Risk Assessment of Patients algorithm attempts to extract ICD-10 codes from patient records and use these codes to compute the risk assessment of the patients.
In this session, we will see the idea to combine the Bio+Clinical BERT transformers models and fine-tune that with Knowledge Graph generated by meta map tool over the UMLS dataset for the MedNLI task.
We are able to download the MedNLI dataset and then ran multiple runs of the MedNLI dataset with various different hyperparameters for Bio+Clinical BERT, ClinicalBERT models, Bio+ ClinicalBERT fine-tuned with discharge summary.