Face recognition is one of the techniques used to identify and classify individuals that has distinctive features of their face for automated healthcare monitoring. Intelligent face analysis is indispensable in the medical field for improving patient safety, rapidly assessing patient status, and live monitoring. However, many existing methods rely on conventional facial analysis or basic classifiers, which limits recognition accuracy and reduces performance under varying facial expressions and real-time conditions. To overcome these restrictions, a Convolutional Neural Network with MediaPipe Face Mesh (CNN-MPFM) model is proposed. Initially, the input patient facial images are preprocessed using face detection, resizing, and normalization. The output of the processed facial images is then passed to the MediaPipe Face Mesh module, which provides more detailed facial information with precise facial landmark coordinates and structural facial information. CNN model is an intelligent classification to determine if patient is in normal or suffocated condition, given the prominent features that were retrieved. The suggested CNN-MPFM framework is evaluated with accuracy, precision, recall, F1- score, loss and classification efficiency. The experimental analysis results show that the accuracy, precision, recall, and F1-score of the suggested CNN-MPFM method are 99.52%, 99.46%, 99.49%, and 99.47% respectively. Furthermore, the proposed CNN-MPFM approach enhances the classification accuracy over the previous deep transfer learning, ConvNet, MobileNet-V1 and multi-modal recognition approaches.