About Tuberculosis and the Widespread Concern

Tuberculosis is a type of infectious disease. It’s commonly also known simply as TB. The disease is caused by a specific bacterium, known as Mycobacterium tuberculosis. It’s generally considered an infectious disease of the lungs, as this is the primary area of the human body it infects. However, there are cases where tuberculosis also affects other body parts.

When it comes to the management of tuberculosis, being able to identify the symptoms, especially early on, is crucial. This can help patients seek medical treatment before the infection progresses and becomes severe.

The specific symptoms depend on the region of the patient’s body where the infection is growing. It is most commonly found to infect the lungs, making some of the crucial symptoms to look out for [1]:

• A cough that lasts for several weeks
• Bringing up phlegm or blood with coughs
• Pain located in the chest

These are symptoms that tend to appear in many cases as they relate to an infection of the lungs. In addition to coughing and chest pain, other symptoms related to TB include:

• Weight loss
• A loss of appetite
• Nightsweats
• Fever
• Chills
• Weakness
• Fatigue

Another major concern with this infectious disease is inactive tuberculosis. Not every person will start to show signs of TB when they are infected. Some people have inactive TB, but if it goes untreated, it can develop into active tuberculosis.

Worldwide, tuberculosis claimed the lives of 1.25 million people in 2023 alone [2]. In that same year, about 10.8 million people contracted this infectious disease.

Tuberculosis Strains: The Key to Successful Treatment

Not every case of tuberculosis is caused by the same strain of the bacteria behind this infectious disease. There are different strains, and they are often first categorized based on the location of the body they infect.

The two main categories of tuberculosis strains include:

• Pulmonary Tuberculosis: This is the most common kind of tuberculosis and affects the lungs.
• Extrapulmonary Tuberculosis: When tuberculosis affects the kidneys, larynx, brain, lymph nodes, joints, or bones, it is categorized as extrapulmonary tuberculosis [3].

There’s also a category called miliary tuberculosis, which is a particularly severe form that spreads through the patient’s body. It is linked to serious complications, including meningitis.

Strains are also categorized by drug resistance, as this has become a key point in scientific studies. The specific categorization depends on whether the strains respond to current treatment or have shown resistance.

These are the main strain categorizations based on their drug resistance:

• Mono-resistant
• Poly-resistant
• Multidrug-resistant
• pre-XDR-TB
• XDR-TB

Current Challenges in Drug Resistance For Tuberculosis

One of the most significant challenges the healthcare industry faces in regards to tuberculosis right now is drug resistance. This continues to be a burden and makes it harder for medical experts to further reduce the mortality rate associated with this infectious disease.

Research shows that the most pressing matter right now is multidrug-resistant strains [4]. These strains are harder to address due to the fact that they show resistance to more than two drugs. In 2021, about 450,000 new cases of multidrug-resistant tuberculosis were reported.

This poses a major threat, as treating a patient with multidrug-resistant tuberculosis can be a difficult process. Standard drugs may not be efficient in treating the tuberculosis infection, which often leads to trials with several different medications. This delays the physician’s ability to help effectively treat a patient’s case of tuberculosis, which also creates an opportunity for the infection to become more serious.

When patients present with multidrug-resistant TB, it requires a more complex treatment regimen that also lasts significantly longer. This can sometimes mean the patient has to stay on a treatment program for as long as 24 months.

There are a couple of major concerns that arise in these situations. It becomes expensive for patients to continue getting their monthly refill of the medication over a period of two years. Additionally, long-term use of these drugs also greatly increases the risk of patients experiencing side effects. Medical experts have also talked about the potential increased risk for toxicity from using these drugs over such a long period of time, as they are generally indicated for a shorter-term treatment.

Another challenge is the complexity of diagnosis in cases where tuberculosis bacterium strains are resistant to multiple drugs. Standard diagnostic tools are often not sufficient in identifying these strains and helping physicians determine the most appropriate treatment regimen for the patient. Issues like a lack of access to the advanced diagnostic tools needed for an accurate diagnosis remain a problem in widespread areas [5].

Machine Learning Predicts Drug Resistance in Tuberculosis Strains

As machine learning continues to advance and become more integrated in the medical industry, researchers have started to use this, alongside artificial intelligence, more efficiently in recent years. According to studies from the past couple of months, there lies potential in using machine learning to predict drug resistance in tuberculosis strains.

While this research is still ongoing, it has already demonstrated significant potential in advancing the diagnosis and treatment of tuberculosis.

One interesting research paper [6] provides an overview of how advanced machine learning models were used for this purpose. A total of 10 input classes were used, and the researchers utilized a variety of metrics to rate the efficiency of the models. The tests were given to a total of three different machine learning models.

Some of the factors that were considered to rate these models included:

• Specificity
• Sensitivity
• F1 score
• Receiver Operating Characteristic curve (ROC)
• Area Under Curve (AUC)

The accuracy in detecting and predicting drug resistance in tuberculosis cases was considered significant for all three of these advanced machine learning models. However, one model had deep learning integrated, and it was able to efficiently outperform the other models used in the study.

Apart from the deep neural network, other algorithms included:

• XGBoost (Extreme Gradient Boosting)
• LightGBM (Light Gradient Boosting)

There are other research papers that have considered the use of machine learning to assist in predicting drug resistance in tuberculosis as well. The results of studies continue to seem very promising, which means there is definitely a future for AI and ML models in the advancements against tuberculosis.

One particular study [7] used gene-specific machine learning models. They found that these models were very successful and accurate in predicting resistance to TB drugs. It was a study that provided a proof-of-concept, which means further research is still needed. However, this advancement in machine learning for tuberculosis already shows how drug resistance can be detected early on.

Future Directions

Currently, there is already a significant body of evidence that supports the use of machine learning to help detect and predict potential drug resistance in tuberculosis strains. The research is still considered to be at an early stage. This means there is still a lot of work and research to be done on this topic.

Looking at the future, we will likely see studies start to fine-tune their machine learning models even more. This would help to further enhance the accuracy of these predictions.

When this kind of prediction system is rolled out into real-world situations, it can help physicians identify resistance against tuberculosis drugs before a treatment regimen is prescribed to a patient. There are several potential advantages that this could hold, including a reduced need for long-term treatment that can sometimes last for two years.

The long-term side effects of tuberculosis drugs would also be reduced, as patients may find their symptoms improve faster when a more efficient drug is chosen as a first-line treatment. This would be made possible by knowing which drugs (or drugs) a tuberculosis strain is likely to be resistant to from the beginning of the treatment program. In addition to reducing side effects, it would also result in greater affordability and reduced toxicity associated with the use of these medications to treat TB in patients.

We’ll also likely see researchers turn to other tuberculosis strains, such as cases of miliary tuberculosis. This rare disease is currently a major threat to individuals affected by TB. Researchers could also use machine learning to make predictions and potentially even choose a better treatment for these patients from the start.

Conclusion

While treatments for tuberculosis have advanced significantly in the 21st century, we are still seeing this disease claim lives, especially in developing countries. Drug resistance in tuberculosis strains is one major issue that the medical industry has faced, but with machine learning continuing to make progress, researchers are now turning to this technology. Emerging research shows taht machine learning could be the key to help predict drug resistance in tuberculosis strains, thus allowing medical experts to make critical decisions beforehand.

References

[1] Signs and Symptoms of Tuberculosis. CDC. 17 Jan 2025. https://www.cdc.gov/tb/signs-symptoms/index.html

[2] Tuberculosis. World Health Organization. 14 Mar 2025. https://www.who.int/news-room/fact-sheets/detail/tuberculosis

[3] J.Y. Lee. Diagnosis and Treatment of Extrapulmonary Tuberculosis. Tuberculosis & Respiratory Diseases. 2 Apr 2015. https://pmc.ncbi.nlm.nih.gov/articles/PMC4388900/

[4] V. Srivastava. Current Challenges in the Management of Tuberculosis. Journal of Young Pharmacists. 3 Jun 2024. https://archives.jyoungpharm.org/7857/

[5] S. Dommati, P. Gottimukkula, R. Patil. Multidrug-Resistant and Rifampicin-Resistant Tuberculosis: Challenges and Advances in Diagnosis and Management. International Journal of Pharmaceutical Sciences. https://www.ijpsjournal.com/article

[6] N. Sirri, C. Guyeux, C. Sola. Advanced Machine Learning for Predicting Drug Resistance in Clinical Isolates of Mycobacterium Tuberculosis Complex. 2024 World Conference on Complex Systems (WCCS). https://ieeexplore.ieee.org/document/10765528

[7] A.T. Subalakshmi, A. Mahesh. Machine learning approaches to predict drug resistance in tuberculosis. Computational Biology and Chemistry. https://www.sciencedirect.com/science/article/abs/pii/S1476927125003664