Students Mohammad Abu Tame’, researcher in the Master’s program in Data Science and Business Analysis, and Sari Al-Masry, researcher in the Master’s program in Artificial Intelligence at the Faculty of Graduate Studies at the Arab American University, have published a peer-reviewed research article in the Information - MDPI Journal with an impact factor of 3.1 titled “Transformer-Based Approach to Pathology Diagnosis Using Audio Spectrogram”

The research article was part of the machine learning course, under the supervision of Dr. Ahmed Al-Hassasneh, and in cooperation with researchers at the Ecole Supérieure de Technologie (ETS) at the University of Quebec in Montreal - Canada.

The research article included proposing an advanced deep machine learning methodology that analyzes and diagnoses the crying of infants in order to early identify diseases they may suffer from. The proposed methodology was based on the use of audio spectrograms without the need for reprocessing and deep algorithms based on transformers to classify crying infants into specific disease categories (septicaemia and respiratory distress syndrome).

The study achieved high classification results, reaching 98.69%, exceeding the classification results of traditional machine learning models and deep machine learning models such as convolutional neural networks. This research is an effective, new, scalable diagnostic tool that opens up ways to improve newborn care through early and accurate detection of diseases.

More information about the published research, click here