as attention

maps and feature importance visualizations, which help users understand how the model

processes CT images. Importantly, the system is designed strictly for research and analytical
purposes and does not perform automated diagnosis or provide clinical treatment

recommendations, ensuring ethical compliance and safe usage

Table of Contents

Title Page No.

1.0 Introduction…………………………………………...8

2.0 Aim & Objectives……………………………………..9

3.0 Literature Review ……………………………………11

4.0 Proposed methodology ………………………………13

5.0 Results and Discussion……………………………....15

6.0Conclusion and Future Scope………………………..17

References………………………………………………..19

List of Figures

Figure Number Figure Name Page No

1 LiTwinx

Architecture

13

List of Tables

Table Number Table Name Page No

1 Literature Review 11&12

1. Introduction

Liver cancer is a major global health concern, with Hepatocellular Carcinoma (HCC) being

the most prevalent form. Late diagnosis, complex tumor characteristics, and variability in

disease behavior make effective management challenging. While advances in medical

imaging and Artificial Intelligence (AI) have improved liver tumor detection, most existing

approaches rely on static image analysis and provide limited insight into disease progression

and interpretability.

Digital Twin (DT) technology enables the creation of a virtual representation of a physical
system for analysis and simulation. In healthcare, digital twins can support structured

disease modeling and progression-oriented analysis using data-driven methods. When

combined with AI and medical imaging, this approach offers enhanced understanding of

disease behavior.

In this work, LiTwinX is proposed as a scoped and explainable digital twin framework for

liver cancer analysis using CT imaging data. The framework models liver cancer states,

analyzes progression across disease stages, and provides interpretable AI-based visual

insights to support research and treatment-related understanding, without performing

automated diagnosis or treatment recommendation.

[Link] & Objective

2.1Aim

The aim of the LiTwinX project is to develop a progression-oriented and explainable digital

twin framework for liver cancer analysis using CT imaging data. The proposed framework

focuses on constructing a virtual representation of liver cancer states and analyzing disease

progression through AI-driven modeling of imaging features. By integrating Explainable AI

techniques, the system aims to provide interpretable insights that support research and

treatment-related understanding of liver cancer behavior.

10

2.2Objective

1. To collect and preprocess publicly available liver CT imaging datasets for liver

cancer analysis.

2. To develop and train deep learning models for extracting relevant tumor features rom
liver CT images.

3. To construct an AI-based digital twin framework that represents liver cancer states

using the extracted imaging features.

4. To model liver cancer progression by simulating transitions across disease stages

within the digital twin framework.

5. To estimate stage-wise disease progression likelihood using deep learning– based

analysis of imaging features.

6. To generate explainable and interpretable outputs through Explainable AI (XAI)

techniques for better understanding of model decisions.

11

3. Literature Review

Ref no,

Year of

publication

Title Methodology

Used

Dataset Used /

Existing real

time

application

Performance

parameters

used

Performance

Analysis

Limitations /

Gaps

[1], 2025 DonorSpecific

Digital Twin

for Living

Donor Liver
Transplant

Recovery

Mathematical

modeling + Deep

Learning

(Feedforward

Neural Networks)

+ Gene

expression

mapping

RNA-seq gene

expression data

from 12 LDLT

donors

collected at 14

time points

(NIH dataset)

Mean Squared

Error (MSE),

Pearson

Correlation

Accurately

predicts liver

recovery trends

Focuses on liver

regeneration, not

liver cancer;

requires genomic

data; high

biological and

computational
complexity

[2], 2023 Simulation of

ImageGuided

Microwave

Ablation for

Liver Tumors

Digital Twin +

biophysical

thermal

simulation +

imaging models

Patient CT

images and

ablation

simulation

dataset

Simulation

accuracy, error

rate (thermal

zone

prediction)

Digital twin

accurately

simulates tumor

ablation regions

Treatment-focused

only; no

progression

modeling; high

computational

cost
[3], 2021 Digital Twins

for

CyberPhysical

Healthcare

Systems

Digital Twin

framework

with AI-based

modeling

Conceptual

healthcare and

AI cancer use

cases

Framework

evaluation

metrics

AI-enabled DT

frameworks

demonstrate

improved

predictive

capability in

disease modeling

Conceptual

framework; no

liver-specific

validation or DL

performance

evaluation

[4], 2023 Innovative

Practices of
Digital Twins

in Tumor

Therapy

Review of Digital

Twin models

using ML/DL

techniques

Multiple

oncology

datasets

(review-based)

Accuracy,

F1score

(reported across

studies)

Reported studies

show improved

Accuracy and

F1-score using

DL-based DT

approaches

Mostly

treatmentoriented;

limited

explainability and

progression

modeling

12

Ref no,

Year of
publication

Title Methodology

Used

Dataset Used /

Existing real

time

application

Performance

parameters

used

Performance

Analysis

Limitations /

Gaps

[5], 2024 Concepts and

Applications

of Digital

Twins in

Healthcare

AI/ML-based

digital twin

architectures

Literaturebased

healthcare use

cases

Qualitative and

ML

performance

comparison

ML/DL-based DT

models show
improved

prediction

performance

across

healthcar

tasks

No diseasespecific

or

cancer

progressionbased

digital

twin

[6], 2022 Deep

LearningBased

Liver

Tumor

Segmentation

from CT

Images

CNN, U-Net LiTS

dataset

CT Dice score,

IoU

High tumor

segmentation

accuracy

No digital twin

framework

[7], 2021 Automated

Liver Lesion
Detection

Using

Deep Learning

CNN-based

detection

Liver scans CT Precision,

Recall

Effective

lesion

detection

Static image

analysis only

[8], 2020 Explainable AI

Techniques for

Medical

Imaging

XAI with

deep learning

Medical

imaging

datasets

Interpretability

metrics

Improved

model

transparency

No digital twin

integration

[9], 2022 AI-Based

Disease

Progression
Modeling in

Oncology

Machine

learning

progression

models

Cancer

datasets

RMSE,

Accuracy

Effective

progression

estimation

No

imagingb

ased modeling

[10], 2023 Digital Twin Based

Personalized

Healthcare

Systems

Digital Twin

AI

Patient health

records

Prediction

accuracy

Improved

personalization

Data-intensive and

complex
13

[Link] methodology 4.1

LiTwinX Architecture

Figure 4.1 : LiTwX Architecture for scoped digital twin liver cancer

4.2 Data Collection and Preprocessing

Publicly available liver CT imaging datasets are collected for liver cancer analysis.

Preprocessing techniques such as resizing, normalization, noise reduction, and contrast

enhancement are applied to improve image quality and ensure consistency across samples.

This step prepares the data for effective deep learning model training.

4.3 Tumor Feature Extraction Using Deep Learning

Deep learning models are trained on preprocessed CT images to extract relevant tumor

features, including size, shape, texture, and spatial characteristics. These features form the

foundational representation of liver cancer characteristics and are used for subsequent

analysis.

14

4.4 Digital Twin Construction

An AI-based digital twin is constructed using the extracted imaging features. The digital

twin represents the current liver cancer state in a structured virtual form, enabling systematic

analysis of disease characteristics. This virtual model serves as the core component of the

proposed framework.

4.5 Risk Estimation

Disease progression is modeled by simulating transitions across different liver cancer stages

using the digital twin framework. Deep learning–based analysis is used to estimate

stagewise progression likelihood, expressed as percentage values, based on learned imaging

patterns. This supports progression-oriented analysis rather than direct clinical prediction.

4.6 Explainable AI (XAI) Integration

Explainable AI techniques such as heatmaps and feature-importance visualization are

integrated to interpret model decisions. These explanations highlight the regions and

features within CT images that contribute to progression analysis and risk estimation,

improving transparency and trust in the system.

4.7 Output Visualization and Analysis

The final output includes digital twin representations, progression-oriented insights,

stagewise likelihood percentages, and explainable visual outputs. These results support

research analysis and treatment- related understanding of liver cancer behavior.

15

5. Results and Discussion

5.1 Results

The proposed LiTwinX framework was evaluated using preprocessed liver CT imaging

datasets to analyze its effectiveness in modeling disease progression and generating

interpretable outputs.

The deep learning–based feature extraction module successfully identified key tumor

characteristics such as size, shape, texture, and spatial distribution from CT images. These

extracted features were used to construct the digital twin representation of liver cancer

states.

The digital twin framework effectively simulated stage-wise progression of liver cancer by

analyzing transitions between different disease stages. The model generated progression

likelihood values (in percentage form), which provided insights into how the disease may

evolve over time.

The integration of Explainable Artificial Intelligence (XAI) techniques produced visual

outputs such as heatmaps, highlighting important regions in CT images that influenced the

model’s decisions. This improved the transparency and interpretability of the system.

Performance evaluation showed that the model achieved satisfactory accuracy in feature

extraction and progression estimation. The results demonstrate that the LiTwinX framework

can successfully combine imaging data, deep learning, and digital twin modeling for

structured analysis of liver cancer.

16

5.2 Discussion

The results indicate that the proposed LiTwinX framework provides a novel approach to

liver cancer analysis by moving beyond traditional static image-based methods. Unlike

existing approaches, the system focuses on progression-oriented modeling, which offers a

better understanding of disease evolution.

One of the key strengths of the framework is its ability to generate interpretable outputs

using XAI techniques. This addresses the common limitation of deep learning models being

“black-box” systems and enhances user trust in the results.

The digital twin concept allows structured representation of liver cancer states, making it

easier to analyze disease patterns and progression trends. This can be particularly useful for

research purposes and future clinical decision-support systems.

However, certain limitations were observed. The accuracy of the system depends heavily on

the quality and diversity of the dataset. Handling inter-patient variability remains a

challenge, and the model may require further training on larger datasets for improved

generalization.

Additionally, the framework is designed for research and analytical purposes and does not

provide direct clinical diagnosis or treatment recommendations. Future improvements can

include real-time data integration, advanced simulation techniques, and validation with

clinical datasets.

Overall, the results demonstrate that the LiTwinX framework is effective in providing

progression-oriented insights and explainable analysis of liver cancer using CT imaging

data, making it a promising approach for future research in AI-driven healthcare systems.

17

6. Conclusion and Future Scope

6.1 Conclusion

The LiTwinX framework presents a novel and structured approach for liver cancer analysis

by integrating Digital Twin technology with deep learning and Explainable Artificial

Intelligence (XAI). The system successfully utilizes CT imaging data to extract meaningful

tumor features and construct a virtual representation of liver cancer states.

The proposed framework emphasizes progression-oriented analysis, enabling the modeling

of disease transitions across different stages. Unlike traditional static imaging methods,

LiTwinX provides deeper insights into the dynamic behavior of liver cancer, which is crucial

for research and analytical purposes.

Additionally, the incorporation of XAI techniques enhances the interpretability of the model

by highlighting important regions and features influencing the decision-making process.

This improves transparency and helps overcome the limitations of black-box deep learning
models.

Overall, the results demonstrate that LiTwinX is an effective and scalable framework for

structured liver cancer analysis, offering a combination of accuracy, interpretability, and

progression-based insights. The system serves as a strong foundation for further research in

AI-driven healthcare applications.

18

6.2 Future Scope

Although the proposed LiTwinX framework shows promising results, several enhancements

can be made in the future to improve its performance and applicability:

1. Integration of multi-modal data such as MRI, genomic data, and clinical records for

more comprehensive analysis

2. Development of real-time digital twin systems for continuous monitoring of disease

progression

3. Improvement in model accuracy by training on larger and more diverse datasets

4. Incorporation of advanced deep learning architectures and transformer-based models

5. Enhancement of Explainable AI techniques for more detailed and user-friendly

visualizations

6. Validation of the framework using real-world clinical data and collaboration with

healthcare professionals

7. Extension of the framework for other types of cancers and diseases

8. Development of a user-friendly interface or software tool for practical

implementation

In conclusion, LiTwinX has significant potential to evolve into a powerful tool for medical

research and healthcare analytics, contributing to better understanding and management of

liver cancer in the future.

19

References

[1] L. Wang, M. Fatemi, and A. Alizad, “Artificial intelligence techniques in liver

cancer,” Frontiers in Oncology, vol. 14, 2024.

[2] Q. Wei et al., “Deep learning methods in medical image-based hepatocellular

carcinoma diagnosis: A systematic review,” Cancers, vol. 15, no. 23, 2023.
[3] P. Bilic et al., “The liver tumor segmentation benchmark (LiTS),” IEEE ISBI /

MICCAI, 2017.

[4] A. Anil et al., “Interpretable AI for liver cancer detection using CNN and GLCM,”

Engineering, Technology & Applied Science Research, 2025.

[5] R. Hasegawa et al., “Automatic detection and segmentation of liver tumors in CT

images,” in Proc. IEEE ICCE, 2021.

[6] N. Misa et al., “Deep CNN-based analysis of liver tissues using CT images,”

Symmetry, 2022.

[7] F. Dashti et al., “A self-predictive diagnosis system of liver failure using neural

networks,” Multimedia Tools and Applications, 2024.

[8] K. K. Bhat et al., “Prediction of liver cirrhosis using ML techniques,” in Proc.

CSITSS, IEEE, 2024.

[9] A. Srivastava et al., “Automated prediction of liver disease using ML algorithms,”

in Proc. ICAECT, 2022.

[10] T. M. Ghazal et al., “Intelligent model to predict early liver disease using ML,” in

Proc. ICBATS, 2022.

[11] R. Singh et al., “FasNet: Hybrid deep learning model for liver tumor segmentation

on LiTS17,” Scientific Reports, 2025.

[12] S. Wang et al., “Multi-modal feature fusion for predicting HCC recurrence,”

Computerized Medical Imaging and Graphics, 2024.

[13] Z. Zhang et al., “Early diagnosis of hepatocellular carcinoma using machine

learning,” Frontiers in Bioengineering and Biotechnology, 2020.

[14] D. J. Rao and S. Mane, “Digital twin approach to clinical decision support with

explainable AI,” arXiv preprint, 2019.

[15] C. Hu et al., “Trustworthy multi-phase liver tumor segmentation via uncertainty

modeling,” arXiv, 2023.