Pneumonia is a form of acute lower respiratory infection. It is characterized by the deposition of fluid and/or pus in the alveoli, limiting oxygen exchange and making breathing painful and difficult.
Pneumonia is usually caused by viruses or bacteria and rarely by other microorganisms. It is the leading cause of death, especially in children below the age of 5 and adults over 65.1 There are over 30 different causes of pneumonia and are grouped as bacterial pneumonia, viral pneumonia, fungal pneumonia, and other pneumonia. They can also be categorized as community-acquired, hospital-acquired, or aspiration pneumonia.2
Pneumonia affects approximately 450 million people worldwide and leads to 4 million deaths each year. India accounts for over 23% of the global pneumonia burden. With over 30 million new cases of childhood pneumonia reported each year, India tops the list of countries with a high disease burden.3
Pneumonia is a difficult disease to diagnose. It could be caused by many pathogens that could lead to bacterial, viral, or fungal infection in the lungs and can be contracted anywhere, including the hospitals.
Until recently, chest X-rays were considered the best tool for detecting all types of pneumonia; however, they may not be perfect. Scans could appear similar for different conditions, and imaging makes it difficult to identify the pathogens causing the infections. This makes it challenging to diagnose pneumonia via X-rays. Also, this is more evident when the patients experience several health problems simultaneously.
Physicians must gather a clear and complete picture of the patient’s condition to come up with an accurate diagnosis of their problems. Although a complete blood work could be arranged for, in addition to an X-ray, the value of such results may be of limited value. The results could be indicative of pre-existing conditions and not necessarily pneumonia. In such cases, the use of Artificial intelligence to support clinical decisions seems to be a rather promising concept.4
Artificial intelligence appears to be a powerful tool that can aid in diagnosing and managing pneumonia by mimicking and supplementing the clinical-decision making of physicians. Computational performance using machine learning (ML) and neural network (NN) models allow for extensive mathematical operations and streamlining of high quantity data quickly, which prove to be extremely useful during clinical practice. The availability of data from electronic health records (EHR) and advances constantly being made in computational performance are two major factors that highlight the connection between AI and regular medical practice.
With the constantly increasing data and variables related to patient care, and the constant up-gradation of medical literature, these models verify and confirm all data retrieved from electronic health records and facilitate optimal improvements in patient care. Treatment strategies could also be strategized and optimized based on results obtained from clinical situations like using multi-drug therapies, antibiotic resistance patterns, potential risks associated with treatment protocols, risks related to readmissions, etc.5
The information on the use of AI in improving pneumonia management is limited. AI has greater application potential in medical imaging in cases where the data is easier to obtain. A handful of studies highlight the detailed data about which AI method was used to diagnose and manage pneumonia.
Chest radiographs are the most common diagnostic imaging tools at present. They are still used to screen, diagnose and manage pneumonia and other related diseases. In the coming years, AI systems could be more than handy and technologically useful for physicians, nurses, and physician assistants in the emergency department. The AI tool needs to be fed with several patients' radiograph files, and the process needs to be fine-tuned. Also, when compared with manually coming up with a diagnosis, the AI tool is faster compared to even board-certified radiologists.6
Use of AI in decision making process for empiric antibiotics
Antibiotic resistance is a massive public health burden and is of utmost priority to hospitals worldwide. Huge amounts of data, prescription patterns, and how patients respond to the treatment need to be gathered and analyzed at all times. AI would come in handy and pose as an important tool to automate this process of auditing the usage of antibiotics to ensure optimal use to improve patient care and reduce mortality. AI in pneumonia management could help identify the treatment profiles of patients and narrow it down to the cases that require immediate attention. It could suggest the appropriate antibiotics for each case and provide optimum guidance to clinicians at the point of prescription. Also, speed is of the essence for the success rate. While manual processes could take several minutes, AI can assess the same data in seconds, helping healthcare providers provide empiric antibiotic therapy for effective antibiotic resistance cases.10
Artificial intelligence can certainly be a powerful ally in the fight against the spread of pneumonia. AI can give doctors and other healthcare professionals a leg up when tracking and understanding the spread of pneumonia. AI can also help diagnose pneumonia and assist doctors to develop a better understanding of how to treat and manage the illness better.
1. Chumbita, M., Cillóniz, C., Puerta-Alcalde, P., Moreno-García, E., Sanjuan, G., Garcia-Pouton, N., Soriano, A., Torres, A. and Garcia-Vidal, C., 2022. Can Artificial Intelligence Improve the Management of Pneumonia.
2. Kwon, T., Lee, S., Kim, D., et al., 2021. Diagnostic performance of artificial intelligence model for pneumonia from chest radiography. PLOS ONE, 16(4), p.e0249399.
3. Farooqui, H., Jit, M., Heymann, D. and Zodpey, S., 2015. Burden of Severe Pneumonia, Pneumococcal Pneumonia and Pneumonia Deaths in Indian States: Modelling Based Estimates. PLOS ONE, 10(6), p.e0129191.
4. Nhlbi.nih.gov. 2022. Pneumonia | NHLBI, NIH. [online] Available at:<https://www.nhlbi.nih.gov/health/pneumonia>
5. Chumbita, M., Cillóniz, C., Puerta-Alcalde, P., Moreno-García, E., Sanjuan, G., Garcia-Pouton, N., Soriano, A., Torres, A. and Garcia-Vidal, C., 2022. Can Artificial Intelligence Improve the Management of Pneumonia
6. Stokes, K., Castaldo, R., Federici, C., Pagliara, S., Maccaro, A., Cappuccio, F., Fico, G., Salvatore, M., Franzese, M. and Pecchia, L., 2022. The use of artificial intelligence systems in diagnosis of pneumonia via signs and symptoms: A systematic review. Biomedical Signal Processing and Control, 72, p.103325.
7. Kermany, D., Goldbaum, M., Cai, W., et al., 2018. Identifying Medical Diagnoses and Treatable Diseases by Image-Based Deep Learning. Cell, 172(5), pp.1122-1131.e9.
8. Stephen, O., Sain, M., Maduh, U. and Jeong, D., 2019. An Efficient Deep Learning Approach to Pneumonia Classification in Healthcare. Journal of Healthcare Engineering, 2019, pp.1-7.
9. Heckerling, P., Gerber, B., Tape, T. and Wigton, R., 2003. Prediction of Community-Acquired Pneumonia Using Artificial Neural Networks. Medical Decision Making, 23(2), pp.112-121.
10. Zhang, M., Yu, S., Yin, X., Zeng, X., Liu, X., Shen, Z., Zhang, X., Huang, C. and Wang, R., 2021. An AI-based auxiliary empirical antibiotic therapy model for children with bacterial pneumonia using low-dose chest CT images. Japanese Journal of Radiology, 39(10), pp.973-983.