FK UI professor Erlina Burhan encourages the use of artificial intelligence for tuberculosis diagnosis

MidLand, Jakarta – In his inauguration speech as Professor of Pneumology and Respiratory Medicine, University of Indonesia (user interface), Erlina Burhan offers tuberculosis (TB) diagnosis using artificial intelligence (AI) as an early diagnosis tool.

The standard diagnosis of tuberculosis currently used involves the use of an approach that aims to detect the presence of the bacterium Mycobacterium tuberculosis (Mtb). Second Erlina Burhan, Diagnosis involves examining the genetic material or observing the growth of bacteria in a medium. However, this standard diagnosis often faces obstacles due to insufficient specimens or inaccurate sputum samples examined.

This problem is usually found when making a diagnosis in children, the elderly, or other conditions who have trouble expelling phlegm. Therefore, it is necessary to adopt an innovative approach, one of which is the use of artificial intelligence.

The algorithm used in artificial intelligence is deep learning AND radiomic. Erlina said it Deep learning Artificial intelligence offers many opportunities for new solutions to eradicate tuberculosis. Artificial intelligence can be used in both the diagnosis and management of tuberculosis.

ON deep learningmost of which use convolutional neural networks/ convolutional neural networks (CNN) consisting of several layers, including input, convolutional, pooling, fully connected, AND production. From this process it is possible to obtain images, speech, genetic sequences and clinical text information.

ON radiomicsinputs in the form of images will be collected, then identified region of interestt (ROI) of the image. These various ROIs will be separated from the non-ROI parts and then extracted from various existing features such as texture, shape, intensity and waves.

The various extracted features will go through a machine learning phase that connects the information to each other and produces output. In the context of using artificial intelligence in the medical field of tuberculosis, the outcome can take the form of diagnosis, treatment and prognosis.

Announcement

Furthermore, the use of artificial intelligence to promote tuberculosis diagnosis can be seen in radiological and microbiological examinations. An example is the research conducted by Xiong Y et al. This research uses a CNN network model called TB-AI which is a support system that can be used to detect TB bacteria.

After examination based on a dual diagnosis confirmed by a pathologist both via microscope and digital slide, TB-AI achieved a sensitivity of 97.94% and a specificity of 83.65%. These figures show that the diagnosis of AI-TB can be considered successful. In radiological examinations, the use of artificial intelligence is necessary due to intra-reader variability and the lack of radiologists in most countries with the highest tuberculosis rate, including Indonesia.

Artificial intelligence for tuberculosis diagnosis has been studied and tested abroad. One way is with a photo of your chest or chest. According to Erlina, chest x-rays are easily accessible, affordable, very sensitive and specific for active pulmonary tuberculosis, and available in primary healthcare.

Some of the AI ​​software used are artificial neural network (ANN), deep learning based automatic detection (DLAD), deep convolutional neural network (D-CNN), CheXaid, InferRead®, Genki™ and CAD4TB™. The use of various AI is believed to increase the efficiency and accuracy of diagnosis.

Tuberculosis It is the second deadliest infectious disease in the world and still represents a burden on global health today. The death rate due to tuberculosis is estimated to reach 1.3 million in 2022. According to Erlina Burhan, early detection with correct diagnosis is one way to eliminate tuberculosis disease in Indonesia.

Editors Choice: Professor Erlina Burhan from FK UI offers GIS to detect active tuberculosis cases in Indonesia