The COVID-19 pandemic has deeply affected the respiratory health of people, leaving many sufferers with long term pulmonary problems. Artificial intelligence based physiological analysis of structured exercise program on lung function of recovered COVID 19 patient is studied. The research introduces an integrated data driven approach for assessing the improvement of respiratory through physical training. The approach is to integrate wearable sensor technology with machine learning algorithms. A controlled experimental study with three groups (recovered COVID-19 patients, smokers, healthy individuals) was used as a method. To that aim, each of the participants underwent an eight-week structured aerobic training program that included continuous monitoring through wearable devices of key physiological metrics, namely oxygen saturation, heart rate, respiratory rate and lactic acid levels. Trends were analyzed using machine learning models such as Random Forest and Long Short-Term Memory (LSTM) networks and used in the prediction of individual recovery progress. The trained recovered COVID-19 patients showed statistically significant improvement in lung function demonstrated by an average 5% increase in oxygen saturation and significant reduction in lactic acid. Further, the predictive models confirmed that participants who became more adapted to aerobic exercises prior to the respiratory virus had a higher probability of long-term respiratory recovery. This can serve as an indication of the potential of AI-driven personalized rehabilitation programs for increased efficacy of respiratory therapy. The role of artificial intelligence in rehabilitation sciences demonstrated by this research provides a new and transformative way for artificial intelligent algorithms to be tailored to individual patient’s physiological responses in real time. Future work will scale this approach to more patients and with more precise recovery prediction using advanced deep learning.
