Fatty Liver Disease

Description

Fatty liver disease (FLD) is characterized by the accumulation of triglyceride lipids within the liver cells, leading to an abnormal increase in liver size. A normal liver typically contains about 5% lipid content, but with the progression of the disease, this can escalate to as much as 50% of the liver's mass.

The development of fatty liver disease involves multiple contributing factors. One significant factor is alcohol consumption. Excessive intake of alcohol disrupts the normal process of alcohol detoxification in the liver, resulting in the accumulation of alcohol metabolites and impairment of liver cell function. This, in turn, promotes the retention of fatty acids within the liver cells, contributing to the development of the disease. It is important to note that alcohol consumption thresholds of 20 g/day for women and 30 g/day for men are typically considered in relation to alcoholic fatty liver disease.

Another contributing factor to fatty liver disease is metabolic syndrome, which encompasses multiple conditions including obesity, insulin resistance, type 2 diabetes, dyslipidemia, and hypertension. In individuals with metabolic syndrome, factors such as insulin resistance play a central role. Insulin resistance hampers the normal response of cells to insulin and increases the release of free fatty acids from adipose tissue. These fatty acids are then taken up by the liver, resulting in the accumulation of triglycerides within liver cells.

Furthermore, dietary factors can contribute to the development of fatty liver disease. High consumption of fructose, often found in sweetened beverages or high fructose corn syrup, has been associated with a specific form of fatty liver disease known as non-alcoholic fatty liver disease (NAFLD). Excess fructose intake contributes to the storage of excessive fat in the liver cells.

Participants

10,349 adults (18+ years of age) recruited at a hospital health clinic in China.

Additional Data Collection Procedure

Subjects were also asked whether they have been diagnosed with FLD.

Modeling Approach

Blood flow signal was extracted and processed from facial video, and then blood flow features were extracted from blood flow signal.

Feature selection was carried out on blood flow and demographic features to identify features predictive of FLD status. A machine-learning based classifier was then created to predict an individual’s FLD status based on those features.

The model was created/trained using 8,185 subjects. The distribution of sex and FLD status in this group is summarized in Table 21.

Subjects in videos	Composition
Sex distribution	51.7 % Male; 48.3% Female
FLD distribution	20%

Table 21: FLD distribution of training set

Model Performance

The fatty liver disease model was then validated for accuracy on an independent portion of the dataset (n=1082) that was not used in training - validation set. Then the final accuracy was obtained on an independent portion of the dataset (n=1082) that was not used in training - test set. The distribution of sex and FLD status in these groups is summarized in Table 22.

Subjects in videos	Composition in Validation C	Composition in Test
Sex distribution	49.1% Male; 50.9 % Female	54.1% Male; 45.9% Female
FLD distribution	20% FLD	20% FLD

Table 22: FLD distribution of validation and test sets

The accuracy on the test set calculated as area under the curve was 89.0% as shown in Figure 26.

Figure 26: AUC of FLD risk prediction

Predictions are displayed to the user as a percentage likelihood of having FLD. A percentage of 45% or more suggests a risk of FLD, while 55% or more suggests high risk of FLD.