Thumbnails:
List:
Year:
Category:
Session:
Poster:
Getting poster data...
Jayashree Vijaya Raghavan, Shruthi KS, Vinod Kumar Dorai, Rebecca Diya Samuel, Priyanka Arunachalam, H.C. Chaluvanarayana, Pavan Belahalli, Kalpana S.R., Siddharth Jhunjhunwala (3rd Floor, BSSE, Biological Sciences Building, Indian Institute of Science, Bengaluru, Karnataka, India - 560012)
Effective combination of clinical information with biochemical and immunological data assists clinicians in making better prognosis. However, analyzing multi-parameter data is not facile. We use unsupervised and supervised learning techniques to analyze biochemical and immune cell data collected by us on diabetic foot ulcer (DFU) patients. Unsupervised learning methods such as PCA, tSNE and hierarchical clustering failed to capture the differences between healed and non-healed ulcers. Hence, we resorted to use various supervised learning methods such as logistic regression (LR), Support Vector Machine (SVM), k-Nearest Neighbours and Linear Discriminant Analyses to classify patients based on healing status. Feature selection techniques were used to build the model and model with best performance was chosen upon performing 10-fold cross validation. We observed that LR and SVM showed best performance with accuracy of prediction 76.9 % and 88.6% respectively. In conclusion, we show that measuring 5 specific parameters in DFU patients could guide the clinician in determining which wounds will heal. Currently, we are establishing a new dataset to validate the performance of the model.