Machine learning-based texture analysis for differentiation of large adrenal cortical tumours on CT

Patient classification and attribute assessment based on machine learning techniques in the qualification process for surgical treatment of adrenal tumours

Adrenal gland incidentaloma is frequently identified through computed tomography and poses a common clinical challenge. Only selected cases require surgical intervention. The primary aim of this study was to compare the effectiveness of selected machine learning (ML) techniques in proper qualifying patients for adrenalectomy and to identify the most accurate algorithm, providing a valuable tool for doctors to simplify their therapeutic decisions. The secondary aim was to assess the significance of attributes for classification accuracy. In total, clinical data were collected from 33 patients who underwent adrenalectomy. Histopathological assessments confirmed the proper selection of 21 patients for surgical intervention according to the guidelines, with accuracy reaching 64%. Statistical analysis showed that Supported Vector Machines (linear) were significantly better than the baseline (p < 0.05), with accuracy reaching 91%, and imaging features of the tumour were found to be the most crucial attributes. In summarise, ML methods may be helpful in qualifying patients for adrenalectomy.

Deep learning approach for differentiating indeterminate adrenal masses using CT imaging

PURPOSE

Distinguishing stage 1-2 adrenocortical carcinoma (ACC) and large, lipid poor adrenal adenoma (LPAA) via imaging is challenging due to overlapping imaging characteristics. This study investigated the ability of deep learning to distinguish ACC and LPAA on single time-point CT images.

METHODS

Retrospective cohort study from 1994 to 2022. Imaging studies of patients with adrenal masses who had available adequate CT studies and histology as the reference standard by method of adrenal biopsy and/or adrenalectomy were included as well as four patients with LPAA determined by stability or regression on follow-up imaging. Forty-eight (48) subjects with pathology-proven, stage 1-2 ACC and 43 subjects with adrenal adenoma >3 cm in size demonstrating a mean non-contrast CT attenuation > 20 Hounsfield Units centrally were included. We used annotated single time-point contrast-enhanced CT images of these adrenal masses as input to a 3D Densenet121 model for classifying as ACC or LPAA with five-fold cross-validation. For each fold, two checkpoints were reported, highest accuracy with highest sensitivity (accuracy focused) and highest sensitivity with the highest accuracy (sensitivity focused).

RESULTS

We trained a deep learning model (3D Densenet121) to predict ACC versus LPAA. The sensitivity-focused model achieved mean accuracy: 87.2% and mean sensitivity: 100%. The accuracy-focused model achieved mean accuracy: 91% and mean sensitivity: 96%.

CONCLUSION

Deep learning demonstrates promising results distinguishing between ACC and large LPAA using single time-point CT images. Before being widely adopted in clinical practice, multicentric and external validation are needed.

Feasibility study of computed tomography texture analysis for evaluation of canine primary adrenal gland tumors

Objective

This study aimed to investigate the feasibility of computed tomography (CT) texture analysis for distinguishing canine adrenal gland tumors and its usefulness in clinical decision-making.

Materials and methods

The medical records of 25 dogs with primary adrenal masses who underwent contrast CT and a histopathological examination were retrospectively reviewed, of which 12 had adenomas (AAs), 7 had adenocarcinomas (ACCs), and 6 had pheochromocytomas (PHEOs). Conventional CT evaluation of each adrenal gland tumor included the mean, maximum, and minimum attenuation values in Hounsfield units (HU), heterogeneity of the tumor parenchyma, and contrast enhancement (type, pattern, and degree), respectively, in each phase. In CT texture analysis, precontrast and delayed-phase images of 18 adrenal gland tumors, which could be applied for ComBat harmonization were used, and 93 radiomic features (18 first-order and 75 second-order statistics) were extracted. Then, ComBat harmonization was applied to compensate for the batch effect created by the different CT protocols. The area under the receiver operating characteristic curve (AUC) for each significant feature was used to evaluate the diagnostic performance of CT texture analysis.

Results

Among the conventional features, PHEO showed significantly higher mean and maximum precontrast HU values than ACC (p < 0.05). Eight second-order features on the precontrast images showed significant differences between the adrenal gland tumors (p < 0.05). However, none of them were significantly different between AA and PHEO, or between precontrast images and delayed-phase images. This result indicates that ACC exhibited more heterogeneous and complex textures and more variable intensities with lower gray-level values than AA and PHEO. The correlation, maximal correlation coefficient, and gray level non-uniformity normalized were significantly different between AA and ACC, and between ACC and PHEO. These features showed high AUCs in discriminating ACC and PHEO, which were comparable or higher than the precontrast mean and maximum HU (AUC = 0.865 and 0.860, respectively).

Conclusion

Canine primary adrenal gland tumor differentiation can be achieved with CT texture analysis on precontrast images and may have a potential role in clinical decision-making. Further prospective studies with larger populations and cross-validation are warranted.

Emerging diagnostic methods and imaging modalities in cushing's syndrome

Endogenous Cushing's syndrome (CS) is a rare disease characterized by prolonged glucocorticoid excess. Timely diagnosis is critical to allow prompt treatment and limit long-term disease morbidity and risk for mortality. Traditional biochemical diagnostic modalities each have limitations and sensitivities and specificities that vary significantly with diagnostic cutoff values. Biochemical evaluation is particularly complex in patients whose hypercortisolemia fluctuates daily, often requiring repetition of tests to confirm or exclude disease, and when delineating CS from physiologic, nonneoplastic states of hypercortisolism. Lastly, traditional pituitary MRI may be negative in up to 60% of patients with adrenocorticotropic hormone (ACTH)-secreting pituitary adenomas (termed "Cushing's disease" [CD]) whereas false positive pituitary MRI findings may exist in patients with ectopic ACTH secretion. Thus, differentiating CD from ectopic ACTH secretion may necessitate dynamic testing or even invasive procedures such as bilateral inferior petrosal sinus sampling. Newer methods may relieve some of the diagnostic uncertainty in CS, providing a more definitive diagnosis prior to subjecting patients to additional imaging or invasive procedures. For example, a novel method of cortisol measurement in patients with CS is scalp hair analysis, a non-invasive method yielding cortisol and cortisone values representing long-term glucocorticoid exposure of the past months. Hair cortisol and cortisone have both shown to differentiate between CS patients and controls with a high sensitivity and specificity. Moreover, advances in imaging techniques may enhance detection of ACTH-secreting pituitary adenomas. While conventional pituitary MRI may fail to identify microadenomas in patients with CD, high-resolution 3T-MRI with 3D-spoiled gradient-echo sequence has thinner sections and superior soft-tissue contrast that can detect adenomas as small as 2 mm. Similarly, functional imaging may improve the identification of ACTH-secreting adenomas noninvasively; Gallium-68-tagged corticotropin-releasing hormone (CRH) combined with PET-CT can be used to detect CRH receptors, which are upregulated on corticotroph adenomas. This technique can delineate functionality of adenomas in patients with CD from patients with ectopic ACTH secretion and false positive pituitary lesions on MRI. Here, we review emerging methods and imaging modalities for the diagnosis of CS, discussing their diagnostic accuracy, strengths and limitations, and applicability to clinical practice.

Review of Diagnostic Modalities for Adrenal Incidentaloma

Adrenal incidentalomas are common findings in clinical practice, with a prevalence of up to 4.2% in radiological studies. Due to the large number of focal lesions in the adrenal glands, it can be challenging to make a definitive diagnosis and determine the appropriate management. The purpose of this review is to present current diagnostic modalities used to preoperatively distinguish between adrenocortical adenoma (ACA) and adrenocortical cancer (ACC). Proper management and diagnosis are crucial in avoiding unnecessary adrenalectomies, which occur in over 40% of cases. A literature analysis was conducted to compare ACA and ACC using imaging studies, hormonal evaluation, pathological workup, and liquid biopsy. Before deciding on surgical treatment, the nature of the tumor can be accurately determined using noncontrast CT imaging combined with tumor size and metabolomics. This approach helps to narrow down the group of patients with adrenal tumors who require surgical treatment due to the suspected malignant nature of the lesion.

Adrenal lesion classification with abdomen caps and the effect of ROI size

Accurate classification of adrenal lesions on magnetic resonance (MR) images are very important for diagnosis and treatment planning. The detection and classification of lesions in medical imaging heavily rely on several key factors, including the specialist's level of experience, work intensity, and fatigue of the clinician. These factors are critical determinants of the accuracy and effectiveness of the diagnostic process, which in turn has a direct impact on patient health outcomes. With the spread of artificial intelligence, the use of computer-aided diagnosis (CAD) systems in disease diagnosis has also increased. In this study, adrenal lesion classification was performed using deep learning on MR images. The data set used was obtained from the Department of Radiology, Faculty of Medicine, Selcuk University, and all adrenal lesions were identified and reviewed in consensus by two radiologists experienced with abdominal MR. Studies were carried out on two different data sets created by T1- and T2-weighted MR images. The data set consisted of 112 benign and 10 malignant lesions for each mode. Experiments were performed with regions of interest (ROIs) of different sizes to increase the working performance. Thus, the effect of the selected ROI size on the classification performance was assessed. In addition, instead of the convolutional neural network (CNN) models used in deep learning, a unique classification model structure called Abdomen Caps was proposed. When the data sets used in classification studies are manually separated for training, validation, and testing, different results are obtained with different data sets for each stage. To eliminate this imbalance, tenfold cross-validation was used in this study. The best results obtained were 0.982, 0.999, 0.969, 0.983, 0.998, and 0.964 for accuracy, precision, recall, F1-score, area under the curve (AUC) score, and kappa score, respectively.

Can CT texture analysis parameters be used as imaging biomarkers for prediction of malignancy in canine splenic tumors?

Splenic hemangiosarcoma has morphological similarities to benign nodular hyperplasia. Computed tomography (CT) texture analysis can analyze the texture of images that the naive human eye cannot detect. Recently, there have been attempts to incorporate CT texture analysis with artificial intelligence in human medicine. This retrospective, analytical design study aimed to assess the feasibility of CT texture analysis in splenic masses and investigate predictive biomarkers of splenic hemangiosarcoma in dogs. Parameters for dogs with hemangiosarcoma and nodular hyperplasia were compared, and an independent parameter that could differentiate between them was selected. Discriminant analysis was performed to assess the ability to discriminate the two splenic masses and compare the relative importance of the parameters. A total of 23 dogs were sampled, including 16 splenic nodular hyperplasia and seven hemangiosarcoma. In each dog, total 38 radiomic parameters were extracted from first-, second-, and higher-order matrices. Thirteen parameters had significant differences between hemangiosarcoma and nodular hyperplasia. Skewness in the first-order matrix and GLRLM_LGRE and GLZLM_ZLNU in the second, higher-order matrix were determined as independent parameters. A discriminant equation consisting of skewness, GLZLM_LGZE, and GLZLM_ZLNU was derived, and the cross-validation verification result showed an accuracy of 95.7%. Skewness was the most influential parameter for the discrimination of the two masses. The study results supported using CT texture analysis to help differentiate hemangiosarcoma from nodular hyperplasia in dogs. This new diagnostic approach can be used for developing future machine learning-based texture analysis tools.

Two nomograms based on radiomics models using triphasic CT for differentiation of adrenal lipid-poor benign lesions and metastases in a cancer population: an exploratory study

OBJECTIVES

To investigate the effectiveness of CT-based radiomics nomograms in differentiating adrenal lipid-poor benign lesions and metastases in a cancer population.

METHODS

This retrospective study enrolled 178 patients with cancer history from three medical centres categorised as those with adrenal lipid-poor benign lesions or metastases. Patients were divided into training, validation, and external testing cohorts. Radiomics features were extracted from triphasic CT images (unenhanced, arterial, and venous) to establish three single-phase models and one triphasic radiomics model using logistic regression. Unenhanced and triphasic nomograms were established by incorporating significant clinico-radiological factors and radscores. The models were evaluated by the receiver operating characteristic curve, Delong's test, calibration curve, and decision curve.

RESULTS

Lesion side, diameter, and enhancement ratio resulted as independent factors and were selected into nomograms. The areas under the curves (AUCs) of unenhanced and triphasic radiomics models in validation (0.878, 0.914, p = 0.381) and external testing cohorts (0.900, 0.893, p = 0.882) were similar and higher than arterial and venous models (validation: 0.842, 0.765; testing: 0.814, 0.806). Unenhanced and triphasic nomograms yielded similar AUCs in validation (0.903, 0.906, p = 0.955) and testing cohorts (0.928, 0.946, p = 0.528). The calibration curves showed good agreement and decision curves indicated satisfactory clinical benefits.

CONCLUSION

Unenhanced and triphasic CT-based radiomics nomograms resulted as a useful tool to differentiate adrenal lipid-poor benign lesions from metastases in a cancer population. They exhibited similar predictive efficacies, indicating that enhanced examinations could be avoided in special populations.

KEY POINTS

• All four radiomics models and two nomograms using triphasic CT images exhibited favourable performances in three cohorts to characterise the cancer population's adrenal benign lesions and metastases. • Unenhanced and triphasic radiomics models had similar predictive performances, outperforming arterial and venous models. • Unenhanced and triphasic nomograms also exhibited similar efficacies and good clinical benefits, indicating that contrast-enhanced examinations could be avoided when identifying adrenal benign lesions and metastases.

Using CT radiomic features based on machine learning models to subtype adrenal adenoma

BACKGROUND

Functioning and non-functioning adrenocortical adenoma are two subtypes of benign adrenal adenoma, and their differential diagnosis is crucial. Current diagnostic procedures use an invasive method, adrenal venous sampling, for endocrinologic assessment.

METHODS

This study proposes establishing an accurate differential model for subtyping adrenal adenoma using computed tomography (CT) radiomic features and machine learning (ML) methods. Dataset 1 (289 patients with adrenal adenoma) was collected to develop the models, and Dataset 2 (54 patients) was utilized for external validation. Cuboids containing the lesion were cropped from the non-contrast, arterial, and venous phase CT images, and 1,967 features were extracted from each cuboid. Ten discriminative features were selected from each phase or the combined phases. Random forest, support vector machine, logistic regression (LR), Gradient Boosting Machine, and eXtreme Gradient Boosting were used to establish prediction models.

RESULTS

The highest accuracies were 72.7%, 72.7%, and 76.1% in the arterial, venous, and non-contrast phases, respectively, when using radiomic features alone with the ML classifier of LR. When features from the three CT phases were combined, the accuracy of LR reached 83.0%. After adding clinical information, the area under the receiver operating characteristic curve increased for all the machine learning methods except for LR. In Dataset 2, the accuracy of LR was the highest, reaching 77.8%.

CONCLUSION

The radiomic features of the lesion in three-phase CT images can potentially suggest the functioning or non-functioning nature of adrenal adenoma. The resulting radiomic models can be a non-invasive, low-cost, and rapid method of minimizing unnecessary testing in asymptomatic patients with incidentally discovered adrenal adenoma.

Endocrine Tumor Classification via Machine-Learning-Based Elastography: A Systematic Scoping Review

Elastography complements traditional medical imaging modalities by mapping tissue stiffness to identify tumors in the endocrine system, and machine learning models can further improve diagnostic accuracy and reliability. Our objective in this review was to summarize the applications and performance of machine-learning-based elastography on the classification of endocrine tumors. Two authors independently searched electronic databases, including PubMed, Scopus, Web of Science, IEEEXpress, CINAHL, and EMBASE. Eleven (n = 11) articles were eligible for the review, of which eight (n = 8) focused on thyroid tumors and three (n = 3) considered pancreatic tumors. In all thyroid studies, the researchers used shear-wave ultrasound elastography, whereas the pancreas researchers applied strain elastography with endoscopy. Traditional machine learning approaches or the deep feature extractors were used to extract the predetermined features, followed by classifiers. The applied deep learning approaches included the convolutional neural network (CNN) and multilayer perceptron (MLP). Some researchers considered the mixed or sequential training of B-mode and elastographic ultrasound data or fusing data from different image segmentation techniques in machine learning models. All reviewed methods achieved an accuracy of ≥80%, but only three were ≥90% accurate. The most accurate thyroid classification (94.70%) was achieved by applying sequential training CNN; the most accurate pancreas classification (98.26%) was achieved using a CNN-long short-term memory (LSTM) model integrating elastography with B-mode and Doppler images.

A modified method for CT radiomics region-of-interest segmentation in adrenal lipid-poor adenomas: a two-institution comparative study

Objective

This study aimed to investigate the application of modified region-of-interest (ROI) segmentation method in unenhanced computed tomography in the radiomics model of adrenal lipid-poor adenoma, and to evaluate the diagnostic performance using an external medical institution data set and select the best ROI segmentation method.

Methods

The imaging data of 135 lipid-poor adenomas and 102 non-adenomas in medical institution A and 30 lipid-poor adenomas and 43 non-adenomas in medical institution B were retrospectively analyzed, and all cases were pathologically or clinically confirmed. The data of Institution A builds the model, and the data of Institution B verifies the diagnostic performance of the model. Semi-automated ROI segmentation of tumors was performed using uAI software, using maximum area single-slice method (MAX) and full-volume method (ALL), as well as modified single-slice method (MAX_E) and full-volume method (ALL_E) to segment tumors, respectively. The inter-rater correlation coefficients (ICC) was performed to assess the stability of the radiomics features of the four ROI segmentation methods. The area under the curve (AUC) and at least 95% specificity pAUC (Partial AUC) were used as measures of the diagnostic performance of the model.

Results

A total of 104 unfiltered radiomics features were extracted using each of the four segmentation methods. In the ROC analysis of the radiomics model, the AUC value of the model constructed by MAX was 0.925, 0.919, and 0.898 on the training set, the internal validation set, and the external validation set, respectively, and the AUC value of MAX_E was 0.937, 0.931, and 0.906, respectively. The AUC value of ALL was 0.929, 0.929, and 0.918, and the AUC value of ALL_E was 0.942, 0.926, and 0.927, respectively. In all samples, the pAUCs of MAX, MAX_E, ALL, and ALL_E were 0.021, 0.025, 0.018, and 0.028, respectively.

Conclusion

The diagnostic performance of the radiomics model constructed based on the full-volume method was better than that of the model based on the single-slice method. The model constructed using the ALL_E method had a stronger generalization ability and the highest AUC and pAUC value.

Artificial intelligence in adrenal imaging: A critical review of current applications

In the elective field of adrenal imaging, artificial intelligence (AI) can be used for adrenal lesion detection, characterization, hypersecreting syndrome management and patient follow-up. Although a perfect AI tool that includes all required steps from detection to analysis does not exist yet, multiple AI algorithms have been developed and tested with encouraging results. However, AI in this setting is still at an early stage. In this regard, most published studies about AI in adrenal gland imaging report preliminary results that do not have yet daily applications in clinical practice. In this review, recent developments and current results of AI in the field of adrenal imaging are presented. Limitations and future perspectives of AI are discussed.

Application of radiomics in adrenal incidentaloma: a literature review

Assessment of adrenal incidentaloma relies on imaging analysis and evaluation of adrenal function. Radiomics as a tool for quantitative image analysis is useful for evaluation of adrenal incidentaloma. In this review, we examined radiomic literature on adrenal incidentaloma including both adrenal functional assessment and structural differentiation of benign versus malignant adrenal tumors. In this review, we summarized the status of radiomic application on adrenal incidentaloma and suggested potential direction for future research.

Diagnostic performance of radiomics in adrenal masses: A systematic review and meta-analysis

Objectives

(1) To assess the methodological quality and risk of bias of radiomics studies investigating the diagnostic performance in adrenal masses and (2) to determine the potential diagnostic value of radiomics in adrenal tumors by quantitative analysis.

Methods

PubMed, Embase, Web of Science, and Cochrane Library databases were searched for eligible literature. Methodological quality and risk of bias in the included studies were assessed by the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) and Radiomics Quality Score (RQS). The diagnostic performance was evaluated by pooled sensitivity, specificity, diagnostic odds ratio (DOR), and area under the curve (AUC). Spearman’s correlation coefficient and subgroup analysis were used to investigate the cause of heterogeneity. Publication bias was examined using the Deeks’ funnel plot.

Results

Twenty-eight studies investigating the diagnostic performance of radiomics in adrenal tumors were identified, with a total of 3579 samples. The average RQS was 5.11 (14.2% of total) with an acceptable inter-rater agreement (ICC 0.94, 95% CI 0.93–0.95). The risk of bias was moderate according to the result of QUADAS-2. Nine studies investigating the use of CT-based radiomics in differentiating malignant from benign adrenal tumors were included in the quantitative analysis. The pooled sensitivity, specificity, DOR and AUC with 95% confidence intervals were 0.80 (0.68-0.88), 0.83 (0.73-0.90), 19.06 (7.87-46.19) and 0.88 (0.85–0.91), respectively. There was significant heterogeneity among the included studies but no threshold effect in the meta-analysis. The result of subgroup analysis demonstrated that radiomics based on unenhanced and contrast-enhanced CT possessed higher diagnostic performance, and second-order or higher-order features could enhance the diagnostic sensitivity but also increase the false positive rate. No significant difference in diagnostic ability was observed between studies with machine learning and those without.

Conclusions

The methodological quality and risk of bias of studies investigating the diagnostic performance of radiomics in adrenal tumors should be further improved in the future. CT-based radiomics has the potential benefits in differentiating malignant from benign adrenal tumors. The heterogeneity between the included studies was a major limitation to obtaining more accurate conclusions.

Systematic Review Registration

https://www.crd.york.ac.uk/PROSPERO/ CRD 42022331999 .

Optimal and novel imaging of the adrenal glands

PURPOSE OF REVIEW

Adrenal imaging forms an important role in the workup of adrenal masses. The purpose of this review is to briefly review the traditional role of imaging in adrenal diseases and highlight the most recent research and new applications aimed to improve diagnostic accuracy.

RECENT FINDINGS

The current review will focus on new applications of computed tomography (CT), MRI and PET/CT imaging, addressing the implications of artificial intelligence and radiomics in progressing diagnostic accuracy.

SUMMARY

The new applications of adrenal imaging are improving diagnostic accuracy and expanding the role of imaging, particularly with novel PET radiotracers and the use of artificial intelligence.

Can Radiomics Provide Additional Diagnostic Value for Identifying Adrenal Lipid-Poor Adenomas From Non-Adenomas on Unenhanced CT?

Background

It is difficult for radiologists to differentiate adrenal lipid-poor adenomas from non-adenomas; nevertheless, this differentiation is important as the clinical interventions required are different for adrenal lipid-poor adenomas and non-adenomas.

Purpose

To develop an unenhanced computed tomography (CT)-based radiomics model for identifying adrenal lipid-poor adenomas to assist in clinical decision-making.

Materials and methods

Patients with adrenal lesions who underwent CT between January 2015 and August 2021 were retrospectively recruited from two independent institutions. Patients from institution 1 were randomly divided into training and test sets, while those from institution 2 were used as the external validation set. The unenhanced attenuation and tumor diameter were measured to build a conventional model. Radiomics features were extracted from unenhanced CT images, and selected features were used to build a radiomics model. A nomogram model combining the conventional and radiomic features was also constructed. All the models were developed in the training set and validated in the test and external validation sets. The diagnostic performance of the models for identifying adrenal lipid-poor adenomas was compared.

Results

A total of 292 patients with 141 adrenal lipid-poor adenomas and 151 non-adenomas were analyzed. Patients with adrenal lipid-poor adenomas tend to have lower unenhanced attenuation and smoother image textures. In the training set, the areas under the curve of the conventional, radiomic, and nomogram models were 0.94, 0.93, and 0.96, respectively. There was no difference in diagnostic performance between the conventional and nomogram models in all datasets (all p < 0.05).

Conclusions

Our unenhanced CT-based nomogram model could effectively distinguish adrenal lipid-poor adenomas. The diagnostic power of conventional unenhanced CT imaging features may be underestimated, and further exploration is worthy.

Radiomics in Cross-Sectional Adrenal Imaging: A Systematic Review and Quality Assessment Study

In this study, we aimed to systematically review the current literature on radiomics applied to cross-sectional adrenal imaging and assess its methodological quality. Scopus, PubMed and Web of Science were searched to identify original research articles investigating radiomics applications on cross-sectional adrenal imaging (search end date February 2021). For qualitative synthesis, details regarding study design, aim, sample size and imaging modality were recorded as well as those regarding the radiomics pipeline (e.g., segmentation and feature extraction strategy). The methodological quality of each study was evaluated using the radiomics quality score (RQS). After duplicate removal and selection criteria application, 25 full-text articles were included and evaluated. All were retrospective studies, mostly based on CT images (17/25, 68%), with manual (19/25, 76%) and two-dimensional segmentation (13/25, 52%) being preferred. Machine learning was paired to radiomics in about half of the studies (12/25, 48%). The median total and percentage RQS scores were 2 (interquartile range, IQR = −5–8) and 6% (IQR = 0–22%), respectively. The highest and lowest scores registered were 12/36 (33%) and −5/36 (0%). The most critical issues were the absence of proper feature selection, the lack of appropriate model validation and poor data openness. The methodological quality of radiomics studies on adrenal cross-sectional imaging is heterogeneous and lower than desirable. Efforts toward building higher quality evidence are essential to facilitate the future translation into clinical practice.

Machine intelligence in non-invasive endocrine cancer diagnostics

Artificial intelligence (AI) has illuminated a clear path towards an evolving health-care system replete with enhanced precision and computing capabilities. Medical imaging analysis can be strengthened by machine learning as the multidimensional data generated by imaging naturally lends itself to hierarchical classification. In this Review, we describe the role of machine intelligence in image-based endocrine cancer diagnostics. We first provide a brief overview of AI and consider its intuitive incorporation into the clinical workflow. We then discuss how AI can be applied for the characterization of adrenal, pancreatic, pituitary and thyroid masses in order to support clinicians in their diagnostic interpretations. This Review also puts forth a number of key evaluation criteria for machine learning in medicine that physicians can use in their appraisals of these algorithms. We identify mitigation strategies to address ongoing challenges around data availability and model interpretability in the context of endocrine cancer diagnosis. Finally, we delve into frontiers in systems integration for AI, discussing automated pipelines and evolving computing platforms that leverage distributed, decentralized and quantum techniques.

Diagnostic Accuracy of CT Texture Analysis in Adrenal Masses: A Systematic Review

Adrenal incidentalomas (AIs) are incidentally discovered adrenal neoplasms. Overt endocrine secretion (glucocorticoids, mineralocorticoids, and catecholamines) and malignancy (primary or metastatic disease) are assessed at baseline evaluation. Size, lipid content, and washout characterise benign AIs (respectively, <4 cm, <10 Hounsfield unit, and rapid release); nonetheless, 30% of adrenal lesions are not correctly indicated. Recently, image-based texture analysis from computed tomography (CT) may be useful to assess the behaviour of indeterminate adrenal lesions. We performed a systematic review to provide the state-of-the-art of texture analysis in patients with AI. We considered 9 papers (from 70 selected), with a median of 125 patients (range 20–356). Histological confirmation was the most used criteria to differentiate benign from the malignant adrenal mass. Unenhanced or contrast-enhanced data were available in all papers; TexRAD and PyRadiomics were the most used software. Four papers analysed the whole volume, and five considered a region of interest. Different texture features were reported, considering first- and second-order statistics. The pooled median area under the ROC curve in all studies was 0.85, depicting a high diagnostic accuracy, up to 93% in differentiating adrenal adenoma from adrenocortical carcinomas. Despite heterogeneous methodology, texture analysis is a promising diagnostic tool in the first assessment of patients with adrenal lesions.

Machine Learning-Based Texture Analysis in the Characterization of Cortisol Secreting vs. Non-Secreting Adrenocortical Incidentalomas in CT Scan

New radioimaging techniques, exploiting the quantitative variables of imaging, permit to identify an hypothetical pathological tissue. We have applied this potential in a series of 72 adrenal incidentalomas (AIs) followed at our center, subdivided in functioning and non-functioning using laboratory findings. Each AI was studied in the preliminary non-contrast phase with a specific software (Mazda), surrounding a region of interest within each lesion. A total of 314 features were extrapolated. Mean and standard deviations of features were obtained and the difference in means between the two groups was statistically analyzed. Receiver Operating Characteristic (ROC) curves were used to identify an optimal cutoff for each variable and a prediction model was constructed via multivariate logistic regression with backward and stepwise selection. A 11-variable prediction model was constructed, and a ROC curve was used to differentiate patients with high probability of functioning AI. Using a threshold value of >-275.147, we obtained a sensitivity of 93.75% and a specificity of 100% in diagnosing functioning AI. On the basis of these results, computed tomography (CT) texture analysis appears a promising tool in the diagnostic definition of AIs.

Differential diagnostic value of plain CT scan in adrenal adenoma and non-adenoma: A two-center control study of mean attenuation value, minimum attenuation value, and CT histogram

OBJECTIVES

To investigate the value of mean attenuation value (AV_mean), minimum attenuation value (AV_min), and CT histogram (CTH) for the differential diagnosis of adrenal adenoma and non-adenoma in two medical centers.

METHODS

The plain CT data of 403 cases of adrenal adenoma and 141 cases of non-adenoma in center A were retrospectively analyzed, and compared with data of 86 cases of adenoma and 71 cases of non-adenoma in center B. All cases were confirmed by pathology or clinical follow-up. The diagnostic efficacy of AV_mean ≤ 10 Hounsfield units (HU), AV_min ≤ 0 HU, and CTH negative pixels ≥ 10% for adrenal adenoma, and AV_min and CTH for adenoma with AV_mean > 10Hu were compared between the two medical centers.

RESULTS

In medical centers A and B, the AUC of AV_mean for the differential diagnosis of adenoma and non-adenoma was 0.956 and 0.956, respectively, and the corresponding sensitivity, specificity, and accuracy were, 0.591 and 0.663, 1.000 and 1.000, 0.697, and 0.815, respectively, when the threshold was ≤ 10 HU. The AUC of AV_min was 0.941 and 0.958, respectively, and the corresponding sensitivity, specificity, and accuracy were 0.869 and 0.826, 0.986, and 0.972, 0.899, and 0.892, respectively, when the threshold was ≤ 0 HU. The AUC of CTH negative pixels was 0.948 and 0.952, respectively, and the corresponding sensitivity, specificity, and accuracy were 0.759 and 0.674, 1.000 and 1.000, 0.822, and 0.822, respectively, when the threshold was ≥ 10%. Among adenoma with AV_mean >10 HU, the best threshold of AV_min in center A and center B were -0.250HU and 2.375HU, and the corresponding AUC, sensitivity and specificity were 0.858 and 0.846, 0.691 and 0.586, 0.986 and 0.958; the best threshold of CTH in center A and center B were 0.895% and 0.775%, and the corresponding AUC, sensitivity and specificity were 0.873 and 0.822, 0.818 and 0.724, 0.837 and 0.915.

CONCLUSION

AV_mean, AV_min, and CTH are all important parameters for differentiating adrenal adenoma from non-adenoma. Even for adenomas with AV_mean > 10 HU, AV_min and CTH still had high diagnostic efficiency. The three parameters are complementary, assisting clinicians to develop personalized treatments.

Machine learning-based texture analysis for differentiation of radiologically indeterminate small adrenal tumors on adrenal protocol CT scans

GOAL

To evaluate the ability of radiomic feature extraction and a machine learning algorithm to differentiate between benign and malignant indeterminate adrenal lesions on contrast-enhanced computed tomography (CT) studies.

BACKGROUND

Adrenal "incidentalomas" are adrenal lesions that are accidentally discovered during workup not related to the adrenal glands; they have an incidence as high as 5%. Small adrenal incidentalomas (< 4 cm) with high attenuation values on pre-contrast CT(> 10 HU) need further evaluation to calculate the absolute percentage of washout (APW). If the APW is < 60%, these lesions are considered non-adenomas and commonly classified as indeterminate adrenal lesions. Further workup for indeterminate lesions includes more complicated and expensive radiological studies or invasive procedures like biopsy or surgical resection.

METHODS

We searched our institutional database for indeterminate adrenal lesions with the following characteristics: < 4 cm, pre-attenuation value > 10 HU, and APW < 60%. Exclusion criteria included pheochromocytoma and no histopathological examination. CT images were converted to Nifti format, and adrenal tumors were segmented using Amira software. Radiomic features from the adrenal mask were extracted using PyRadiomics software after removing redundant features (highly pairwise correlated features and low-variance features) using recursive feature extraction to select the final discriminative set of features. Lastly, the final features were used to build a binary classification model using a random forest algorithm, which was validated and tested using leave-one-out cross-validation, confusion matrix, and receiver operating characteristic curve.

RESULTS

We found 40 indeterminate adrenal lesions (21 benign and 19 malignant). Feature extraction resulted in 3947 features, which reduced down to 62 features after removing redundancies. Recursive feature elimination resulted in the following top 4 discriminative features: gray-level size zone matrix-derived size zone non-uniformity from pre-contrast and delayed phases, gray-level dependency matrix-derived large dependence high gray-level emphasis from venous-phase, and gray-level co-occurrence matrix-derived cluster shade from delayed-phase. A binary classification model with leave-one-out cross-validation showed AUC = 0.85, sensitivity = 84.2%, and specificity = 71.4%.

CONCLUSION

Machine learning and radiomic features extraction can differentiate between benign and malignant indeterminate adrenal tumors and can be used to direct further workup with high sensitivity and specificity.

Relative Enhancement Ratio of Portal Venous Phase to Unenhanced CT in the Diagnosis of Lipid-poor Adrenal Adenomas

Background The development of an accurate, practical, noninvasive, and widely available diagnostic approach to characterize lipid-poor adrenal lesions (greater than 10 HU at unenhanced CT) remains an ongoing demand. Purpose To investigate whether combined assessment of unenhanced and portal venous phase CT allows for the differentiation of lipid-poor adrenal adenomas from nonadenomas. Materials and Methods Patients with lipid-poor adrenal lesions who underwent unenhanced and portal venous phase CT with a single-energy scanner between January 2016 and March 2020 were identified retrospectively. For each lesion, the unenhanced and contrast-enhanced attenuation were measured; the absolute enhancement (contrast-enhanced minus unenhanced attenuation [HU]) and relative enhancement ratio ([absolute enhancement divided by unenhanced attenuation] × 100%) were calculated. The sensitivity achieved at 95% specificity to distinguish adenomas from nonadenomas was determined with receiver operating characteristic curve analysis and compared among parameters with use of the McNemar test. Results A total of 220 patients (mean age ± standard deviation, 66 years ± 12; 134 men) with 131 lipid-poor adenomas and 89 nonadenomas were analyzed. The sensitivity (achieved at 95% specificity) of the relative enhancement ratio (86% [113 of 131 adenomas; 95% CI: 79, 92] at a threshold of >210%) was higher than that of unenhanced attenuation (50% [66 of 131 adenomas; 95% CI: 42, 59] at a threshold of ≤21 HU), contrast-enhanced attenuation (3% [four of 131 adenomas; 95% CI: 1, 8] at a threshold of >120 HU), and absolute enhancement (24% [32 of 131 adenomas; 95% CI: 17, 33] at a threshold of >74 HU; all P < .001). The sensitivities of the relative enhancement ratio were 100% (58 of 58 adenomas; 95% CI: 94, 100), 83% (52 of 63 adenomas; 95% CI: 71, 91), and 30% (three of 10 adenomas; 95% CI: 7, 65) for adenomas measuring unenhanced attenuation of more than 10 HU up to 20 HU, 21-30 HU, and more than 30 HU, respectively. Conclusion A relative enhancement ratio threshold of greater than 210%, measured at unenhanced and portal venous phase CT, accurately differentiated lipid-poor adenomas from nonadenomas, particularly for lesions with unenhanced attenuation of 10-30 HU. © RSNA, 2021 Online supplemental material is available for this article.

[Adrenal imaging: anatomy and pathology (literature review)]

This literature review focuses on the normal adrenal gland anatomy and typical imaging features necessary to evaluate benign and malignant lesions. In particular, adenoma, pheochromocytoma, metastases and adrenocortical carcinoma were discussed as some of the most common lesions. For this purpose, a review of relevant local and international literature sources up to January 2021 was conducted.In many cases, adrenal incidentalomas have distinctive features allowing characterization using noninvasive methods. It is possible to suspect a malignant nature and promptly refer the patient for the necessary invasive examinations in some cases. -Computed tomography, especially with intravenous contrast enhancement, is the primary imaging modality because it enables differential diagnosis. Magnetic resonance tomography remains a sensitive method in lesion detection and follow-up but is not very specific for determining the malignant potential. Positron emission computed tomography also remains an additional method and is used mainly for differential diagnosis of malignant tumors, detecting metastases and recurrences after surgical treatment. Ultrasound has a limited role but is nevertheless of great importance in the pediatric population, especially newborns. Promising techniques such as radiomics and dual-energy CT can expand imaging capabilities and improve diagnostic accuracy.Because adrenal lesions are often incidentally detected by imaging performed for other reasons, it is vital to interpret such findings correctly. This review should give the reader a broad overview of how different imaging modalities can evaluate adrenal pathology and guide radiologists and clinicians.

Challenges and opportunities for artificial intelligence in oncological imaging

Imaging plays a key role in oncology, including the diagnosis and detection of cancer, determining clinical management, assessing treatment response, and complications of treatment or disease. The current use of clinical oncology is predominantly qualitative in nature with some relatively crude size-based measurements of tumours for assessment of disease progression or treatment response; however, it is increasingly understood that there may be significantly more information about oncological disease that can be obtained from imaging that is not currently utilized. Artificial intelligence (AI) has the potential to harness quantitative techniques to improve oncological imaging. These may include improving the efficiency or accuracy of traditional roles of imaging such as diagnosis or detection. These may also include new roles for imaging such as risk-stratifying patients for different types of therapy or determining biological tumour subtypes. This review article outlines several major areas in oncological imaging where there may be opportunities for AI technology. These include (1) screening and detection of cancer, (2) diagnosis and risk stratification, (3) tumour segmentation, (4) precision oncology, and (5) predicting prognosis and assessing treatment response. This review will also address some of the potential barriers to AI research in oncological imaging.

Handcrafted MRI radiomics and machine learning: Classification of indeterminate solid adrenal lesions

PURPOSE

To assess a radiomic machine learning (ML) model in classifying solid adrenal lesions (ALs) without fat signal drop on chemical shift (CS) as benign or malignant.

METHOD

55 indeterminate ALs (21 lipid poor adenomas, 15 benign pheocromocytomas, 1 oncocytoma, 12 metastases, 6 primary tumors) showing no fat signal drop on CS were retrospectively included. Manual 3D segmentation on T2-weighted and CS images was performed for subsequent radiomic feature extraction. After feature stability testing and an 80-20% train-test split, the train set was balanced via oversampling. Following a multi-step feature selection, an Extra Trees model was tuned with 5-fold stratified cross-validation in the train set and then tested on the hold-out test set.

RESULTS

A total of 3396 features were extracted from each AL, of which 133 resulted unstable while none had low variance (< 0.01). Highly correlated (r > 0.8) features were also excluded, leaving 440 parameters. Among these, Support Vector Machine 5-fold stratified cross-validated recursive feature elimination selected a subset of 6 features. ML obtained a cross-validation accuracy of 0.94 on the train and 0.91 on the test sets. Precision, recall and F1 score were respectively 0.92, 0.91 and 0.91.

CONCLUSIONS

Our MRI handcrafted radiomics and ML pipeline proved useful to characterize benign and malignant solid indeterminate adrenal lesions.

Radiomics: a new tool to differentiate adrenocortical adenoma from carcinoma

Background

The main challenge in the management of indeterminate incidentally discovered adrenal tumours is to differentiate benign from malignant lesions. In the absence of clear signs of invasion or metastases, imaging techniques do not always precisely define the nature of the mass. The present pilot study aimed to determine whether radiomics may predict malignancy in adrenocortical tumours.

Methods

CT images in unenhanced, arterial, and venous phases from 19 patients who had undergone resection of adrenocortical tumours and a cohort who had undergone surveillance for at least 5 years for incidentalomas were reviewed. A volume of interest was drawn for each lesion using dedicated software, and, for each phase, first-order (histogram) and second-order (grey-level colour matrix and run-length matrix) radiological features were extracted. Data were revised by an unsupervised machine learning approach using the K-means clustering technique.

Results

Of operated patients, nine had non-functional adenoma and 10 carcinoma. There were 11 patients in the surveillance group. Two first-order features in unenhanced CT and one in arterial CT, and 14 second-order parameters in unenhanced and venous CT and 10 second-order features in arterial CT, were able to differentiate adrenocortical carcinoma from adenoma (P < 0.050). After excluding two malignant outliers, the unsupervised machine learning approach correctly predicted malignancy in seven of eight adrenocortical carcinomas in all phases.

Conclusion

Radiomics with CT texture analysis was able to discriminate malignant from benign adrenocortical tumours, even by an unsupervised machine learning approach, in nearly all patients.

Future Directions in Diagnosis, Prognosis and Disease Monitoring of Adrenocortical Carcinoma: Novel Non-Invasive Biomarkers

Adrenocortical carcinoma (ACC) is a rare endocrine malignancy with frequent metastatic spread and poor prognosis. The disease can occur at any age with unexpected biological behavior. Recent genome-wide studies of ACC have contributed to our understanding of the disease, but diagnosis of ACC remains a challenge, even for multidisciplinary expert teams. Patients with ACC are frequently diagnosed in advanced stages and have limited therapeutic options. Therefore, for earlier diagnosis and better clinical management of adrenocortical carcinoma, specific, sensitive, and minimal invasive markers are urgently needed. Over several decades, great efforts have been made in discovering novel and reliable diagnostic and prognostic biomarkers including microRNAs, steroid profilings, circulating tumor cells, circulating tumor DNAs and radiomics. In this review, we will summarize these novel noninvasive biomarkers and analyze their values for diagnosis, predicting prognosis, and disease monitoring. Current problems and possible future application of these non-invasive biomarkers will also be discussed.

Preresection Radiologic Assessment and Imaging Features of 156 Pathologically Proven Adrenal Adenomas

PURPOSE

The aims of the study were to assess the typical and atypical radiologic features of pathologically proven adrenal adenomas and to determine the relationship between the radiologic and histopathologic classification.

METHODS

We retrospectively studied 156 pathologically proven adrenal adenomas in 154 patients from our institutional databases who have computed tomography (CT) and/or magnetic resonance imaging (MRI) examinations before intervention. We determined the histopathologic diagnosis (typical or atypical) using Weiss scoring and classified the adenomas radiologically into typical, atypical, or indeterminate based on lesion size, precontrast CT attenuation, absolute percentage washout, calcification, and necrosis. The κ statistic was used to assess the agreement between radiologists. The Fisher exact test was used to compare the radiologic and pathological classifications.

RESULTS

In consensus, there were 83 typical, 42 atypical, and 31 indeterminate adrenal lesions. Logistic regression model showed that radiologically atypical adenoma was significantly associated with larger size, lobulated shape, higher unenhanced CT attenuation, heterogeneous appearance, nonfunctioning status, absolute percentage washout of less than 60%, and a signal intensity index of less than 16.5%.Pathologically, 147 adenomas were pathologically typical (Weiss 0), and 9 adenomas were pathologically atypical (Weiss 1-2). Radiologically, there was substantial agreement between both readers, with Cohen κ at 0.71. Approximately 98% of radiologically typical adenomas were pathologically typical. Only 17% of radiologically atypical adenomas were pathologically atypical. All radiologically indeterminate adenomas were pathologically typical. However, some of the radiologically indeterminate and typical adenomas still had an atypical component on pathologic analysis, such as necrosis, nuclear atypia, or oncocytic features.

CONCLUSIONS

Radiologically atypical lesion was significantly associated with larger size and higher unenhanced CT attenuation. Approximately 27% of the cases demonstrated atypical features on imaging. Most radiologically atypical adrenal adenomas are pathologically typical.

Adrenal cortical carcinoma: pathology, genomics, prognosis, imaging features, and mimics with impact on management

Adrenocortical carcinoma (ACC) is a rare tumor with a poor prognosis. Most tumors are either metastatic or locally invasive at the time of diagnosis. Differentiation between ACC and other adrenal masses depends on clinical, biochemical, and imaging factors. This review will discuss the genetics, pathological, and imaging feature of ACC.