From Simple Labels to Descriptive Sentences in Leukemia Image Classification

In many leukemia classification studies, each image is assigned to only one class such as AML, CML, ALL, or CLL. In this work, we used short descriptive sentences instead of simple labels to give more meaningful information to the AI model.

Published in Social Sciences

From Simple Labels to Descriptive Sentences in Leukemia Image Classification
Like

Share this post

Choose a social network to share with, or copy the URL to share elsewhere

This is a representation of how your post may appear on social media. The actual post will vary between social networks

Explore the Research

Springer International Publishing
Springer International Publishing Springer International Publishing

Advancing WBC Classification: A Hybrid ConvNextV2-Swin Transformer Framework with R3GAN Data Balancing and CLAHE Preprocessing - Journal of Imaging Informatics in Medicine

White blood cell (WBC) classification remains a critical challenge in hematological diagnostics, particularly for rare cell types such as basophils and imbalanced datasets. This study introduces a novel three-component hybrid framework that synergistically integrates: (1) ConvNeXtV2-Swin Transformer for dual-scale hierarchical feature extraction—combining ConvNeXtV2’s depthwise convolutions with Swin Transformer’s shifted window attention to capture both local cellular morphology and global contextual dependencies; (2) R3GAN (Reinforced Reliable Robust Generative Adversarial Network) for intelligent minority class augmentation through reinforcement learning-guided sample generation, effectively mitigating class imbalance while preserving biological fidelity; and (3) CLAHE (Contrast-Limited Adaptive Histogram Equalization) for adaptive preprocessing to normalize imaging variations. Evaluated on the challenging Raabin dataset—characterized by severe class imbalance (301 basophils vs. 8887 neutrophils) and limited diversity—the proposed architecture achieves 99.1% accuracy, surpassing state-of-the-art methods by 2–10%. Notably, the framework demonstrates exceptional data efficiency, maintaining 94% accuracy with only 50% training data. The synergistic integration of architectural innovation, intelligent data synthesis, and adaptive preprocessing establishes a robust paradigm for clinical deployment in resource-constrained environments. Source code is publicly available at https://github.com/momenianmohammad/wbc-convnextv2swin-r3gan-eccgan .

The idea of this paper was very simple. In traditional leukemia classification, each cell image is usually assigned to only one class such as AML, CML, ALL, or CLL. I thought this method is limited because a single label cannot describe the full meaning of a medical image.

So, we tried a different idea. Instead of assigning only one class label, we connected each image to a short descriptive sentence.

AML → “Large immature blood cells with abnormal structure.”

CML → “High number of abnormal white blood cells in blood cells.”

ALL → “Fast-growing immature lymphocyte cells in the blood.”

CLL → “Small abnormal lymphocyte cells with slow progression.”

In this work, we used both text-to-image and image-to-text learning. The model tried to understand the relation between medical images and written descriptions at the same time. We also used GAN-based methods to help balance the dataset because some classes had fewer images than others.

One challenge in this project was creating meaningful text descriptions that were simple but still useful for the model. Another challenge was the limited and unbalanced medical dataset. Training vision-language models also needed careful tuning and many experiments.

The model performance was evaluated using the metrics discussed in the paper. The final results were around the 70% range. This is not considered perfect or extremely high accuracy. However, the method is more practical and informative than traditional classification systems that only assign simple labels like 0, 1, 2, or 3.

I think the interesting part of this project was trying to make AI understand medical images in a more human-like way using language. This project showed that combining text and image understanding can open new directions in medical AI research.

In the future, this idea can be improved using larger datasets, better text descriptions, and stronger multimodal models. It may also help doctors better understand AI decisions instead of only seeing simple class outputs.

Link of paper: https://link.springer.com/article/10.1007/s44163-026-01239-7

Please sign in or register for FREE

If you are a registered user on Research Communities by Springer Nature, please sign in

Follow the Topic

Social Science Matters
Humanities and Social Sciences > Social Science Matters

Introducing: Social Science Matters

Social Science Matters is a campaign from the team at Palgrave Macmillan that aims to increase the visibility and impact of the social sciences

Continue reading announcement

Related Collections

With Collections, you can get published faster and increase your visibility.

Transforming Education through Artificial Intelligence: Opportunities, Challenges, and Future Directions

Artificial Intelligence (AI) is rapidly changing the educational field by enabling personalized learning, intelligent tutoring systems, automated assessments, learning analytics, and administrative automation.

This collection invites original research, systematic reviews, and visionary perspectives on the transformative impact of AI in education. It aims to explore how AI technologies can enhance equity, inclusion, and efficiency in educational settings across different contexts, including higher education, K-12, vocational training, and lifelong learning. This collection will address technical, pedagogical, ethical, and policy aspects, fostering interdisciplinary perspectives and evidence-based insights.

This Collection supports and amplifies research related to SDG 4 and SDG 9.

Keywords: Artificial Intelligence, AI in Education, Educational Technology, Data Analytics, AI Ethics

Publishing Model: Open Access

Deadline: May 31, 2026

AI for Image and Video Analysis: Emerging Trends and Applications

The application of AI in image and video analysis has revolutionized a wide range of domains, offering more accurate and efficient visual data processing. Thanks to advances in neural networks, large-scale datasets, and computational power, AI algorithms have surpassed traditional computer vision techniques in performance. This transformation has had a profound impact on areas like healthcare (where AI aids in diagnosing diseases through medical imaging), security (with real-time video surveillance), and entertainment (enhancing video quality and enabling automated content tagging). As AI continues to evolve, new challenges emerge, including the need for explainability, handling large datasets efficiently, improving robustness in real-world environments, and addressing biases in AI models. These open questions necessitate continued research, collaboration, and discourse. The proposed Collection focuses on the intersection of artificial intelligence (AI) and image and video analysis, exploring the latest advancements, challenges, and applications in this rapidly evolving field. As AI-powered techniques such as deep learning, computer vision, and generative models mature, they are increasingly being leveraged for tasks like image classification, object detection, video segmentation, activity recognition, facial recognition, and more. These technologies are pivotal in industries including healthcare, security, autonomous vehicles, entertainment, and smart cities, to name a few. We invite researchers and practitioners to submit articles related to, but not limited to, the following topics:

- Deep learning techniques for image and video analysis

- AI-based object detection and recognition

- Image segmentation and annotation using AI

- Video classification and activity recognition

- Real-time video surveillance and security systems

- AI for medical image analysis and diagnostics

- Generative adversarial networks (GANs) for image and video generation

- AI in autonomous driving and smart transportation systems

- AI-powered multimedia search and retrieval

- Human-Computer Interaction (HCI) through AI-based video analysis

- AI techniques for image and video compression

- Ethical concerns and responsible AI in image and video analysis

This Collection supports and amplifies research related to SDG 9 and SDG 11.

Keywords: computer vision; image segmentation; object detection; video surveillance

Publishing Model: Open Access

Deadline: Sep 15, 2026