Analysis of model development strategies: predicting ventral hernia recurrence

Development and Validation of Contrast-Enhanced CT-Based Deep Transfer Learning and Combined Clinical-Radiomics Model to Discriminate Thymomas and Thymic Cysts: A Multicenter Study

We aimed to construct and validate a radiomics-based nomogram model derived from gadoxetic acid-enhanced magnetic resonance (MR) images to predict cytokeratin (CK) 19-positive (+) hepatocellular carcinoma (HCC) and patients' prognosis. A two-center and time-independent cohort of 311 patients were retrospectively enrolled (training cohort, n = 168; internal validation cohort, n = 72; external validation cohort, n = 71). A total of 2286 radiomic features were extracted from multisequence MR images with the uAI Research Portal (uRP), and a radiomic feature model was established. A combined model was established by incorporating the clinic-radiological features and the fusion radiomics signature using logistic regression analysis. Receiver operating characteristic curve (ROC) was used to evaluate the predictive efficacy of these models. Kaplan-Meier survival analysis was used to assess 1-year and 2-year progression-free survival (PFS) and overall survival (OS) in the cohort. By combining radiomic features extracted in DWI phase, arterial phase, venous and delay phase, the fusion radiomics signature achieved AUCs of 0.865, 0.824, and 0.781 in the training, internal, and external validation cohorts. The final combined clinic-radiological model showed higher AUC values in the three datasets compared with the fusion radiomics model. The nomogram based on the combined model showed satisfactory prediction performance in the training (C-index, 0.914), internal (C-index, 0.855), and external validation (C-index, 0.795) cohort. The 1-year and 2-year PFS and OS of the patients in the CK19+ group were 76% and 73%, and 78% and 68%, respectively. The 1-year and 2-year PFS and OS of the patients in the CK19-negative (-) group were 81% and 77%, and 80% and 74%, respectively. Kaplan-Meier survival analysis showed no significant differences in 1-year PFS and OS between the groups (P = 0.273 and 0.290), but it did show differences in 2-year PFS and OS between the groups (P = 0.032 and 0.040). Both PFS and OS were lower in CK19+ patients. The combined model based on clinic-radiological radiomics features can be used for predicting CK19+ HCC noninvasively to assist in the development of personalized treatment.

Radiation-induced acute rectal injury (RARI) is a common early toxicity after radiotherapy for cervical cancer (CC) and remains difficult to predict before treatment, which can adversely affect life quality of patients. We aimed to develop a combined dual-radiomics and deep learning (DL) model to improve the prediction of RARI in CC patients treated with radiotherapy. This retrospective study included 200 CC patients from one hospital, randomly divided into training (n=160), internal validation (n=40) cohorts and external validation (n=40) from another hospital. Patients were classified as RARI (CTCAE v5.0 grade≥2) or Non-RARI (grade<2). Radiomic and dosiomic features were extracted from CT images and dose distributions, and DL features were learned using 3D CNNs. The performance of radiomics, dosiomics, DL and hybrid features models for RARI prediction was compared using the receiver operating characteristic (ROC) curve with measurement of the area under the curve (AUC). For radiomics combining dosiomics, XGBoost achieved the best performance with AUCs of 0.786 and 0.755 in internal and external validation cohorts, respectively. For DL, Resnet_with_CBAM achieved the best performance in the input of combining CT and dose distribution with AUCs of 0.786 and 0.773 in internal and external validation cohorts, respectively. Nomogram integrating radiomics, dosiomics, DL features, and clinical factor improved the AUC to 0.810, 0.803 in internal and external validation cohorts, respectively. The nomogram integrating radiomics, dosiomics, DL, and clinical factors can improve the predictive performance for RARI in CC patients followed by radiotherapy.

Preoperative differentiation of borderline from malignant epithelial ovarian tumors (BEOT from MEOT) can impact surgical management. MRI has improved this assessment but subjective interpretation by radiologists may lead to inconsistent results. To develop and validate an objective MRI-based machine-learning (ML) assessment model for differentiating BEOT from MEOT, and compare the performance against radiologists' interpretation. Retrospective study of eight clinical centers. In all, 501 women with histopathologically-confirmed BEOT (n = 165) or MEOT (n = 336) from 2010 to 2018 were enrolled. Three cohorts were constructed: a training cohort (n = 250), an internal validation cohort (n = 92), and an external validation cohort (n = 159). Preoperative MRI within 2 weeks of surgery. Single- and multiparameter (MP) machine-learning assessment models were built utilizing the following four MRI sequences: T2 -weighted imaging (T2 WI), fat saturation (FS), diffusion-weighted imaging (DWI), apparent diffusion coefficient (ADC), and contrast-enhanced (CE)-T1 WI. Diagnostic performance of the models was assessed for both whole tumor (WT) and solid tumor (ST) components. Assessment of the performance of the model in discriminating BEOT vs. early-stage MEOT was made. Six radiologists of varying experience also interpreted the MR images. Mann-Whitney U-test: significance of the clinical characteristics; chi-square test: difference of label; DeLong test: difference of receiver operating characteristic (ROC). The MP-ST model performed better than the MP-WT model for both the internal validation cohort (area under the curve [AUC] = 0.932 vs. 0.917) and external validation cohort (AUC = 0.902 vs. 0.767). The model showed capability in discriminating BEOT vs. early-stage MEOT, with AUCs of 0.909 and 0.920, respectively. Radiologist performance was considerably poorer than both the internal (mean AUC = 0.792; range, 0.679-0.924) and external (mean AUC = 0.797; range, 0.744-0.867) validation cohorts. Performance of the MRI-based ML model was robust and superior to subjective assessment of radiologists. If our approach can be implemented in clinical practice, improved preoperative prediction could potentially lead to preserved ovarian function and fertility for some women. Level 4. Stage 2. J. Magn. Reson. Imaging 2020;52:897-904.

Periventricular white matter injury (PVWMI) is the most common form of brain injury and the leading cause of cerebral palsy (CP). Early prediction of CP within the first 2 years of life is crucial for timely and effective intervention. Early CP prediction tools for infants with PVWMI are lacking. This study aimed to develop and validate a conventional Magnetic Resonance Imaging (MRI)-based model to predict CP in infants with PVWMI. In this multicentre retrospective cohort study in China, infants with PVWMI who underwent conventional MRI between 6 and 24 months of corrected age (CA) were included from five hospitals and confirmed to have CP or non-CP by 5 years of age. Between April 2013 and September 2018, a multivariable regression logistic model was developed and internally validated using data from one hospital to identify significant independent MRI features associated with CP, followed by external validation across four other hospitals. A visual nomogram was constructed based on these factors. Predictive performance was evaluated via the area under the receiver operating characteristic curve (AUC), calibration curves, and decision curves. Between October 2018 and January 2021, data from one hospital was included in a multiple readers test cohort (nine radiologists and two paediatric neurologists with varying experience) to assess the diagnostic performance and generalisability of the model. Subgroup analyses were conducted by age and sex. Across the two recruitment periods, 383 infants (65% male) with MRI-diagnosed PVWMI were included: 191 infants (122 with CP) in the derivation cohort, 115 (75 with CP) in the external validation cohort, and 77 (46 with CP) in the multiple readers test cohort. Five MRI features were associated with CP: abnormal signals in the posterior limb of the internal capsule (odds ratio [OR] 16.52; 95% confidence interval (CI) 5.78-52.67; P < 0.001), corticospinal tract in centrum semiovale (13.01; 3.49-62.30; P < 0.001), and cerebral peduncle (5.54; 1.20-32.15; P = 0.04), abnormal signals or atrophy in the thalamus (4.76; 1.41-19.32; P = 0.02) and lenticular nucleus (4.58; 1.24-21.35; P = 0.03). The model yielded an AUC of 0.94 (95% CI 0.91-0.98) in the derivation cohort. Similar AUCs were achieved in the internal (0.96 [0.93-0.99]) and external (0.92 [0.86-0.97]) validation cohorts. In the multiple readers test cohort, the average AUC, average sensitivity, and average specificity of 11 readers were 0.96 (95% CI 0.93-0.99), 0.90 (0.84-0.96), and 0.88 (0.77-0.98), respectively. Subgroup analyses were robust, yielding similar AUCs. The conventional MRI-based model showed good performance for predicting CP in infants aged 6-24 months with PVWMI and also had good diagnostic performance and generalisability, which may assist in identifying high-risk infants of CP and facilitating timely interventions. Future work with external validation in diverse countries and socioeconomic contexts are needed. National Natural Science Foundation of China, Key R&D Program of Shanxi Province, National Medical Centre Project of the First Affiliated Hospital of Xi'an Jiaotong University, Henan Provincial Health Commission National Traditional Chinese Medicine Clinical Research Base Scientific Research Special Fund, and Clinical Research Award of the First Affiliated Hospital of Xi'an Jiaotong University.

Radiomics can non-invasively predict the prognosis of a tumour by applying advanced imaging feature algorithms.The aim of this study was to predict the chance of postoperative recurrence by modelling tumour radiomics and peritumour radiomics and clinical features in patients with stage I lung adenocarcinoma (LUAD). Retrospective analysis of 190 patients with postoperative pathologically confirmed stage I LUAD from centre 1, who were divided into training cohort and internal validation cohort, with centre 2 added as external validation cohort. To develop a combined radiation-clinical omics model nomogram incorporating clinical features based on images from low-dose lung cancer screening CT plain for predicting postoperative recurrence and to evaluate the performance of the nomogram in the training cohort, internal validation cohort and external validation cohort. A total of 190 patients were included in the model in centre 1 and randomised into a training cohort of 133 and an internal validation cohort of 57 in a ratio of 7:3, and 39 were included in centre 2 as an external validation cohort. In the training cohort (AUC=0.865, 95% CI 0.824-0.906), internal validation cohort (AUC=0.902, 95% CI 0.851-0.953) and external validation cohort (AUC=0.830,95% CI 0.751-0.908), the combined radiation-clinical omics model had a good predictive ability. The combined model performed significantly better than the conventional single-modality models (clinical model, radiomic model), and the calibration curve and decision curve analysis (DCA) showed high accuracy and clinical utility of the nomogram. The combined preoperative radiation-clinical omics model provides good predictive value for postoperative recurrence in stage ILUAD and combines the model's superiority in both internal and external validation cohorts, demonstrating its potential to aid in postoperative treatment strategies.

Preoperative MRI prediction and molecular pathway study of perineural invasion in intrahepatic cholangiocarcinoma: Insights from bioinformatics approach.

BackgroundThis study aimed to validate apparent diffusion coefficient (ADC) values and thresholds to predict poor neurological outcomes in out-of-hospital cardiac arrest (OHCA) survivors by quantitatively analysing the ADC values via brain magnetic resonance imaging (MRI).MethodsThis observational study used prospectively collected data from two tertiary academic hospitals. The derivation cohort comprised 70% of the patients randomly selected from one hospital, whereas the internal validation cohort comprised the remaining 30%. The external validation cohort used the data from another hospital, and the MRI data were restricted to scans conducted at 3 T within 72–96 h after an OHCA experience. We analysed the percentage of brain volume below a specific ADC value at 50-step intervals ranging from 200 to 1200 × 10–6 mm2/s, identifying thresholds that differentiate between good and poor outcomes. Poor neurological outcomes were defined as cerebral performance categories 3–5, 6 months after experiencing an OHCA.ResultsA total of 448 brain MRI scans were evaluated, including a derivation cohort (n = 224) and internal/external validation cohorts (n = 96/128, respectively). The proportion of brain volume with ADC values below 450, 500, 550, 600, and 650 × 10–6 mm2/s demonstrated good to excellent performance in predicting poor neurological outcomes in the derivation group (area under the curve [AUC] 0.89–0.91), and there were no statistically significant differences in performances among the derivation, internal validation, and external validation groups (all P > 0.5). Among these, the proportion of brain volume with an ADC below 600 × 10–6 mm2/s predicted a poor outcome with a 0% false-positive rate (FPR) and 76% (95% confidence interval [CI] 68–83) sensitivity at a threshold of > 13.2% in the derivation cohort. In both the internal and external validation cohorts, when using the same threshold, a specificity of 100% corresponded to sensitivities of 71% (95% CI 58–81) and 78% (95% CI 66–87), respectively.ConclusionsIn this validation study, by consistently restricting the MRI types and timing during quantitative analysis of ADC values in brain MRI, we observed high reproducibility and sensitivity at a 0% FPR. Prospective multicentre studies are necessary to validate these findings.

BackgroundSystemic inflammation (SI) and insulin resistance (IR) are correlated to the progression of gastrointestinal (GI) cancer. Therefore, this study aimed to analyze the potential clinical value of the C-reactive protein-triglyceride-glucose index (CTI) in relation to SI and IR in patients with GI cancer.MethodsThis prospective cohort study included patients with GI cancer. Patient data were collected from Fujian Cancer Hospital as an external validation cohort. Prognostic AUC, time-dependent ROC curve, C-index, and calibration curve analyses were used to predict the efficacy and accuracy of CTI survival prediction. Multivariate survival analysis was performed to evaluate the potential prognostic value of the CTI. Multiple logistic regression was performed to evaluate the relationship between the CTI and 90-day and 180-day mortalities.ResultsWe divided 1520 patients with GI cancer (mean age, 60.39 ± 11.3 years; male sex, 67%) into a training cohort and internal validation cohort; the external validation cohort included 476 patients. Prognostic AUC, time-dependent ROC curve, C-index, and calibration curve analyses of all cohorts indicated that the CTI could reliably and accurately predict the short- and long-term survival outcomes of patients with GI cancer. Multivariate survival analysis showed that for each standard deviation increase in the CTI, the risk of death increased by 32%, 21%, and 40% in the training, internal validation, and external validation cohorts, respectively. A high CTI was correlated to worse survival in patients with GI cancer (training cohort, hazard ratio [HR]=1.67, 95% confidence interval [CI]=1.35–2.08; internal validation cohort, HR=1.51, 95% CI=1.07–2.14, and external validation cohort, HR=1.59, 95% CI=1.18–2.13). In different tumor subgroups, a high CTI predicted worse survival outcomes for upper GI cancer (HR=1.54, 95% CI=1.18–2.01) and lower GI cancer (HR=1.98, 95% CI=1.36–2.86). Multivariate logistic regression analysis showed that a high CTI was positively correlated with 90-day (odds ratio [OR]=3.25, 95% CI=1.75–6.23) and 180-day mortalities (OR=2.66, 95% CI=1.72–4.15).ConclusionsThe CTI is related to SI and IR and can predict the short- and long-term prognosis of patients with GI cancer. Evaluation of the CTI could provide clinicians with an effective tool for predicting the prognosis of patients with GI cancer.Clinical trial registrationhttps://www.chictr.org.cn/showproj.html?proj=31813, identifier ChiCTR1800020329.

Development and derivation of bacteremia prediction model in patients with hepatobiliary infection

Transarterial chemoembolization (TACE) is a primary treatment for hepatocellular carcinoma (HCC). Despite its benefits, not all patients experience prolonged survival. Hence, this study aimed to develop a prognostic model using data from HCC patients who underwent TACE. This study collected patients with hepatocellular carcinoma who received TACE treatment in the Second Hospital of Dalian Medical University from September 2014 to September 2024 and randomly divided them into the training set and the validation set in a 7:3 ratio. Meanwhile, the data of patients who received TACE treatment at Zhongshan Hospital Affiliated to Dalian University from September 2021 to September 2024 were collected for external validation. Univariable and multivariable Cox analyses in the training cohort identified independent risk factors. Based on these factors, two predictive models were established via Cox regression and random survival forest (RSF). The predictive accuracy of models was appraised using the concordance index (C-index) and the area under the receiver operating characteristic curve (AUC). The calibration of models was assessed via calibration curves. This study collected clinical data from 815 HCC patients receiving TACE. Independent risk factors included tumor size, TNM stage, serum alpha-fetoprotein, TACE therapy count, and post-treatment direct bilirubin, lactate dehydrogenase, and gamma-glutamyl transferase levels. In the RSF model, AUCs for predicting 1-, 3-, and 5-year OS 0.88 (95% CI: 0.81-0.94), 0.80 (95% CI: 0.74-0.86), and 0.75 (95% CI: 0.67-0.81) in the internal validation cohort, and 0.87(95%CI:0.76,0.98), 0.91(95%CI:0.85,0.97) and 0.89(95%CI:0.82,0.96) in the external validation cohort. The Cox regression model yielded AUCs for OS at 1, 3, and 5 years of 0.92 (95% CI: 0.87-0.96), 0.83 (95% CI: 0.76-0.89), and 0.81 (95% CI: 0.74-0.88) in the internal validation cohort, and 0.86(95%CI:0.75,0.97), 0.89(95%CI:0.82,0.96)and 0.90(95%CI:0.81,0.99) in the external validation cohort. Given its superior AUC and C-index, a more interpretable survival nomogram was built using the Cox model. The prognostic nomogram, constructed based on routine, interpretable clinical features in this study, may help estimate the risk of death in TACE-treated HCC patients. This tool may provide a crucial reference for decision-making in follow-up treatment and rehabilitation. Not applicable.

We aimed to develop an effective model to identify the risk of concurrent bacterial infections in older patients with acute-on-chronic liver disease (AoCLD). Data from 809 individuals aged 60-80 sourced from the CATCH-LIFE cohort were analyzed. Participants were randomly assigned to training and internal validation groups at a ratio of 7:3. An independent cohort of 336 older inpatients with AoCLD from Xiangya Hospital, Central South University was used to conduct an external validation of the model. Independent risk factors were identified using LASSO and logistic regression analysis in the training cohort and were subsequently used to develop a user-friendly model. Model performance was evaluated using area under the curve (AUC), calibration plots, and decision curve analysis in the internal and external validation cohorts. Two different cutoff values were determined to stratify infection risk in older patients with AoCLD. The infection rate among older patients with AoCLD was 30.28%. Pulmonary infections were predominant, accounting for 93% of all infections. Gram-negative bacteria were the most frequently isolated pathogens, representing 64% of cases in this population. The novel model developed to identify bacterial infections included three variables: cirrhosis, absolute neutrophil count, and C-reactive protein (CRP) level. The AUC for the training, internal, and external validation datasets demonstrated high accuracy in identifying bacterial infections (AUC of the training dataset = 0.805, AUC of the internal validation dataset = 0.848, and AUC of the external validation dataset = 0.838). The model significantly outperformed neutrophil count, CRP level, and procalcitonin level alone in detecting bacterial infections among older patients with AoCLD. To facilitate clinical decision-making, we defined two cutoff values of prediction probability: a low cutoff of 32.2% to rule out bacterial infections and a high cutoff of 47.9% to confidently confirm bacterial infections. Our model aids in the early and precise diagnosis of bacterial infections in older patients with AoCLD, thereby facilitating prompt interventions to prevent adverse outcomes.

To develop and validate a computed tomography (CT) radiomics nomogram from multicentre datasets for preoperative prediction of perineural invasion (PNI) in colorectal cancer. A total of 299 patients with histologically confirmed colorectal cancer from three hospitals were enrolled in this retrospective study. Radiomic features were extracted from the whole tumour volume. The least absolute shrinkage and selection operator logistic regression was applied for feature selection and radiomics signature construction. Finally, a radiomics nomogram combining the radiomics score and clinical predictors was established. The receiver operating characteristic curve and decision curve analysis (DCA) were used to evaluate the predictive performance of the radiomics nomogram in the training cohort, internal validation and external validation cohorts. Twelve radiomics features extracted from the whole tumour volume were used to construct the radiomics model. The area under the curve (AUC) values of the radiomics model in the training cohort, internal validation cohort, external validation cohort 1, and external validation cohort 2 were 0.82 (0.75-0.90), 0.77 (0.62-0.92), 0.71 (0.56-0.85), and 0.73 (0.60-0.85), respectively. The nomogram, which combined the radiomics score with T category and N category by CT, yielded better performance in the training cohort (AUC = 0.88), internal validation cohort (AUC = 0.80), external validation cohort 1 (AUC = 0.75), and external validation cohort 2 (AUC = 0.76). DCA confirmed the clinical utility of the nomogram. The CT-based radiomics nomogram has the potential to accurately predict PNI in patients with colorectal cancer.

Natural language processing for the surveillance of postoperative venous thromboembolism

Preoperative CT-based topologically distinct intratumoral heterogeneity scores for predicting intratumoral tertiary lymphoid structures and outcomes in hepatocellular carcinoma: A multicenter study.