Prof. Behnam Kiani Kalejahi from the Engineering School of Central Asian University has published a new research article in the prestigious Springer Q1 journal “BMC Medical Imaging”. The study presents an advanced artificial intelligence framework for the diagnosis of papillary thyroid carcinoma by integrating ultrasound imaging and fine-needle aspiration cytology through a multimodal deep learning architecture.
The research introduces a comprehensive deep learning framework that combines state-of-the-art neural network architectures with multimodal data fusion techniques to improve the accuracy and reliability of thyroid cancer diagnosis. By integrating a domain-pretrained pathology encoder, advanced multiple-instance learning, and bidirectional cross-attention mechanisms, the proposed model effectively captures complementary diagnostic information from both ultrasound images and cytological examinations, resulting in more robust and clinically meaningful predictions.
A significant contribution of the study is its emphasis on developing trustworthy clinical artificial intelligence systems through rigorous model evaluation and probability calibration. In addition to achieving excellent diagnostic performance, the proposed framework demonstrates improved confidence estimation and decision reliability — two essential requirements for the safe adoption of AI in clinical practice. Experimental validation on a publicly available multimodal thyroid cancer dataset demonstrated an AUROC of 0.977, while statistically outperforming several conventional machine learning approaches under a rigorous evaluation protocol.
Beyond proposing a high-performing diagnostic model, the research highlights the importance of evaluating medical AI systems using comprehensive performance measures that include calibration, reproducibility, and clinical decision reliability rather than relying solely on conventional accuracy metrics. These findings contribute to the development of more transparent, dependable, and clinically applicable artificial intelligence technologies for medical image analysis and computer-aided diagnosis.
This publication reflects the ongoing commitment of the Engineering School at Central Asian University to conducting high-impact research in artificial intelligence, medical imaging, and digital healthcare. The achievement further strengthens the University's growing international research presence and its contribution to advancing innovative AI technologies for next-generation healthcare solutions.
