Introduction

Capsule Endoscopy (CE) is a minimally invasive technique that allows for detailed visualization of the small bowel mucosa. Anemia and gastrointestinal (GI) bleeding are among the primary indications for CE. Accurate detection of vascular lesions is critical in this context; however, CE interpretation is time-consuming and subject to human error. Artificial intelligence (AI) has the potential to assist in this task, yet real-world studies assessing its clinical impact remain limited.

Aims & Methods

This study aimed to evaluate the performance of a deep learning convolutional neural network in detecting and identifying vascular lesions—specifically red spots and angiectasias—in complete CE videos.
We conducted a multicenter, prospective validation study comparing AI-assisted CE reading with standard, conventional interpretation. A total of 211 CE videos, acquired using three different devices across seven centers in Portugal, Spain, Brazil, and the United States, were included. Following initial standard readings, an independent expert performed AI-assisted readings using a deep learning model designed to detect and classify small bowel vascular lesions. Both reports were reviewed by a third-party expert from an independent center who adjudicated discrepant cases to establish the reference standard. Diagnostic performance was evaluated using standard metrics and the area under the receiver operating characteristic curve (AUC-ROC).

Results

AI-assisted reading identified 241 confirmed vascular lesions, compared to 158 detected through conventional reading. Sensitivity was significantly higher for AI-assisted reading (96.4% vs. 67.7%), though specificity was lower (75.7% vs. 93.5%). Overall accuracy was also higher with AI (89.7% vs. 76.1%). The AUC-ROC was 0.93 for AI-assisted and 0.78 for standard reading. Moreover, AI-assisted reading reduced the mean examination time to 316 seconds per video.

Conclusion

AI-assisted interpretation of full-length CE videos significantly outperforms conventional reading in the detection and classification of vascular lesions. The inclusion of videos from multiple CE systems addresses interoperability challenges, and the involvement of patients from four countries across two continents supports the model's generalizability and real-world clinical applicability.

Introduction

Hepatic decompensation in patients with cirrhosis represents a critical point in the clinical course of the disease, associated with a significant increase in morbidity and mortality. ¹ Common clinical manifestations of decompensation include ascites, hepatic encephalopathy, and variceal bleeding.^2​ Cystatin C, a low molecular weight cysteine protease inhibitor produced by all nucleated cells, has emerged as a reliable endogenous marker of glomerular filtration rate (GFR). Unlike serum creatinine, cystatin C is not significantly influenced by age, sex, muscle mass, or dietary protein intake, making it a more precise indicator of renal function—particularly in populations with altered muscle mass, such as those with chronic liver disease.³ Beyond its role in renal assessment, cystatin C has demonstrated prognostic value in various clinical settings, including cardiovascular disease, sepsis, and general critical illness^​4,5; however, patients with liver cirrhosis have been underrepresented in these cohorts.

Aims & Methods

Prospective longitudinal study including 62 patients with confirmed Child-Pugh class A or B cirrhosis. Serum cystatin C levels were measured at baseline by ELISA (mg/L), using <0.95 mg/L as the normal cutoff. Patients were followed for 6 months to detect decompensation events. Clinical and biochemical variables (age, sex, creatinine, albumin) were analyzed using quartile stratification, linear trend tests, Pearson’s chi-squared test, and diagnostic performance metrics (odds ratio, sensitivity, specificity, likelihood ratios).

Results

Table 1. Clinical and Biochemical Characteristics According to Hepatic Decompensation During Follow-Up.

Variable	Total (n = 62)	With decompensation (n =27)	Without decompensation (n = 35)
Age (mean ± SD, years)	58.05 ± 11.4	59.1 ± 10.9	57.2 ± 11.9
Male sex, n (%)	32 (51.6%)	14 (43%)	18(56%)
Serum cystatin C (mg/L, mean ± SD)	1.07 ± 0.42	1.25 ± 0.40	0.90 ± 0.40
Elevated cystatin C (>0.95 mg/L), n (%)	35 (56.4)	22 (81.5)	20 (57.1)
Serum albumin (g/dL, mean ± SD)	3.1 ± 0.5	2.8 ± 0.5	3.4 ± 0.4
Creatinine/cystatin C ratio (mean ± SD)	8.7 ± 5.8	7.6 ± 5.1	9.7 ± 6.1
Main etiology: Alcoholic liver disease, n (%)	26 (41.9)	12 (44.4)	14 (40.0)

Of 62 patients (32 men, 51.6%; mean age 58.05 ± 11.4 years), major etiologies were alcohol use (41.9%), metabolic liver disease (38.7%), and autoimmune disease (19.3%). Patients with infectious or undefined etiologies were excluded.
During follow-up, 27 patients (43.5%) developed at least one decompensation event (ascites, encephalopathy, or variceal bleeding). Among them, 81.5% had elevated cystatin C (mean 1.25 ± 0.40 mg/L) versus non-decompensated (mean 0.9 ± 0.40 mg/L, p<0.001). Chi-square analysis showed a significant association (χ² = 4.132; p = 0.042), supported by likelihood ratio (χ² = 4.293; p = 0.038) and linear-by-linear association (χ² = 4.065; p = 0.044). The estimated OR was 3.3.
Decompensated patients had a lower creatinine/cystatin ratio (mean 7.6 ± 5.1 vs 9.7 ± 6.1), though not statistically significant (p=0.162). Albumin levels were lower (2.8 ± 0.5 g/dL vs 3.4 ± 0.4 g/dL, p<0.001).
Quartile analysis showed a higher risk of decompensation in the highest cystatin C quartile (>1.25 mg/L; p<0.001, linear trend), even after excluding patients with eGFR <60 mL/min/1.73 m². This trend was confirmed by linear-by-linear association (χ² = 39.88; df = 1; p < 0.001).

Conclusion

Cystatin C is significantly associated with the development of decompensation events in patients with chronic liver disease, providing an OR of 3.3 and a consistent dose-response trend across quartiles. Although the final sample size was relatively small due to exclusions for loss to follow-up or incomplete biochemical studies, the results were statistically significant and clinically relevant. These findings, based on a prospective design and strict follow-up, suggest that cystatin C could be a prognostic marker, and its inclusion in clinical practice warrants prospective validation.

References

1. Zhai M, Long J, Liu S, Liu C, Li L, Yang L, Li Y, Shu B. The burden of liver cirrhosis and underlying etiologies: results from the global burden of disease study 2017. Aging (Albany NY). 2021 Jan 12;13(1):279–300. doi: 10.18632/aging.104127.
2. D'Amico G, Bernardi M, Angeli P. Towards a new definition of decompensated cirrhosis. J Hepatol. 2022 Jan;76(1):202–207. doi: 10.1016/j.jhep.2021.06.018.​
3. Chen DC, Potok OA, Rifkin D, Estrella MM. Ventajas, limitaciones y consideraciones clínicas en el uso de la cistatina C para estimar la TFG. Riñón360. 2022;3(10):1807-14.
4. Wang Y, Li W, Yang J, Zhang M, Tian C, Ma M, Zhang Q. Association between cystatin C and the risk of ischemic stroke: a systematic review and meta-analysis. J Mol Neurosci. 2019;69(3):444–449. doi:10.1007/s12031-019-01373-1.
5. Luo J, Wang LP, Hu HF, Zhang L, Li YL, Ai LM, et al. Cystatin C and cardiovascular or all-cause mortality risk in the general population: a meta-analysis. Clin Chim Acta. 2015 Oct 23;450:39–45. doi: 10.1016/j.cca.2015.07.016.

Disclosure

I declare that I have no conflicts of interest related to this work.