An effective and accurate identification system of Mycobacterium tuberculosis using convolution neural networks

Pulmonary Tuberculosis Diagnosis Using an Intelligent Microscopy Scanner and Image Recognition Model for Improved Acid-Fast Bacilli Detection in Smears

Microscopic examination of acid-fast mycobacterial bacilli (AFB) in sputum smears remains the most economical and readily available method for laboratory diagnosis of pulmonary tuberculosis (TB). However, this conventional approach is low in sensitivity and labor-intensive. An automated microscopy system incorporating artificial intelligence and machine learning for AFB identification was evaluated. The study was conducted at an infectious disease hospital in Jiangsu Province, China, utilizing an intelligent microscope system. A total of 1000 sputum smears were included in the study, with the system capturing digital microscopic images and employing an image recognition model to automatically identify and classify AFBs. Referee technicians served as the gold standard for discrepant results. The automated system demonstrated an overall accuracy of 96.70% (967/1000), sensitivity of 91.94% (194/211), specificity of 97.97% (773/789), and negative predictive value (NPV) of 97.85% (773/790) at a prevalence of 21.1% (211/1000). Incorporating AI and machine learning into an automated microscopy system demonstrated the potential to enhance the sensitivity and efficiency of AFB detection in sputum smears compared to conventional manual microscopy. This approach holds promise for widespread application in TB diagnostics and potentially other fields requiring labor-intensive microscopic examination.

The Impact of Artificial Intelligence on Microbial Diagnosis

Traditional microbial diagnostic methods face many obstacles such as sample handling, culture difficulties, misidentification, and delays in determining susceptibility. The advent of artificial intelligence (AI) has markedly transformed microbial diagnostics with rapid and precise analyses. Nonetheless, ethical considerations accompany AI adoption, necessitating measures to uphold patient privacy, mitigate biases, and ensure data integrity. This review examines conventional diagnostic hurdles, stressing the significance of standardized procedures in sample processing. It underscores AI's significant impact, particularly through machine learning (ML), in microbial diagnostics. Recent progressions in AI, particularly ML methodologies, are explored, showcasing their influence on microbial categorization, comprehension of microorganism interactions, and augmentation of microscopy capabilities. This review furnishes a comprehensive evaluation of AI's utility in microbial diagnostics, addressing both advantages and challenges. A few case studies including SARS-CoV-2, malaria, and mycobacteria serve to illustrate AI's potential for swift and precise diagnosis. Utilization of convolutional neural networks (CNNs) in digital pathology, automated bacterial classification, and colony counting further underscores AI's versatility. Additionally, AI improves antimicrobial susceptibility assessment and contributes to disease surveillance, outbreak forecasting, and real-time monitoring. Despite a few limitations, integration of AI in diagnostic microbiology presents robust solutions, user-friendly algorithms, and comprehensive training, promising paradigm-shifting advancements in healthcare.

Automatic bright-field smear microscopy for diagnosis of pulmonary tuberculosis

In recent decades, many studies have been published on the use of automatic smear microscopy for diagnosing pulmonary tuberculosis (TB). Most of them deal with a preliminary step of the diagnosis, the bacilli detection, whereas sputum smear microscopy for diagnosis of pulmonary TB comprises detecting and reporting the number of bacilli found in at least 100 microscopic fields, according to the 5 grading scales (negative, scanty, 1+, 2+ and 3+) endorsed by the World Health Organization (WHO). Pulmonary TB diagnosis in bright-field smear microscopy, however, depends upon the attention of a trained and motivated technician, while the automated TB diagnosis requires little or no interpretation by a technician. As far as we know, this work proposes the first automatic method for pulmonary TB diagnosis in bright-field smear microscopy, according to the WHO recommendations. The proposed method comprises a semantic segmentation step, using a deep neural network, followed by a filtering step aiming to reduce the number of false positives (false bacilli): color and shape filtering. In semantic segmentation, different configurations of encoders are evaluated, using depth-wise separable convolution layers and channel attention mechanism. The proposed method was evaluated with a large, robust, and annotated image dataset designed for this purpose, consisting of 250 testing sets, 50 sets for each of the 5 TB diagnostic classes. The following performance metrics were obtained for automatic pulmonary TB diagnosis by smear microscopy: mean precision of 0.894, mean recall of 0.896, and mean F1-score of 0.895.

Harnessing of Artificial Intelligence for the Diagnosis and Prevention of Hospital-Acquired Infections: A Systematic Review

Healthcare-associated infections (HAIs) are the most common adverse events in healthcare and constitute a major global public health concern. Surveillance represents the foundation for the effective prevention and control of HAIs, yet conventional surveillance is costly and labor intensive. Artificial intelligence (AI) and machine learning (ML) have the potential to support the development of HAI surveillance algorithms for the understanding of HAI risk factors, the improvement of patient risk stratification as well as the prediction and timely detection and prevention of infections. AI-supported systems have so far been explored for clinical laboratory testing and imaging diagnosis, antimicrobial resistance profiling, antibiotic discovery and prediction-based clinical decision support tools in terms of HAIs. This review aims to provide a comprehensive summary of the current literature on AI applications in the field of HAIs and discuss the future potentials of this emerging technology in infection practice. Following the PRISMA guidelines, this study examined the articles in databases including PubMed and Scopus until November 2023, which were screened based on the inclusion and exclusion criteria, resulting in 162 included articles. By elucidating the advancements in the field, we aim to highlight the potential applications of AI in the field, report related issues and shortcomings and discuss the future directions.

Explainable deep-neural-network supported scheme for tuberculosis detection from chest radiographs

Chest radiographs are examined in typical clinical settings by competent physicians for tuberculosis diagnosis. However, this procedure is time consuming and subjective. Due to the growing usage of machine learning techniques in applied sciences, researchers have begun applying comparable concepts to medical diagnostics, such as tuberculosis screening. In the period of extremely deep neural nets which comprised of hundreds of convolution layers for feature extraction, we create a shallow-CNN for screening of TB condition from Chest X-rays so that the model is able to offer appropriate interpretation for right diagnosis. The suggested model consists of four convolution-maxpooling layers with various hyperparameters that were optimized for optimal performance using a Bayesian optimization technique. The model was reported with a peak classification accuracy, F1-score, sensitivity and specificity of 0.95. In addition, the receiver operating characteristic (ROC) curve for the proposed shallow-CNN showed a peak area under the curve value of 0.976. Moreover, we have employed class activation maps (CAM) and Local Interpretable Model-agnostic Explanations (LIME), explainer systems for assessing the transparency and explainability of the model in comparison to a state-of-the-art pre-trained neural net such as the DenseNet.

Performance Evaluation of No-Code Artificial Intelligence Models for the Detection of Acid-Fast Bacilli: A Comparative Analysis of Three Models

Background Acid-fast bacilli Mycobacterium tuberculosis and Mycobacterium leprae are the causative organisms behind two major diseases of developing nations, tuberculosis and leprosy, respectively. To efficiently tackle these diseases in developing nations, drugs must be augmented with improved detection modalities. This necessitates the development of enhanced tools that can aid the current detection modalities being used in high-incidence areas. A no-code artificial intelligence model based on image classification is one such tool that can be used in the identification of acid-fast bacilli. This study utilizes three such no-code artificial intelligence models that originate from three different platforms but share identical training, testing, and subsequent evaluation. Thereafter, the study is directed at comparing the three models created and identifying the one that can function as a promising support system for the detection of acid-fast bacilli. Methods To begin with, a total of 1000 images per class, i.e., positive and negative for each disease, were captured from the diagnosed slides of tuberculosis and leprosy, taken from the Department of Pathology. Subsequently, these slides were reviewed again by a pathologist to demarcate them as positive or negative for acid-fast bacilli. Once the required number of images was captured, 600 images of each class were selected as the training set, 300 images as the testing set, and the remaining 100 images as the evaluation set. Data augmentation was then performed using techniques such as rotating, mirroring, cropping, and position shifting. These designated data sets were then used to train the image classification software available on the following three platforms: Lobe (Microsoft Corporation, Redmond, Washington, United States), Create ML (Apple Inc., Cupertino, California, United States), Python-based open-source software (PerceptiLabs, Stockholm, Sweden). The final evaluation was based on different parameters such as sensitivity, specificity, ease of use, learning curve, technological resources required, and feasibility of implementation. All parameters put together served the purpose of comparison to identify the most promising model. Results Out of the three models tested, the one built using Lobe is the most promising in terms of the evaluation parameters considered. For tuberculosis, the sensitivity and specificity values obtained were 96% each, while for leprosy, they were 100% and 96%, respectively. Also, the model built using Lobe had a near-negligible learning curve, in addition to being the most cost-effective and feasible model to implement. Furthermore, it had a unique real-time training feature, which constantly improved the model throughout the testing period, till the final sensitivity and specificity values were achieved. Conclusions In clinical situations where a high number of cases are encountered each day, a no-code artificial intelligence model built using Lobe would get exposed to a huge database, getting trained in real time. Subsequently, such a model would reach considerable levels of sensitivity and specificity and in turn, act as a promising support system for the detection of acid-fast bacilli.

Clinical screening of Nocardia in sputum smears based on neural networks

Objective

Nocardia is clinically rare but highly pathogenic in clinical practice. Due to the lack of Nocardia screening methods, Nocardia is often missed in diagnosis, leading to worsening the condition. Therefore, this paper proposes a Nocardia screening method based on neural networks, aiming at quick Nocardia detection in sputum specimens with low costs and thereby reducing the missed diagnosis rate.

Methods

Firstly, sputum specimens were collected from patients who were infected with Nocardia, and a part of the specimens were mixed with new sputum specimens from patients without Nocardia infection to enhance the data diversity. Secondly, the specimens were converted into smears with Gram staining. Images were captured under a microscope and subsequently annotated by experts, creating two datasets. Thirdly, each dataset was divided into three subsets: the training set, the validation set and the test set. The training and validation sets were used for training networks, while the test set was used for evaluating the effeteness of the trained networks. Finally, a neural network model was trained on this dataset, with an image of Gram-stained sputum smear as input, this model determines the presence and locations of Nocardia instances within the image.

Results

After training, the detection network was evaluated on two datasets, resulting in classification accuracies of 97.3% and 98.3%, respectively. This network can identify Nocardia instances in about 24 milliseconds per image on a personal computer. The detection metrics of mAP50 on both datasets were 0.780 and 0.841, respectively.

Conclusion

The Nocardia screening method can accurately and efficiently determine whether Nocardia exists in the images of Gram-stained sputum smears. Additionally, it can precisely locate the Nocardia instances, assisting doctors in confirming the presence of Nocardia.

Distinguishing infectivity in patients with pulmonary tuberculosis using deep learning

Introduction

This study aimed to develop and assess a deep-learning model based on CT images for distinguishing infectivity in patients with pulmonary tuberculosis (PTB).

Methods

We labeled all 925 patients from four centers with weak and strong infectivity based on multiple sputum smears within a month for our deep-learning model named TBINet's training. We compared TBINet's performance in identifying infectious patients to that of the conventional 3D ResNet model. For model explainability, we used gradient-weighted class activation mapping (Grad-CAM) technology to identify the site of lesion activation in the CT images.

Results

The TBINet model demonstrated superior performance with an area under the curve (AUC) of 0.819 and 0.753 on the validation and external test sets, respectively, compared to existing deep learning methods. Furthermore, using Grad-CAM, we observed that CT images with higher levels of consolidation, voids, upper lobe involvement, and enlarged lymph nodes were more likely to come from patients with highly infectious forms of PTB.

Conclusion

Our study proves the feasibility of using CT images to identify the infectivity of PTB patients based on the deep learning method.

Multilayer outperforms single-layer slide scanning in AI-based classification of whole slide images with low-burden acid-fast mycobacteria (AFB)

Manual screening of Ziehl-Neelsen (ZN)-stained slides that are negative or contain rare acid-fast mycobacteria (AFB) is labor-intensive and requires repetitive refocusing to visualize AFB candidates under the microscope. Whole slide image (WSI) scanners have enabled implementation of AI to classify digital ZN-stained slides as AFB+ or AFB-. By default, these scanners acquire a single-layer WSI. However, some scanners can acquire a multilayer WSI with a z-stack and an extended focus image layer embedded. We developed a parameterized WSI classification pipeline to assess whether multilayer imaging improves ZN-stained slide classification accuracy. A CNN built into the pipeline classified tiles in each image layer to form an AFB probability score heatmap. Features extracted from the heatmap were then entered into a WSI classifier. 46 AFB+ and 88 AFB- single-layer WSIs were used for the classifier training. 15 AFB+ (with rare microorganisms) and 5 AFB- multilayer WSIs comprised the test set. Parameters in the pipeline included: (a) a WSI representation: z-stack of image layers, middle image layer (a single image layer equivalent) or an extended focus image layer, (b) 4 methods of aggregating AFB probability scores across the z-stack, (c) 3 classifiers, (d) 3 AFB probability thresholds, and (e) 9 feature vector types extracted from the aggregated AFB probability heatmaps. Balanced accuracy (BACC) was used to measure the pipeline performance for all parameter combinations. Analysis of Covariance (ANCOVA) was used to statistically evaluate the effect of each parameter on the BACC. After adjusting for other factors, a significant effect of the WSI representation (p-value < 1.99E-76), classifier type (p-value < 1.73E-21), and AFB threshold (p-value = 0.003) was observed on the BACC. The feature type had no significant effect (p-value = 0.459) on the BACC. WSIs represented by the middle layer, extended focus layer and the z-stack followed by the weighted averaging of AFB probability scores were classified with the average BACC of 58.80%, 68.64%, and 77.28%, respectively. The multilayer WSIs represented by the z-stack with the weighted averaging of AFB probability scores were classified by a Random Forest classifier with the average BACC of 83.32%. Low classification accuracy of WSIs represented by the middle layer suggests that they contain fewer features permitting identification of AFB than the multilayer WSIs. Our results indicate that single-layer acquisition can introduce a bias (sampling error) into the WSI. This bias can be mitigated by the multilayer or the extended focus acquisitions.

A systematic review and repeatability study on the use of deep learning for classifying and detecting tuberculosis bacilli in microscopic images

Tuberculosis (TB) is among the leading causes of death worldwide from a single infectious agent. This disease usually affects the lungs (pulmonary TB) and can be cured in most cases with a quick diagnosis and proper treatment. Microscopic sputum smear is widely used to diagnose and manage pulmonary TB. Despite being relatively fast and low cost, it can be exhausting because it depends on manually counting TB bacilli (Mycobacterium tuberculosis) in microscope images. In this context, different Deep Learning (DL) techniques are proposed in the literature to assist in performing smear microscopy. This article presents a systematic review based on the PRISMA procedure, which investigates which DL techniques can contribute to classifying TB bacilli in microscopic images of sputum smears using the Ziehl-Nielsen method. After an extensive search and a careful inclusion/exclusion procedure, 28 papers were selected from a total of 400 papers retrieved from nine databases. Based on these articles, the DL techniques are presented as possible solutions to improve smear microscopy. The main concepts necessary to understand how such techniques are proposed and used are also presented. In addition, replication work is also carried out, verifying reproducibility and comparing different works in the literature. In this review, we look at how DL techniques can be a partner to make sputum smear microscopy faster and more efficient. We also identify some gaps in the literature that can guide which issues can be addressed in other works to contribute to the practical use of these methods in laboratories.

Artificial intelligence-based tools applied to pathological diagnosis of microbiological diseases

Infectious diseases still threaten the global community, especially in resource-limited countries. An accurate diagnosis is paramount to proper patient and public health management. Identification of many microbes still relies on manual microscopic examination, a time-consuming process requiring skilled staff. Thus, artificial intelligence (AI) has been exploited for identification of microorganisms. A systematic search was carried out using electronic databases looking for studies dealing with the application of AI to pathology microbiology specimens. Of 4596 retrieved articles, 110 were included. The main applications of AI regarded malaria (54 studies), bacteria (28), nematodes (14), and other protozoa (11). Most publications examined cytological material (95, 86%), mainly analyzing images acquired through microscope cameras (65, 59%) or coupled with smartphones (16, 15%). Various deep-learning strategies were used for the analysis of digital images, achieving highly satisfactory results. The published evidence suggests that AI can be reliably utilized for assisting pathologists in the detection of microorganisms. Further technologic improvement and availability of datasets for training AI-based algorithms would help expand this field and widen its adoption, especially for developing countries.

Artificial intelligence in anatomical pathology: building a strong foundation for precision medicine

With the convergence of digital pathology (DP) and artificial intelligence (AI), anatomic pathology practice has been experiencing an exciting paradigm shifting. Pathologists will be provided with an augmented ability to improve diagnostic accuracy, efficiency, and consistency. There will be subvisual morphometric features discovered and multiomics data integrated to provide better prognostic and theragnostic information to guide individual patients' management. The perspective for future precision medicine is promising. However, there are many challenges before AI-assisted DP diagnostic workflows can be successfully implemented. Herein, we briefly review some examples of AI application in anatomic pathology with an emphasis on the subspecialty of gastrointestinal pathology and discuss potential challenges for clinical implementation.

Review and Updates on the Diagnosis of Tuberculosis

Diagnosis of tuberculosis, and especially the diagnosis of extrapulmonary tuberculosis, still faces challenges in clinical practice. There are several reasons for this. Methods based on the detection of Mycobacterium tuberculosis (Mtb) are insufficiently sensitive, methods based on the detection of Mtb-specific immune responses cannot always differentiate active disease from latent infection, and some of the serological markers of infection with Mtb are insufficiently specific to differentiate tuberculosis from other inflammatory diseases. New tools based on technologies such as flow cytometry, mass spectrometry, high-throughput sequencing, and artificial intelligence have the potential to solve this dilemma. The aim of this review was to provide an updated overview of current efforts to optimize classical diagnostic methods, as well as new molecular and other methodologies, for accurate diagnosis of patients with Mtb infection.

A New Artificial Intelligence-Based Method for Identifying Mycobacterium Tuberculosis in Ziehl–Neelsen Stain on Tissue

Mycobacteria identification is crucial to diagnose tuberculosis. Since the bacillus is very small, finding it in Ziehl–Neelsen (ZN)-stained slides is a long task requiring significant pathologist’s effort. We developed an automated (AI-based) method of identification of mycobacteria. We prepared a training dataset of over 260,000 positive and over 700,000,000 negative patches annotated on scans of 510 whole slide images (WSI) of ZN-stained slides (110 positive and 400 negative). Several image augmentation techniques coupled with different custom computer vision architectures were used. WSIs automatic analysis was followed by a report indicating areas more likely to present mycobacteria. Our model performs AI-based diagnosis (the final decision of the diagnosis of WSI belongs to the pathologist). The results were validated internally on a dataset of 286,000 patches and tested in pathology laboratory settings on 60 ZN slides (23 positive and 37 negative). We compared the pathologists’ results obtained by separately evaluating slides and WSIs with the results given by a pathologist aided by automatic analysis of WSIs. Our architecture showed 0.977 area under the receiver operating characteristic curve. The clinical test presented 98.33% accuracy, 95.65% sensitivity, and 100% specificity for the AI-assisted method, outperforming any other AI-based proposed methods for AFB detection.

Scaphoid Fracture Detection by Using Convolutional Neural Network

Scaphoid fractures frequently appear in injury radiograph, but approximately 20% are occult. While there are few studies in the fracture detection of X-ray scaphoid images, their effectiveness is insignificant in detecting the scaphoid fractures. Traditional image processing technology had been applied to segment interesting areas of X-ray images, but it always suffered from the requirements of manual intervention and a large amount of computational time. To date, the models of convolutional neural networks have been widely applied to medical image recognition; thus, this study proposed a two-stage convolutional neural network to detect scaphoid fractures. In the first stage, the scaphoid bone is separated from the X-ray image using the Faster R-CNN network. The second stage uses the ResNet model as the backbone for feature extraction, and uses the feature pyramid network and the convolutional block attention module to develop the detection and classification models for scaphoid fractures. Various metrics such as recall, precision, sensitivity, specificity, accuracy, and the area under the receiver operating characteristic curve (AUC) are used to evaluate our proposed method’s performance. The scaphoid bone detection achieved an accuracy of 99.70%. The results of scaphoid fracture detection with the rotational bounding box revealed a recall of 0.789, precision of 0.894, accuracy of 0.853, sensitivity of 0.789, specificity of 0.90, and AUC of 0.920. The resulting scaphoid fracture classification had the following performances: recall of 0.735, precision of 0.898, accuracy of 0.829, sensitivity of 0.735, specificity of 0.920, and AUC of 0.917. According to the experimental results, we found that the proposed method can provide effective references for measuring scaphoid fractures. It has a high potential to consider the solution of detection of scaphoid fractures. In the future, the integration of images of the anterior–posterior and lateral views of each participant to develop more powerful convolutional neural networks for fracture detection by X-ray radiograph is probably important to research.

Clinical Laboratory Employees' Attitudes Toward Artificial Intelligence

OBJECTIVE

The objective of this study was to determine the attitudes of laboratory personnel toward the application of artificial intelligence (AI) in the laboratory.

METHODS

We surveyed laboratory employees who covered a range of work roles, work environments, and educational levels.

RESULTS

The survey response rate was 42%. Most respondents (79%) indicated that they were at least somewhat familiar with AI. Very few (4%) classified themselves as experts. Contact with AI varied by educational level (P = .005). Respondents believed that AI could help them perform their work by reducing errors (24%) and saving time (16%). The most common concern (27%) was job security (being replaced by AI). The majority (64%) of the respondents expressed support for the development of AI projects in the organization.

CONCLUSIONS

Laboratory employees see the potential for AI and generally support the adoption of AI tools but have concerns regarding job security and quality of AI performance.

Artificial Intelligence-Based Screening for Mycobacteria in Whole-Slide Images of Tissue Samples

OBJECTIVES

This study aimed to develop and validate a deep learning algorithm to screen digitized acid fast-stained (AFS) slides for mycobacteria within tissue sections.

METHODS

A total of 441 whole-slide images (WSIs) of AFS tissue material were used to develop a deep learning algorithm. Regions of interest with possible acid-fast bacilli (AFBs) were displayed in a web-based gallery format alongside corresponding WSIs for pathologist review. Artificial intelligence (AI)-assisted analysis of another 138 AFS slides was compared to manual light microscopy and WSI evaluation without AI support.

RESULTS

Algorithm performance showed an area under the curve of 0.960 at the image patch level. More AI-assisted reviews identified AFBs than manual microscopy or WSI examination (P < .001). Sensitivity, negative predictive value, and accuracy were highest for AI-assisted reviews. AI-assisted reviews also had the highest rate of matching the original sign-out diagnosis, were less time-consuming, and were much easier for pathologists to perform (P < .001).

CONCLUSIONS

This study reports the successful development and clinical validation of an AI-based digital pathology system to screen for AFBs in anatomic pathology material. AI assistance proved to be more sensitive and accurate, took pathologists less time to screen cases, and was easier to use than either manual microscopy or viewing WSIs.

The Development of a Skin Cancer Classification System for Pigmented Skin Lesions Using Deep Learning

Recent studies have demonstrated the usefulness of convolutional neural networks (CNNs) to classify images of melanoma, with accuracies comparable to those achieved by dermatologists. However, the performance of a CNN trained with only clinical images of a pigmented skin lesion in a clinical image classification task, in competition with dermatologists, has not been reported to date. In this study, we extracted 5846 clinical images of pigmented skin lesions from 3551 patients. Pigmented skin lesions included malignant tumors (malignant melanoma and basal cell carcinoma) and benign tumors (nevus, seborrhoeic keratosis, senile lentigo, and hematoma/hemangioma). We created the test dataset by randomly selecting 666 patients out of them and picking one image per patient, and created the training dataset by giving bounding-box annotations to the rest of the images (4732 images, 2885 patients). Subsequently, we trained a faster, region-based CNN (FRCNN) with the training dataset and checked the performance of the model on the test dataset. In addition, ten board-certified dermatologists (BCDs) and ten dermatologic trainees (TRNs) took the same tests, and we compared their diagnostic accuracy with FRCNN. For six-class classification, the accuracy of FRCNN was 86.2%, and that of the BCDs and TRNs was 79.5% (p = 0.0081) and 75.1% (p < 0.00001), respectively. For two-class classification (benign or malignant), the accuracy, sensitivity, and specificity were 91.5%, 83.3%, and 94.5% by FRCNN; 86.6%, 86.3%, and 86.6% by BCD; and 85.3%, 83.5%, and 85.9% by TRN, respectively. False positive rates and positive predictive values were 5.5% and 84.7% by FRCNN, 13.4% and 70.5% by BCD, and 14.1% and 68.5% by TRN, respectively. We compared the classification performance of FRCNN with 20 dermatologists. As a result, the classification accuracy of FRCNN was better than that of the dermatologists. In the future, we plan to implement this system in society and have it used by the general public, in order to improve the prognosis of skin cancer.

A CNN-based active learning framework to identify mycobacteria in digitized Ziehl-Neelsen stained human tissues

Tuberculosis is the most common mycobacterial disease that affects humans worldwide. Rapid and reliable diagnosis of mycobacteria is crucial to identify infected individuals, to initiate and monitor treatment and to minimize or prevent transmission. Microscopic identification of acid-fast mycobacteria (AFB) in tissue sections is usually accomplished by examining Ziehl-Neelsen (ZN) stained slides in which AFB appear bright red against the blue background. Because the ZN-stained slides require time consuming and meticulous screening by an experienced pathologist, our team developed a machine learning pipeline to classify digitized ZN-stained slides as AFB-positive or AFB-negative. The pipeline includes two convolutional neural network (CNN) models to recognize tiles containing AFB, and a logistic regression (LR) model to classify slides based on features from AFB-probability maps assembled from the CNN tile-based classification results. The first CNN was trained using tiles from 6 AFB-positive and 8 AFB-negative slides, and the second CNN was trained using the initial tile set expanded by additional tiles from 19 AFB-negative slides selected within an active learning framework. When evaluated on a separate set of tiles, the two CNNs yielded F1 scores of 99.03% and 98.75%, respectively, and were used to classify tiles in a separate set of 134 slides (46 AFB-positive and 88 AFB-negative). The classification yielded two AFB-probability maps, one for each CNN. The LR model was then 10-fold cross-validated using the average of feature vectors extracted from the AFB-probability maps generated by each CNN. The feature vector consisted of seven features of the AFB-probability map histogram and the positive tile rate (PTR). The sensitivity (87.13%), specificity (87.62%) and F1 (80.18%) achieved by this model were superior to the baseline performance of PTR-based separation of slides that yielded F1 scores of 73.13% and 66.67% in the AFB-probability maps outputted by the CNN trained within the active learning framework and the CNN trained only on the initial set of slides, respectively. Our CNNs outperformed several recently published models for AFB detection. Active learning induced robust learning of features by the CNN and led to improved LR classification performance of slides. In the 52 AFB-positive slides used in the pipeline development, the AFB were infrequent, predominantly single and only rarely found in small clusters. Our pipeline can classify slides and visualize suspected AFB-positive areas in each slide, and thus potentially facilitate evaluation of ZN-stained tissue sections for AFB.

Artificial intelligence and the future of global health

Concurrent advances in information technology infrastructure and mobile computing power in many low and middle-income countries (LMICs) have raised hopes that artificial intelligence (AI) might help to address challenges unique to the field of global health and accelerate achievement of the health-related sustainable development goals. A series of fundamental questions have been raised about AI-driven health interventions, and whether the tools, methods, and protections traditionally used to make ethical and evidence-based decisions about new technologies can be applied to AI. Deployment of AI has already begun for a broad range of health issues common to LMICs, with interventions focused primarily on communicable diseases, including tuberculosis and malaria. Types of AI vary, but most use some form of machine learning or signal processing. Several types of machine learning methods are frequently used together, as is machine learning with other approaches, most often signal processing. AI-driven health interventions fit into four categories relevant to global health researchers: (1) diagnosis, (2) patient morbidity or mortality risk assessment, (3) disease outbreak prediction and surveillance, and (4) health policy and planning. However, much of the AI-driven intervention research in global health does not describe ethical, regulatory, or practical considerations required for widespread use or deployment at scale. Despite the field remaining nascent, AI-driven health interventions could lead to improved health outcomes in LMICs. Although some challenges of developing and deploying these interventions might not be unique to these settings, the global health community will need to work quickly to establish guidelines for development, testing, and use, and develop a user-driven research agenda to facilitate equitable and ethical use.