Radiology: Artificial intelligence for automatic detection, diagnosis and severity assessment of tuberculosis

Tuberculosis (TB) is an airborne infectious disease caused by mycobacteria
.

Although TB-related morbidity and mortality continue to decline globally at this stage , the burden of the disease remains high in countries where TB is endemic .

Tuberculosis (TB) is an airborne infectious disease caused by mycobacteria

In recent years, artificial intelligence (AI) has gained a great deal of attention, proposing many applications in medical image recognition and parsing .
Deep learning (DL), as the core technology of artificial intelligence application, has made great progress in medical image analysis] .
These DL algorithms can autonomously "learn" to predict features from human-classified initial data .

A study published in the journal Radiology developed and evaluated an artificial intelligence-based fully automated CT image analysis model to provide imaging support for the detection, diagnosis, and quantification of disease severity in pulmonary tuberculosis patients .

A study published in the journal Radiology developed and evaluated an artificial intelligence-based fully automated CT image analysis model to provide imaging support for the detection, diagnosis, and quantification of disease severity in pulmonary tuberculosis patients .

From December 2007 to September 2020, we retrospectively included chest CT scans of 892 patients with pathogen - confirmed tuberculosis .
A deep learning based cascade framework is connected to create processing channels .
To train and validate the model, 1921 lesions were manually labeled and classified according to six key imaging features, and the involvement of lesions was visually scored as ground truth .
A "TB score" was calculated from the network activation map to quantitatively assess disease burden .
Independent test datasets from two other hospitals (Dataset 2, n = 99; Dataset 3, n = 86) and NIH Tuberculosis (n = 171) were used to externally validate the performance of the AI ​​model .
 

This study analyzed CT scan images of 526 participants (mean age, 48.

5±16.
5 years; 206 women) .
The lung lesion detection subsystem yielded an average precision of 0.
68 in the validation cohort .
Six lung key imaging results from independent datasets indicated an overall TB classification accuracy of 81.
08-91.
05% .

This study analyzed CT scan images of 526 participants (mean age, 48.
5±16.
5 years; 206 women) .
The lung lesion detection subsystem yielded an average precision of 0.
68 in the validation cohort .

 

Figure example of chest CT images of a pulmonary tuberculosis patient and the performance of the AI ​​model 

Figure example of chest CT image of pulmonary tuberculosis patient and performance of AI model  Figure of example of chest CT image of pulmonary tuberculosis patient and performance of AI model 

In conclusion, the DL cascade model based on chest CT images can be used clinically for accurate detection, diagnosis and triage of pulmonary tuberculosis
.

The fully automated artificial intelligence system proposed in this study has great potential in clinical practice to rapidly assess the active type of lesions and guide the treatment and management of pulmonary tuberculosis .

In conclusion, the DL cascade model based on chest CT images can be used clinically for accurate detection, diagnosis and triage of pulmonary tuberculosis

Original source :

Chenggong Yan , Lingfeng Wang , Jie Lin ,et al .

A fully automatic artificial intelligence-based CT image analysis system for accurate detection, diagnosis, and quantitative severity evaluation of pulmonary tuberculosis .
DOI : 10.
1007/s00330-021-08365-z

Chenggong Yan Lingfeng Wang Jie Lin ,et al A fully automatic artificial intelligence-based CT image analysis system for accurate detection, diagnosis, and quantitative severity evaluation of pulmonary tuberculosis 10.
1007/s00330-021-08365-z 10.
1007/s00330-021-08365- z leave a message here