Algorithm 1: Pneumonia Status Prediction by Data Preprocessing and Feature Vectorization using Multi-Modal Data Analysis

Input:  DRaw and IRaw, Vital signs record and CXR images of the patients.      DLabelled and ILabelled, Labelling performed by the doctors.      RInference, Inference rules with threshold specified by the doctors.      Threshold, Determines pneumonia status prediction limit. Output:  NNConf, NN model predicted confidence.       Alert, MDA-PSP system alert for discharge or no discharge within 7 days. Initialize (candidateSet 1, candidateSet 2, NNConf, CandidateFinal) = ∅ DProcessedIM, IProcessedIM = Imputation (DRaw, IRaw) DProcessedCG, IProcessedCG = Categorization (DProcessedIM, IProcessedIM) DProcessedVS = Adaptive Imputation (DProcessedCG) ⋃ DLabelled IProcessedIP = Grey-Scale (Resize (IProcessedCG)) ⋃ ILabelled RInference = Pneumonia ˄ General Ward ˄ Clinical Checkup ˄ Comorbidity If DProcessedVS is Consistent and Score (RInference) ≥ Threshold then    Feature VectorsVS, Feature VectorsIP = CNN (DProcessedVS, IProcessedIP)      candidate Set 1, candidate Set 2 = Feature VectorsVS, Feature VectorsIP else    Print “Inconsistent Data”    break candidateFinal = candidate Set 1 ⋃ candidate Set 2 NNConf = Dense-BN(CandidateFinal) If NNConf ≥ Threshold then    Alert “Discharge” else    Alert “No Discharge” Return NNConf, Alert