Page 69 - AIH-2-3

Artificial Intelligence in Health

ORIGINAL RESEARCH ARTICLE
Optimized convolutional neural network model

for multilevel classification in leukemia diagnosis
using Tversky loss

Kumari Pritee * and Rahul Dev Garg 2
1
1 Department of Information System Management, IIM Sambalpur, Sambalpur, India
2 Department of Geomatics Engineering, IIT Roorkee, Roorkee, India

Abstract
Leukemia diagnosis traditionally depends on time-intensive examination of blood
cell morphology, a process prone to human error. To address these challenges, this
study explores the use of convolutional neural networks (CNNs) optimized with
the Tversky loss function for automated, multilevel image classification in leukemia
diagnostics. The model was designed to tackle binary classification for distinguishing
*Corresponding author: normal from abnormal cells, and multiclass classification for identifying leukemia
Kumari Pritee
(preetik@iimsambalpur.ac.in) subtypes, while addressing the challenges of imbalanced datasets inherent in
medical imaging. Trained on publicly available leukemia image datasets, the CNN
Citation: Pritee K, Garg RD. achieved high accuracy in both tasks, effectively capturing subtle morphological
Optimized convolutional neural
network model for multilevel variations critical for precise diagnosis. By incorporating performance metrics such
classification in leukemia diagnosis as accuracy, precision, and recall, the study highlights the model’s reliability and
using Tversky loss. Artif Intell robustness across classification tasks. The findings underscore the potential of CNN-
Health. 2025;2(3):63-76.
doi: 10.36922/aih.4710 based tools in enhancing diagnostic accuracy and efficiency, paving the way for future
innovations in leukemia diagnostics and broader medical imaging applications.
Received: August 30, 2024
1st revised: September 27, 2024
Keywords: Multilevel classification; Deep learning; Leukemia; Convolutional neural
2nd revised: October 28, 2024 networks; Medical image analysis; Automated diagnosis
3rd revised: December 10, 2024
4th revised: December 31, 2024
5th revised: January 3, 2025 1. Introduction
Accepted: January 8, 2025 1.1. Background and motivation
Published online: January 22, Leukemia, a type of cancer affecting blood and bone marrow, requires timely and
2025
accurate diagnosis for effective treatment. However, traditional diagnostic methods such
Copyright: © 2025 Author(s). as microscopic examination are time-consuming, labor-intensive, and prone to human
This is an Open-Access article
distributed under the terms of the error. With the increasing volume of medical imaging data, there is a growing need for
Creative Commons Attribution automated diagnostic tools that can enhance both the speed and accuracy of leukemia
License, permitting distribution,
and reproduction in any medium, diagnosis. This paper aims to address these challenges by leveraging deep learning
provided the original work is (DL) techniques, specifically convolutional neural networks (CNNs), for the multilevel
properly cited. classification of leukemia cells, offering a more reliable and efficient diagnostic solution.
Publisher’s Note: AccScience
Publishing remains neutral with 1.2. Contributions
regard to jurisdictional claims in
published maps and institutional The main contributions of this work are threefold. First, we introduce a multilevel
affiliations. classification framework for leukemia diagnosis, which uses CNNs optimized with the

Volume 2 Issue 3 (2025) 63 doi: 10.36922/aih.4710