Intratumoral and peritumoral MRI radiomics nomogram for predicting parametrial invasion in patients with early-stage cervical adenocarcinoma and adenosquamous carcinoma

Overlooked and underpowered: a meta-research addressing sample size in radiomics prediction models for binary outcomes

OBJECTIVES

To investigate how studies determine the sample size when developing radiomics prediction models for binary outcomes, and whether the sample size meets the estimates obtained by using established criteria.

METHODS

We identified radiomics studies that were published from 01 January 2023 to 31 December 2023 in seven leading peer-reviewed radiological journals. We reviewed the sample size justification methods, and actual sample size used. We calculated and compared the actual sample size used to the estimates obtained by using three established criteria proposed by Riley et al. We investigated which characteristics factors were associated with the sufficient sample size that meets the estimates obtained by using established criteria proposed by Riley et al. RESULTS: We included 116 studies. Eleven out of one hundred sixteen studies justified the sample size, in which 6/11 performed a priori sample size calculation. The median (first and third quartile, Q1, Q3) of the total sample size is 223 (130, 463), and those of sample size for training are 150 (90, 288). The median (Q1, Q3) difference between total sample size and minimum sample size according to established criteria are -100 (-216, 183), and those differences between total sample size and a more restrictive approach based on established criteria are -268 (-427, -157). The presence of external testing and the specialty of the topic were associated with sufficient sample size.

CONCLUSION

Radiomics studies are often designed without sample size justification, whose sample size may be too small to avoid overfitting. Sample size justification is encouraged when developing a radiomics model.

KEY POINTS

Question Sample size justification is critical to help minimize overfitting in developing a radiomics model, but is overlooked and underpowered in radiomics research. Findings Few of the radiomics models justified, calculated, or reported their sample size, and most of them did not meet the recent formal sample size criteria. Clinical relevance Radiomics models are often designed without sample size justification. Consequently, many models are too small to avoid overfitting. It should be encouraged to justify, perform, and report the considerations on sample size when developing radiomics models.

Intratumoral and peritumoral habitat imaging based on multiparametric MRI to predict cervical stromal invasion in early-stage endometrial carcinoma

RATIONALE AND OBJECTIVES

To evaluate the validity of multiparametric MRI-based intratumoral and peritumoral habitat imaging for predicting cervical stromal invasion (CSI) in patients with early-stage endometrial carcinoma (EC) and to compare the performance of structural and functional habitats.

MATERIALS AND METHODS

The preoperative MRI and clinical data of 680 patients with early-stage EC from three centers were retrospectively analyzed. Based on cohort-level, gaussian mixture model (GMM) algorithm was used for habitat clustering of MRI images. Structural habitats were clustered using T2-weighted imaging (T2WI) and contrast-enhanced T1-weighted imaging (CE-T1WI), and functional habitats were clustered using apparent diffusion coefficient (ADC) mapping and CE-T1WI. Habitat parameters were extracted from four volumes of interest (VOIs): intratumoral regions (ROI), peritumoral loops of 3 mm dilation (L3), intratumoral regions + peritumoral loops of 3 mm dilation (R3), and peritumoral loops of 3 mm dilation + peritumoral loops of 3 mm erosion (DE3). Clinical-habitat models were constructed by combining clinical independent predictors and optimal habitat models. The model performance was evaluated by the area under the curve (AUC).

RESULTS

Deep myometrial invasion (DMI) was an independent predictor. L3 models showed the best performance for both structural and functional habitats, and the L3 functional habitat model had the highest average AUC (0.807) in external test groups, and the average AUC increased to 0.815 when combing with the clinical independent predictor.

CONCLUSION

Multiparametric MRI-based intratumoral and peritumoral habitat imaging provides a noninvasive approach to predict CSI in EC patients. The combination of the clinical predictor with the L3 functional habitat model improved predictive performance.

Preoperative prediction of lymph node metastasis in intrahepatic cholangiocarcinoma: an integrative approach combining ultrasound-based radiomics and inflammation-related markers

OBJECTIVES

To develop ultrasound-based radiomics models and a clinical model associated with inflammatory markers for predicting intrahepatic cholangiocarcinoma (ICC) lymph node (LN) metastasis. Both are integrated for enhanced preoperative prediction.

METHODS

This study retrospectively enrolled 156 surgically diagnosed ICC patients. A region of interest (ROI) was manually identified on the ultrasound image of the tumor to extract radiomics features. In the training cohort, we performed a Wilcoxon test to screen for differentially expressed features, and then we used 12 machine learning algorithms to develop 107 models within the cross-validation framework and determine the optimal radiomics model through receiver operating characteristic (ROC) curve analysis. Multivariable logistic regression analysis was used to identify independent risk factors to construct a clinical model. The combined model was established by combining ultrasound-based radiomics and clinical parameters. The Delong test and decision curve analysis (DCA) were used to compare the diagnostic efficacy and clinical utility of different models.

RESULTS

A total of 1239 radiomics features were extracted from the ROIs of tumors. Among the 107 prediction models, the model (Stepglm + LASSO) utilizing 10 radiomics features ultimately yielded the highest average area under the receiver operating characteristic curve (AUC) of 0.872, with an AUC of 0.916 in the training cohort and 0.827 in the validation cohort. The combined model, which incorporates the optimal radiomics score, clinical N stage, and platelet-to-lymphocyte ratio (PLR), achieved an AUC of 0.882 in the validation cohort, significantly outperforming the clinical model with an AUC of 0.687 (P = 0.009). According to the DCA analysis, the combined model also showed better clinical benefits.

CONCLUSIONS

The combined model incorporating ultrasound-based radiomics features and the PLR marker offers an effective, noninvasive intelligence-assisted tool for preoperative LN metastasis prediction in ICC patients.

CLINICAL TRIAL NUMBER

Not applicable.

Construction of a nomogram for personalized prediction of lower limb lymphedema risk after cervical cancer surgery

OBJECTIVE

This study aims to construct and evaluate an individualized nomogram for predicting the risk of lower limb lymphedema after cervical cancer surgery. Healthcare professionals can utilize line chart models to predict the probability of postoperative lower limb lymphedema in different patients, allowing for the early identification of high-risk patients and facilitating early prevention and treatment.

METHODS

A retrospective study was conducted among 411 cervical cancer patients treated at our hospital from May 2021 to December 2023. The patients were randomly divided into a modeling group (313 cases) and a validation group (98 cases) according to an approximate 3:1 ratio. The modeling group was further divided into a lower limb lymphedema group (61 cases) and a non-lower limb lymphedema group (252 cases) based on the presence of postoperative lower limb lymphedema. Multiple factors Logistic regression was used to identify risk factors, and a nomogram was constructed using R software version 4.0.2, with internal and external validation performed.

RESULTS

Risk factors for lower limb lymphedema following cervical cancer surgery include age 60 years or older, a Body Mass Index (BMI) of 24 kg/m² or higher, hypertension, the removal of 30 or more lymph nodes, adjuvant radiotherapy and chemotherapy, and prolonged standing for six hours or more (P < 0.05). Internal and external validation results demonstrated that the calibration curve closely aligned with the ideal curve. The Area Under Curve(AUC) of the Receiver Operating Characteristic(ROC) curve was 0.890 (95% CI: 0.844 ∼ 0.936) and 0.876 (95% CI: 0.821 ∼ 0.930), indicating high model calibration and discrimination. Decision Curve Analysis(DCA) curve revealed that the Logistic model had good net returns and high clinical practicality when the probability range of the high-risk threshold was 0.11 ∼ 0.98.

CONCLUSION

The nomogram, developed using factors such as age, BMI, hypertension, number of lymph nodes dissected, adjuvant radiotherapy and chemotherapy, and duration of standing, has strong predictive value and offers significant clinical benefits, making it a valuable tool for clinical decision-making.

Radiomics based on MRI in predicting lymphovascular space invasion of cervical cancer: a meta-analysis

Objective

The objective of this meta-analysis is to assess the efficacy of radiomics techniques utilizing magnetic resonance imaging (MRI) for predicting lymphovascular space invasion (LVSI) in patients with cervical cancer (CC).

Methods

A comprehensive literature search was conducted in databases including PubMed, Embase, Cochrane Library, Medline, Scopus, CNKI, and Wanfang, with studies published up to 08/04/2024, being considered for inclusion. The meta-analysis was performed using Stata 15 and Review Manager 5.4. The quality of the included studies was evaluated using the Quality Assessment of Diagnostic Accuracy Studies 2 and Radiomics Quality Score tools. The analysis encompassed the pooled sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), and diagnostic odds ratio (DOR). Summary ROC curves were constructed, and the AUC was calculated. Heterogeneity was investigated using meta-regression. Statistical significance was set at p ≤ 0.05.

Results

There were 13 studies involving a total of 2,245 patients that were included in the meta-analysis. The overall sensitivity and specificity of the MRI-based model in the Training set were 83% (95% CI: 77%-87%) and 72% (95% CI: 74%-88%), respectively. The AUC, DOR, PLR, and NLR of the MRI-based model in the Training set were 0.89 (95% CI: 0.86-0.91), 22 (95% CI: 12-40), 4.6 (95% CI: 3.1-7.0), and 0.21 (95% CI: 0.16-0.29), respectively. Subgroup analysis revealed that the AUC of the model combining radiomics with clinical factors [0.90 (95% CI: 0.87-0.93)] was superior to models based on T2-weighted imaging (T2WI) sequence [0.78 (95% CI: 0.74-0.81)], contrast-enhanced T1-weighted imaging (T1WI-CE) sequence [0.85 (95% CI: 0.82-0.88)], and multiple sequences [0.86 (95% CI: 0.82-0.89)] in the Training set. The pooled sensitivity and specificity of the model integrating radiomics with clinical factors [83% (95% CI: 73%-89%) and 86% (95% CI: 73%-93%)] surpassed those of models based on the T2WI sequence [79% (95% CI: 71%-85%) and 72% (95% CI: 67%-76%)], T1WI-CE sequence [78% (95% CI: 67%-86%) and 78% (95% CI: 68%-86%)], and multiple sequences [78% (95% CI: 67%-87%) and 79% (95% CI: 70%-87%)], respectively. Funnel plot analysis indicated an absence of publication bias (p > 0.05).

Conclusion

MRI-based radiomics demonstrates excellent diagnostic performance in predicting LVSI in CC patients. The diagnostic performance of models combing radiomics and clinical factors is superior to that of models utilizing radiomics alone.

Systematic review registration

https://www.crd.york.ac.uk/PROSPERO/#myprospero, identifier CRD42024538007.

MRI grading for informed clinical decision-making in Peutz-Jeghers syndrome patients with cervical lesions

The preoperative diagnosis and management of Peutz-Jeghers syndrome (PJS) patients with cervical lesions remain problematic. This study analysed the associations between pathological types of cervical lesions in PJS patients and their MRI features. A total of 34 PJS patients were included and two experienced radiologists reviewed the MRIs independently. Based on the pathological diagnosis, the patients were categorized into four groups: normal (n  =  4), lobular endocervical glandular hyperplasia (LEGH, n  =  11), atypical LEGH (aLEGH, n  =  8), and gastric-type endocervical adenocarcinoma (G-EAC, n  =  11). By observing the MRI features, we found statistically significant differences in the extent of lesions (P  = 0 .001), distribution of microcysts (P  = 0 .001), proportion of microcysts (P  <  0.001) and endometrial involvement (P  = 0.019) among the four groups. Notably, solid components and disrupted cervical stromal rings were found only in the aLEGH and G-EAC groups (P  < 0.001).  Consequently, we created a novel grading system based on the aforementioned MRI features to align with the potential malignancy of cervical lesions in PJS patients. This system enables patients to receive timely and appropriate treatment recommendations while facilitating collaboration between radiologists and physicians.

MRI-Based Radiomics as a Promising Noninvasive Diagnostic Technique for Adenomyosis

Background: MRI diagnostics are important for adenomyosis, especially in cases with inconclusive ultrasound. This study assessed the potential of MRI-based radiomics as a novel tool for differentiating between uteri with and without adenomyosis. Methods: This retrospective proof-of-principle single-center study included nine patients with and six patients without adenomyosis. All patients had preoperative T2w MR images and histological findings served as the reference standard. The uterus of each patient was segmented in 3D using dedicated software, and 884 radiomics features were extracted. After dimension reduction and feature selection, the diagnostic yield of individual and combined features implemented in the machine learning models were assessed by means of receiver operating characteristics analyses. Results: Eleven relevant radiomics features were identified. The diagnostic performance of individual features in differentiating adenomyosis from the control group was high, with areas under the curve (AUCs) ranging from 0.78 to 0.98. The performance of ML models incorporating several features was excellent, with AUC scores of 1 and an area under the precision-recall curve of 0.4. Conclusions: The set of radiomics features derived from routine T2w MRI enabled accurate differentiation of uteri with adenomyosis. Radiomics could enhance diagnosis and furthermore serve as an imaging biomarker to aid in personalizing therapies and monitoring treatment responses.

Multitask prediction models for serous ovarian cancer by preoperative CT image assessments based on radiomics

Objective

High-grade serous ovarian cancer (HGSOC) has the highest mortality rate among female reproductive system tumors. Accurate preoperative assessment is crucial for treatment planning. This study aims to develop multitask prediction models for HGSOC using radiomics analysis based on preoperative CT images.

Methods

This study enrolled 112 patients diagnosed with HGSOC. Laboratory findings, including serum levels of CA125, HE-4, and NLR, were collected. Radiomic features were extracted from manually delineated ROI on CT images by two radiologists. Classification models were developed using selected optimal feature sets to predict R0 resection, lymph node invasion, and distant metastasis status. Model evaluation was conducted by quantifying receiver operating curves (ROC), calculating the area under the curve (AUC), De Long's test.

Results

The radiomics models applied to CT images demonstrated superior performance in the testing set compared to the clinical models. The area under the curve (AUC) values for the combined model in predicting R0 resection were 0.913 and 0.881 in the training and testing datasets, respectively. De Long's test indicated significant differences between the combined and clinical models in the testing set (p = 0.003). For predicting lymph node invasion, the AUCs of the combined model were 0.868 and 0.800 in the training and testing datasets, respectively. The results also revealed significant differences between the combined and clinical models in the testing set (p = 0.002). The combined model for predicting distant metastasis achieved AUCs of 0.872 and 0.796 in the training and test datasets, respectively. The combined model displayed excellent agreement between observed and predicted results in predicting R0 resection, while the radiomics model demonstrated better calibration than both the clinical model and combined model in predicting lymph node invasion and distant metastasis. The decision curve analysis (DCA) for predicting R0 resection favored the combined model over both the clinical and radiomics models, whereas for predicting lymph node invasion and distant metastasis, DCA favored the radiomics model over both the clinical model and combined model.

Conclusion

The identified radiomics signature holds potential value in preoperatively evaluating the R0, lymph node invasion and distant metastasis in patients with HGSC. The radiomics nomogram demonstrated the incremental value of clinical predictors for surgical outcome and metastasis estimation.