Machine Learning Diagnosis of Small-Bowel Crohn Disease Using T2-Weighted MRI Radiomic and Clinical Data.

BACKGROUND. Radiologists have variable diagnostic performance and considerable interreader variability when interpreting MR enterography (MRE) examinations for suspected Crohn disease (CD). OBJECTIVE. The purposes of this study were to develop a machine learning method for predicting ileal CD by use of radiomic features of ileal wall and mesenteric fat from noncontrast T2-weighted MRI and to compare the performance of the method with that of expert radiologists. METHODS. This single-institution study included retrospectively identified patients who underwent MRE for suspected ileal CD from January 1, 2020, to January 31, 2021, and prospectively enrolled participants (patients with newly diagnosed ileal CD or healthy control participants) from December 2018 to October 2021. Using axial T2-weighted SSFSE images, a radiologist selected two slices showing greatest terminal ileal wall thickening. Four ROIs were segmented, and radiomic features were extracted from each ROI. After feature selection, support-vector machine models were trained to classify the presence of ileal CD. Three fellowship-trained pediatric abdominal radiologists independently classified the presence of ileal CD on SSFSE images. The reference standard was clinical diagnosis of ileal CD based on endoscopy and biopsy results. Radiomic-only, clinical-only, and radiomic-clinical ensemble models were trained and evaluated by nested cross-validation. RESULTS. The study included 135 participants (67 female, 68 male; mean age, 15.2 ± 3.2 years); 70 were diagnosed with ileal CD. The three radiologists had accuracies of 83.7% (113/135), 88.1% (119/135), and 86.7% (117/135) for diagnosing CD; consensus accuracy was 88.1%. Interradiologist agreement was substantial (κ = 0.78). The best-performing ROI was bowel core (AUC, 0.95; accuracy, 89.6%); other ROIs had worse performance (whole-bowel AUC, 0.86; fat-core AUC, 0.70; whole-fat AUC, 0.73). For the clinical-only model, AUC was 0.85 and accuracy was 80.0%. The ensemble model combining bowel-core radiomic and clinical models had AUC of 0.98 and accuracy of 93.5%. The bowel-core radiomic-only model had significantly greater accuracy than radiologist 1 (p = .009) and radiologist 2 (p = .02) but not radiologist 3 (p > .99) or the radiologists in consensus (p = .05). The ensemble model had greater accuracy than the radiologists in consensus (p = .02). CONCLUSION. A radiomic machine learning model predicted CD diagnosis with better performance than two of three expert radiologists. Model performance improved when radiomic data were ensembled with clinical data. CLINICAL IMPACT. Deployment of a radiomic-based model including T2-weighted MRI data could decrease interradiologist variability and increase diagnostic accuracy for pediatric CD.

Untargeted metabolomics of perfusate and their association with hypothermic machine perfusion and allograft failure.

Although hypothermic machine perfusion (HMP) is associated with improved kidney graft viability and function, the underlying biological mechanisms are unknown. Untargeted metabolomic profiling may identify potential metabolites and pathways that can help assess allograft viability and contribute to organ preservation. Therefore, in this multicenter study, we measured all detectable metabolites in perfusate collected at the beginning and end of deceased-donor kidney perfusion and evaluated their associations with graft failure. In our cohort of 190 kidney transplants, 33 (17%) had death-censored graft failure over a median follow-up of 5.0 years (IQR 3.0-6.1 years). We identified 553 known metabolites in perfusate and characterized their experimental and biological consistency through duplicate samples and unsupervised clustering. After perfusion-time adjustment and false discovery correction, six metabolites in post-HMP perfusate were significantly associated with death-censored graft failure, including alpha-ketoglutarate, 3-carboxy-4-methyl-5-propyl-2-furanpropanoate, 1-carboxyethylphenylalanine, and three glycerol-phosphatidylcholines. All six metabolites were associated with an increased risk of graft failure (Hazard Ratio per median absolute deviation range 1.04-1.45). Four of six metabolites also demonstrated significant interaction with donation after cardiac death with notably greater risk in the donation after cardiac death group (Hazard Ratios up to 1.69). Discarded kidneys did not have significantly different levels of any death-censored graft failure-associated metabolites. On interrogation of pathway analysis, production of reactive oxygen species and increased metabolism of fatty acids were upregulated in kidneys that subsequently developed death-censored graft failure. Thus, further understanding the role of these metabolites may inform the HMP process and help improve the objective evaluation of allograft offers, thereby reducing the discard of potentially viable organs.

Comparison of proteomic methods in evaluating biomarker-AKI associations in cardiac surgery patients.

Although immunoassays are the most widely used protein measurement method, aptamer-based methods such as the SomaScan platform can quantify up to 7000 proteins per biosample, creating new opportunities for unbiased discovery. However, there is limited research comparing the consistency of biomarker-disease associations between immunoassay and aptamer-based platforms. In a substudy of the TRIBE-AKI cohort, preoperative and postoperative plasma samples from 294 patients with previous immunoassay measurements were analyzed using the SomaScan platform. Inter-platform Spearman correlations (rs) and biomarker-AKI associations were compared across 30 preoperative and 34 postoperative immunoassay-aptamer pairs. Possible factors contributing to inter-platform differences were examined including target protein characteristics, immunoassay, and SomaScan coefficients of variation, other assay characteristics, and sample storage time. The median rs was 0.54 (interquartile range [IQR] 0.34-0.83) in postoperative samples and 0.41 (IQR 0.21-0.69) in preoperative samples. We observed a trend of greater rs in biomarkers with greater concentrations; the Spearman correlation between the concentration of protein and the inter-platform correlation was 0.64 in preoperative pairs and 0.53 in postoperative pairs. Of proteins measured by immunoassays, we observed significant biomarker-AKI associations for 13 proteins preop and 24 postop; of all corresponding aptamers, 8 proteins preop and 12 postop. All proteins significantly associated with AKI as measured by SomaScan were also significantly associated with AKI as measured by immunoassay. All biomarker-AKI odds ratios were significantly different (P < 0.05) between platforms in 14% of aptamer-immunoassay pairs, none of which had high (rs > 0.50) inter-platform correlations. Although similar biomarker-disease associations were observed overall, biomarkers with high physiological concentrations tended to have the highest-confidence inter-platform operability in correlations and biomarker-disease associations. Aptamer assays provide excellent precision and an unprecedented coverage and promise for disease associations but interpretation of results should keep in mind a broad range of correlations with immunoassays.