Non-invasive HIV Viral Load Prediction Based on Liver Quantitative Ultrasound

Human immune-deficient virus (HIV) remains a major public health problem facing humanity and affects lots of people worldwide. Since the advent of highly active antiretroviral therapy, the survival rate of acquired immune deficiency syndrome (AIDS) patients has improved, and AIDS prognosis has received more attention. HIV viral load (HVL) is a key determinant in the assessment of AIDS prognosis, but traditional assessment methods are invasive, and related operators have a high risk of iatrogenic infection. Based on the correlation between the levels of HVL and the degree of liver injury, this paper proposes to use quantitative ultrasound (QUS) to determine the degree of liver injury and predict the HVL levels non-invasively. QUS features were extracted from ultrasound radio-frequency (RF) data and envelope signals of the liver. After feature selection, classification was performed using the random forest. With accuracy (ACC), precision (PRE), sensitivity (SEN), F1-score, specificity (SPE) and area under the receiver operating characteristic curve (AUC) as the evaluation indexes, the proposed method performs well with most evaluation indexes higher than 80%, SPE of 92.83% and AUC of 90.42%. The results show that the proposed method may be a promising tool for the non-invasive HVL prediction to reduce the risk of iatrogenic infection, with certain clinical significance.