Subcutaneous Vein Recognition System Using Deep Learning for Intravenous (IV) Access Procedure

Intravenous(IV) access is animportant daily clinical procedure that delivers fluids or medication into a patient’s vein. However, IV insertion very challenging where clinicians aresuffering in locating the subcutaneous vein due to patients’physiological factorssuch as hairy forearm and thick dermis fat, alsomedical staff’s level of fatigue. To resolve this issue, researchers have proposed autonomous machines be usedfor access,but such equipmentarelackingcapabilityin detecting accurately. Therefore, project proposes an automatic veindetection algorithm using deep learning for IVaccess purpose. U-Net, fully connected network (FCN) architecture employed thisproject its capability near-infrared (NIR)subcutaneous Data augmentation applied toincrease dataset size reduce bias from overfitting. The original U-Net optimized by replacing up-sampling with transpose convolution well theadditional implementation ofbatch normalizationbesides reducing number oflayers diminish risk overfitting.After fine-tuning retraining hypermodel, unsupervised used evaluate hypermodel selecting 10 checkpoints foreach image comparing onpredicted outputs determine true positive pixels. lightweight has achieved slightly lower accuracy (0.8871) than architecture. Even so, sensitivity, specificity, precisionare greatly improved byachieving 0.7806, 0.9935,and 0.9918 respectively. This result indicates can venipuncture machine toaccurately locate intravenous (IV)procedures.