New Approaches in the Classification and Prognosis of Sign Clusters on Pulmonary CT Images in Patients With Multidrug-Resistant Tuberculosis

Differences in pulmonary nodular consolidation and pulmonary cavity among drug-sensitive, rifampicin-resistant and multi-drug resistant tuberculosis patients: the Guangzhou computerized tomography study

Background

Pulmonary nodular consolidation (PN) and pulmonary cavity (PC) may represent the two most promising imaging signs in differentiating multidrug-resistant (MDR)-pulmonary tuberculosis (PTB) from drug-sensitive (DS)-PTB. However, there have been concerns that literature described radiological feature differences between DS-PTB and MDR-PTB were confounded by that MDR-PTB cases tend to have a longer history. This study seeks to further clarify this point.

Methods

All cases were from the Guangzhou Chest Hospital, Guangzhou, China. We retrieved data of consecutive new MDR cases [n=46, inclusive of rifampicin-resistant (RR) cases] treated during the period of July 2020 and December 2021, and according to the electronic case archiving system records, the main PTB-related symptoms/signs history was ≤3 months till the first computed tomography (CT) scan in Guangzhou Chest Hospital was taken. To pair the MDR-PTB cases with assumed equal disease history length, we additionally retrieved data of 46 cases of DS-PTB patients. Twenty-two of the DS patients and 30 of the MDR patients were from rural communities. The first CT in Guangzhou Chest Hospital was analysed in this study. When the CT was taken, most cases had anti-TB drug treatment for less than 2 weeks, and none had been treated for more than 3 weeks.

Results

Apparent CT signs associated with chronicity were noted in 10 cases in the DS group (10/46) and 9 cases in the MDR group (10/46). Thus, the overall disease history would have been longer than the assumed <3 months. Still, the history length difference between DS patients and MDR patients in the current study might not be substantial. The lung volume involvement was 11.3%±8.3% for DS cases and 8.4%±6.6% for MDR cases (P=0.022). There was no statistical difference between DS cases and MDR cases both in PN prevalence and in PC prevalence. For positive cases, MDR cases had more PN number (mean of positive cases: 2.63 vs. 2.28, P=0.38) and PC number (mean of positive cases: 2.14 vs. 1.38, P=0.001) than DS cases. Receiver operating characteristic curve analysis shows, PN ≥4 and PC ≥3 had a specificity of 86% (sensitivity 25%) and 93% (sensitivity 36%), respectively, in suggesting the patient being a MDR cases.

Conclusions

A combination of PN and PC features allows statistical separation of DS and MDR cases.

Clinical and Computed Tomography Features Associated with Multidrug-Resistant Pulmonary Tuberculosis: A Retrospective Study in China

Purpose

To explore the value of integrating clinical and computed tomography (CT) features to predict multidrug-resistant pulmonary tuberculosis (MDR-PTB).

Patients and Methods

The study included 212 patients with MDR-PTB and 180 patients with drug-sensitive pulmonary tuberculosis (DS-PTB) who referred to our institute in China between January 2016 and March 2021. The clinical and CT characteristics were analyzed and compared between both groups. Multivariable logistic regression analysis was performed to identify independent factors that can be used to predict MDR-PTB. Furthermore, 115 patients admitted to another center from January 2019 to January 2022 were included as external validation cohort.

Results

For clinical characteristics, five parameters were significantly different between the two groups (all P < 0.05). With regard to CT features, nine parameters were significantly different between the two groups (all P < 0.05). Multivariable logistic regression analysis using the aforementioned differential features showed that male sex, retreated history, longer duration of previous anti-TB treatment, lower CD4⁺ T lymphocyte count, thick-walled cavity, centrilobular micronodules and tree-in-bud sign, bronchial stenosis, pleural and pericardial thickening were the most effective variations associated with MDR-PTB with an area under the curve (AUC) of 0.849 and accuracy of 78.6%. Furthermore, the external validation cohort that contains 115 patients obtained an AUC of 0.933 and accuracy of 81.7%.

Conclusion

MDR-PTB and DS-PTB have different clinical and imaging characteristics. A combined model incorporating these differential features can promptly diagnose MDR-PTB and develop subsequent therapeutic strategies.

Computer-Aided Diagnosis of Spinal Tuberculosis From CT Images Based on Deep Learning With Multimodal Feature Fusion

Background

Spinal tuberculosis (TB) has the highest incidence in remote plateau areas, particularly in Tibet, China, due to inadequate local healthcare services, which not only facilitates the transmission of TB bacteria but also increases the burden on grassroots hospitals. Computer-aided diagnosis (CAD) is urgently required to improve the efficiency of clinical diagnosis of TB using computed tomography (CT) images. However, classical machine learning with handcrafted features generally has low accuracy, and deep learning with self-extracting features relies heavily on the size of medical datasets. Therefore, CAD, which effectively fuses multimodal features, is an alternative solution for spinal TB detection.

Methods

A new deep learning method is proposed that fuses four elaborate image features, specifically three handcrafted features and one convolutional neural network (CNN) feature. Spinal TB CT images were collected from 197 patients with spinal TB, from 2013 to 2020, in the People’s Hospital of Tibet Autonomous Region, China; 3,000 effective lumbar spine CT images were randomly screened to our dataset, from which two sets of 1,500 images each were classified as tuberculosis (positive) and health (negative). In addition, virtual data augmentation is proposed to enlarge the handcrafted features of the TB dataset. Essentially, the proposed multimodal feature fusion CNN consists of four main sections: matching network, backbone (ResNet-18/50, VGG-11/16, DenseNet-121/161), fallen network, and gated information fusion network. Detailed performance analyses were conducted based on the multimodal features, proposed augmentation, model stability, and model-focused heatmap.

Results

Experimental results showed that the proposed model with VGG-11 and virtual data augmentation exhibited optimal performance in terms of accuracy, specificity, sensitivity, and area under curve. In addition, an inverse relationship existed between the model size and test accuracy. The model-focused heatmap also shifted from the irrelevant region to the bone destruction caused by TB.

Conclusion

The proposed augmentation effectively simulated the real data distribution in the feature space. More importantly, all the evaluation metrics and analyses demonstrated that the proposed deep learning model exhibits efficient feature fusion for multimodal features. Our study provides a profound insight into the preliminary auxiliary diagnosis of spinal TB from CT images applicable to the Tibetan area.