A Joint Model Based on Deep Learning to Predict Multidrug-resistant Klebsiella Pneumoniae Liver Abscess

Liver Abscess

Deep-learning; Liver Abscess

The goal of this observational study is to train a deep learning-based model to predict multidrug-resistant Klebsiella pneumoniae liver abscess and evaluate it on a multi-center database.

Liver abscess is one of the most common abdominal organ infections worldwide, with a mortality rate that once reached as high as 70%. Before the widespread use of antibiotics, suppurative appendicitis was the main causative factor for liver abscesses. In recent years, with the widespread application of antibiotics and the promotion and popularization of interventional therapy, the main causative factor for liver abscesses has gradually shifted from suppurative appendicitis to biliary tract diseases, and the mortality rate of liver abscesses has also been gradually declining. However, due to the increasing number of infections caused by various multidrug-resistant organisms (MDROs), the reported mortality rate currently remains between 2% and 18%. In the treatment of liver abscesses, it is very important to apply antibiotic therapy as early and as quickly as possible after diagnosis. Particularly in China, with its large population and varying medical conditions across different regions, for many primary medical institutions that lack the ability to carry out interventional treatments, antibiotic therapy is the only means of treating such infectious diseases as liver abscesses. However, precise antibiotic therapy relies on the results of bacterial cultures and drug susceptibility tests. Because many patients have received treatment at other medical institutions or have self-administered antibiotics before coming to the hospital, the results of blood cultures are somewhat affected. Currently, the common pathogens of community-acquired liver abscesses are mainly Gram-negative enterobacteria, among which Klebsiella pneumoniae (KP) and Escherichia coli are the majority, with a major shift from Escherichia coli to Klebsiella pneumoniae. In China, Klebsiella pneumoniae has become the primary pathogen of community-acquired liver abscesses. Therefore, empirical treatment regimens typically prioritize the treatment of infections caused by Gram-negative enterobacteria, especially KP. However, infections caused by MDROs often result in poor or directly ineffective treatment due to resistance to empirically chosen antibiotics, thereby delaying treatment, causing greater economic burdens on patients, and consuming more medical resources. Infections caused by MDROs have become a global public health issue of great concern. The increasing occurrence of MDRO infections and the emergence of new types of MDROs pose higher demands on clinical physicians. Under this new trend of pathogenic bacteria, it is unacceptable to rely solely on empirical diagnostic and treatment methods but need efficient and convenient new ways to guide the selection of clinical treatment regimens. Among MDRO infections in liver abscess patients, the largest proportion is caused by extended-spectrum beta-lactamase (ESBL)-producing enterobacteria, mainly ESBL-producing Klebsiella pneumoniae and Escherichia coli. Therefore, if it could be based on general understanding on the most predominant Gram-negative bacteria and explore the clinical manifestations, biochemical indicators, CT images, and other characteristics and differences between liver abscesses caused by Gram-negative enterobacteria and other types of pathogens, the differences between ESBL bacteria and other non-resistant bacteria as well as other MDROs within Gram-negative bacteria would be explored. Ultimately, constructing a predictive model that can identify the types of pathogenic bacteria and the presence of MDROs in the early stages of the disease through basic clinical characteristics, laboratory indicators, and CT images would have significant theoretical and practical value.

No related papers found yet.

None listed

China