Varicose rupture bleeding is one of the most serious complications of cirrhosis, with a mortality rate of 10 to 20 per cent.
larger varicose veins can be treated with medication or endoscopy to reduce the occurrence of ruptured bleeding, it is recommended that patients with cirrhosis have a routine endoscopy to prevent the occurrence of varicose vein rupture bleeding.
, however, more than half of patients with cirrhosis do not have gastroesoesopal varicose veins, resulting in a lower positive rate for endoscopy.
, it is important to establish a non-invasive way to detect the presence of high-risk varicose veins in patients with cirrhosis.
Studies have shown that the volume of the liver and spleen measured on CT can help identify clinically significant frontal hypertension and liver cirrhosis loss, but the manual segmentation process is cumbersome and hinders its use in clinical practice.
stage, deep learning algorithm can achieve fully automatic liver and spleen volume measurement on CT.
Therefore, if high-risk varicose veins can be detected by using the volume of the liver and spleen alone or in combination with other clinical factors, deep learning-based CT volume measurements can provide additional information without the need for additional time and effort.
Given the widespread use of CT in evaluating patients with cirrhosis, a reliable CT-based standard can identify or exclude high-risk varicose veins, which is useful for selecting the best endoscopic screening patients and for low-risk patients who can safely avoid endoscopic screening.
Recently, a study published in the journal European Radiology developed a noninvasive indicator that combines spleen volume and clinical factors based on CT deep learning analysis to detect high-risk varicose veins, and evaluated ct's usefulness in screening patients with hepatitis B viral cirrhosis for the presence of high-risk varicose veins and predicting their risk of ruptured bleeding.
this retrospective study included 419 patients with viral cirrhosis B who underwent endoscopic and CT examinations from 2007 to 2008 (derivative queue, n s 239) and 2009 to 2010 (validation queue, n s 180).
use deep learning algorithms to measure liver and spleen volume on CT images.
-variable logic regression analysis of the derived queue established the index of high-risk varicose veins to detect endoscopic diagnosis.
the five-year cumulative risk of varicose bleeding by stratumining patients by indicator value.
the spleen volume-plateboard ratio index was designed based on a derived queue.
In the validation queue, the detection sensitivity of the equilibrium sensitivity and specificity truncation index value (> 3.78) was 69.4%, the specificity was 78.5%, and the high sensitivity truncation index value (> 1.63) detected all high-risk varicose veins.
the index straties all patients to a low level (index value≤ 1.63; n s 118), medium (n s 162), high (index value sgt; 3.78; In the n-139) risk group, the cumulative 5-year rates of varicose bleeding in these three groups were 0%, 1.0% and 12.0%, respectively.
and multivariable analysis of clinical and volume parameters for detecting high-risk varicose veins in a table 1 derived queue.
table 2 is strated with the risk of high-risk varicose and varicose bleeding.
CT can be used to detect the detection of high-risk gastroesovascular varicose veins in patients with hepatitis B and to assess the risk of future varicose bleeding.
the ratio of spleen volume to plate plates obtained using deep learning-based CT analysis can identify patients with high-risk varicose veins and very low probability of occurrence to avoid unnecessary endoscopy in this group of patients.