Details

• We are pleased to announce that two of our lab members have successfully published their research at the 8th International Artificial Intelligence and Data Processing Symposium, held online from September 21-22, 2024, in Malatya, Turkey. This symposium brings together global researchers and professionals to present cutting-edge developments in artificial intelligence and data processing.

Paper 1:

Title: A Novel CAD Framework with Visual and Textual Interpretability: Multimodal Insights for Predicting Respiratory Diseases
Authors: Raza Mukhlis, Saied Saleem, Hyunwook Kwon, Jamil Hussain, Ahmet Arif Aydın, Dr. Mugahed A. Al-antari (Corresponding)

This research introduces an innovative Computer-Aided Diagnosis (CAD) system that integrates both visual and textual interpretations for more accurate predictions of respiratory diseases from Chest X-ray (CXR) images. The system bridges the gap in diagnostic communication by embedding detailed radiology reports from the MIMIC-CXR dataset into a Chroma Vector Database, enhancing disease classification across 13 respiratory pathologies.
Key features of this system include:

• Multimodal disease prediction using pre-trained AI models.
• Text retrieval optimization for clinical details, using techniques like rank fusion.
• Detailed report generation employs LLAVA and LLAMA3.1 large language models (LLMs).

This work significantly improves the interpretability of AI predictions, making it easier for healthcare providers and patients to understand diagnostic findings.

Paper 2:

Title: A Novel AI-based Hybrid Ensemble Segmentation CAD System for Lumbar Spine Stenosis Pathological Regions Using MRI Axial Images
Authors: Saied Saleem, Raza Mukhlis, Oguzhan Katar, Bilal Ertuğrul, Ozal Yildirim, Dr. Mugahed A. Al-antari (Corresponding)

This paper presents a hybrid ensemble deep learning CAD system designed to segment pathological regions in Lumbar Spine Stenosis using MRI axial images. The system combines the strengths of Vision Transformers (ViT) and Convolutional Neural Networks (CNN) to improve the accuracy of segmentation in degenerative spinal diseases.
The research involved:

• Two ensemble methods: Feature space fusion and stacking techniques.
• Validation on a private MRI dataset, demonstrating the robustness of the proposed model.
• International collaboration through a joint project between NRF Korea and TUBITAK Turkey.

The paper highlights how advanced AI techniques can provide more accurate diagnostics for lumbar spinal stenosis, aiding in early detection and treatment.

Acknowledgements:

Both papers were supported by the National Research Foundation of Korea (NRF), under grants RS-2022-00166402 and RS-2023-00256517, and by TUBITAK (The Scientific and Technological Research Council of Turkey) under grant number 123N325. We sincerely thank NRF and TUBITAK for their generous support in making this research possible.