Computer-aided detection for the identification of pulmonary nodules in pediatric oncology patients: initial experience

Performance Analysis in Children of Traditional and Deep Learning CT Lung Nodule Computer-Aided Detection Systems Trained on Adults

BACKGROUND. Although primary lung cancer is rare in children, chest CT is commonly performed to assess for lung metastases in children with cancer. Lung nodule computer-aided detection (CAD) systems have been designed and studied primarily using adult training data, and the efficacy of such systems when applied to pediatric patients is poorly understood. OBJECTIVE. The purpose of this study was to evaluate in children the diagnostic performance of traditional and deep learning CAD systems trained with adult data for the detection of lung nodules on chest CT scans and to compare the ability of such systems to generalize to children versus to other adults. METHODS. This retrospective study included pediatric and adult chest CT test sets. The pediatric test set comprised 59 CT scans in 59 patients (30 boys, 29 girls; mean age, 13.1 years; age range, 4-17 years), which were obtained from November 30, 2018, to August 31, 2020; lung nodules were annotated by fellowship-trained pediatric radiologists as the reference standard. The adult test set was the publicly available adult Lung Nodule Analysis (LUNA) 2016 subset 0, which contained 89 deidentified scans with previously annotated nodules. The test sets were processed through the traditional FlyerScan (github.com/rhardie1/FlyerScanCT) and deep learning Medical Open Network for Artificial Intelligence (MONAI; github.com/Project-MONAI/model-zoo/releases) lung nodule CAD systems, which had been trained on separate sets of CT scans in adults. Sensitivity and false-positive (FP) frequency were calculated for nodules measuring 3-30 mm; nonoverlapping 95% CIs indicated significant differences. RESULTS. Operating at two FPs per scan, on pediatric testing data FlyerScan and MONAI showed significantly lower detection sensitivities of 68.4% (197/288; 95% CI, 65.1-73.0%) and 53.1% (153/288; 95% CI, 46.7-58.4%), respectively, than on adult LUNA 2016 subset 0 testing data (83.9% [94/112; 95% CI, 79.1-88.0%] and 95.5% [107/112; 95% CI, 90.0-98.4%], respectively). Mean nodule size was smaller (p < .001) in the pediatric testing data (5.4 ± 3.1 [SD] mm) than in the adult LUNA 2016 subset 0 testing data (11.0 ± 6.2 mm). CONCLUSION. Adult-trained traditional and deep learning-based lung nodule CAD systems had significantly lower sensitivity for detection on pediatric data than on adult data at a matching FP frequency. The performance difference may relate to the smaller size of pediatric lung nodules. CLINICAL IMPACT. The results indicate a need for pediatric-specific lung nodule CAD systems trained on data specific to pediatric patients.

Generative Adversarial Network (Generative Artificial Intelligence) in Pediatric Radiology: A Systematic Review

Generative artificial intelligence, especially with regard to the generative adversarial network (GAN), is an important research area in radiology as evidenced by a number of literature reviews on the role of GAN in radiology published in the last few years. However, no review article about GAN in pediatric radiology has been published yet. The purpose of this paper is to systematically review applications of GAN in pediatric radiology, their performances, and methods for their performance evaluation. Electronic databases were used for a literature search on 6 April 2023. Thirty-seven papers met the selection criteria and were included. This review reveals that the GAN can be applied to magnetic resonance imaging, X-ray, computed tomography, ultrasound and positron emission tomography for image translation, segmentation, reconstruction, quality assessment, synthesis and data augmentation, and disease diagnosis. About 80% of the included studies compared their GAN model performances with those of other approaches and indicated that their GAN models outperformed the others by 0.1-158.6%. However, these study findings should be used with caution because of a number of methodological weaknesses. For future GAN studies, more robust methods will be essential for addressing these issues. Otherwise, this would affect the clinical adoption of the GAN-based applications in pediatric radiology and the potential advantages of GAN could not be realized widely.

Diagnostic Performance of Artificial Intelligence-Based Computer-Aided Detection and Diagnosis in Pediatric Radiology: A Systematic Review

Artificial intelligence (AI)-based computer-aided detection and diagnosis (CAD) is an important research area in radiology. However, only two narrative reviews about general uses of AI in pediatric radiology and AI-based CAD in pediatric chest imaging have been published yet. The purpose of this systematic review is to investigate the AI-based CAD applications in pediatric radiology, their diagnostic performances and methods for their performance evaluation. A literature search with the use of electronic databases was conducted on 11 January 2023. Twenty-three articles that met the selection criteria were included. This review shows that the AI-based CAD could be applied in pediatric brain, respiratory, musculoskeletal, urologic and cardiac imaging, and especially for pneumonia detection. Most of the studies (93.3%, 14/15; 77.8%, 14/18; 73.3%, 11/15; 80.0%, 8/10; 66.6%, 2/3; 84.2%, 16/19; 80.0%, 8/10) reported model performances of at least 0.83 (area under receiver operating characteristic curve), 0.84 (sensitivity), 0.80 (specificity), 0.89 (positive predictive value), 0.63 (negative predictive value), 0.87 (accuracy), and 0.82 (F1 score), respectively. However, a range of methodological weaknesses (especially a lack of model external validation) are found in the included studies. In the future, more AI-based CAD studies in pediatric radiology with robust methodology should be conducted for convincing clinical centers to adopt CAD and realizing its benefits in a wider context.

Possible Bias in Supervised Deep Learning Algorithms for CT Lung Nodule Detection and Classification

Simple Summary

Artificial Intelligence (AI) algorithms can assist clinicians in their daily tasks by automatically detecting and/or classifying nodules in chest CT scans. Bias of such algorithms is one of the reasons why implementation of them in clinical practice is still not widely adopted. There is no published review on the bias that these algorithms may contain. This review aims to present different types of bias in such algorithms and present possible ways to mitigate them. Only then it would be possible to ensure that these algorithms work as intended under many different clinical settings.

Abstract

Artificial Intelligence (AI) algorithms for automatic lung nodule detection and classification can assist radiologists in their daily routine of chest CT evaluation. Even though many AI algorithms for these tasks have already been developed, their implementation in the clinical workflow is still largely lacking. Apart from the significant number of false-positive findings, one of the reasons for that is the bias that these algorithms may contain. In this review, different types of biases that may exist in chest CT AI nodule detection and classification algorithms are listed and discussed. Examples from the literature in which each type of bias occurs are presented, along with ways to mitigate these biases. Different types of biases can occur in chest CT AI algorithms for lung nodule detection and classification. Mitigation of them can be very difficult, if not impossible to achieve completely.

One-stop local and whole-body staging of children with cancer

Accurate staging and re-staging of cancer in children is crucial for patient management. Currently, children with a newly diagnosed cancer must undergo a series of imaging tests, which are stressful, time-consuming, partially redundant, expensive, and can require repetitive anesthesia. New approaches for pediatric cancer staging can evaluate the primary tumor and metastases in a single session. However, traditional one-stop imaging tests, such as CT and positron emission tomography (PET)/CT, are associated with considerable radiation exposure. This is particularly concerning for children because they are more sensitive to ionizing radiation than adults and they live long enough to experience secondary cancers later in life. In this review article we discuss child-tailored imaging tests for tumor detection and therapy response assessment - tests that can be obtained with substantially reduced radiation exposure compared to traditional CT and PET/CT scans. This includes diffusion-weighted imaging (DWI)/MRI and integrated [F-18]2-fluoro-2-deoxyglucose (18F-FDG) PET/MRI scans. While several investigators have compared the value of DWI/MRI and 18F-FDG PET/MRI for staging pediatric cancer, the value of these novel imaging technologies for cancer therapy monitoring has received surprisingly little attention. In this article, we share our experiences and review existing literature on this subject.

Pediatric Pulmonary Nodules: Imaging Guidelines and Recommendations

Incidental pulmonary nodules are not infrequently identified on computed tomography imaging in the pediatric population and can be a challenge in suggesting appropriate follow-up recommendations. An evidence-based and practical imaging approach for diagnosis and appropriate directed management is essential for optimal patient care. This article provides an up-to-date review of the pediatric pulmonary nodule literature and suggests a practical algorithm to manage pulmonary nodules in the pediatric population.

On the performance of lung nodule detection, segmentation and classification

Computed tomography (CT) screening is an effective way for early detection of lung cancer in order to improve the survival rate of such a deadly disease. For more than two decades, image processing techniques such as nodule detection, segmentation, and classification have been extensively studied to assist physicians in identifying nodules from hundreds of CT slices to measure shapes and HU distributions of nodules automatically and to distinguish their malignancy. Thanks to new parallel computation, multi-layer convolution, nonlinear pooling operation, and the big data learning strategy, recent development of deep-learning algorithms has shown great progress in lung nodule screening and computer-assisted diagnosis (CADx) applications due to their high sensitivity and low false positive rates. This paper presents a survey of state-of-the-art deep-learning-based lung nodule screening and analysis techniques focusing on their performance and clinical applications, aiming to help better understand the current performance, the limitation, and the future trends of lung nodule analysis.

Artificial intelligence applications for pediatric oncology imaging

Machine learning algorithms can help to improve the accuracy and efficiency of cancer diagnosis, selection of personalized therapies and prediction of long-term outcomes. Artificial intelligence (AI) describes a subset of machine learning that can identify patterns in data and take actions to reach pre-set goals without specific programming. Machine learning tools can help to identify high-risk populations, prescribe personalized screening tests and enrich patient populations that are most likely to benefit from advanced imaging tests. AI algorithms can also help to plan personalized therapies and predict the impact of genomic variations on the sensitivity of normal and tumor tissue to chemotherapy or radiation therapy. The two main bottlenecks for successful AI applications in pediatric oncology imaging to date are the needs for large data sets and appropriate computer and memory power. With appropriate data entry and processing power, deep convolutional neural networks (CNNs) can process large amounts of imaging data, clinical data and medical literature in very short periods of time and thereby accelerate literature reviews, correct diagnoses and personalized treatments. This article provides a focused review of emerging AI applications that are relevant for the pediatric oncology imaging community.

Frequency and characteristics of pulmonary nodules in children at computed tomography

BACKGROUND

Normative data on pulmonary nodules in children without malignancy are limited. Knowledge of the frequency and characteristics of pulmonary nodules in healthy children can influence care decisions in children with malignant disease.

OBJECTIVE

To provide normative data concerning the frequency and characteristics of pulmonary nodules on computed tomography (CT) in young children.

MATERIALS AND METHODS

All children ages 1 year-12 years who underwent chest CT after high-energy trauma were retrospectively investigated. Exclusion criteria were a history of malignancy, thick image slices, motion artefacts and extensive post-traumatic pulmonary changes. Two radiologists were asked to independently identify all nodules and to characterize each nodule with respect to location, size, perifissural location and calcification. Discrepancies were adjudicated by a third reader, who set the reference standard in this study. Interobserver agreement in detection and characterization was assessed using the kappa coefficient (κ).

RESULTS

Identified were 120 patients, of whom 72 (75% male; median age: 8.0 years [interquartile range: 4-11]) were included. A total of 59 pulmonary nodules were present in 27 patients (38%; 95% confidence interval: 26-49%; range: 1-5 nodules per patient, with a mean diameter of 3.2 mm [standard deviation: 0.9 mm]). For nodule detection, the per-patient interobserver agreement was substantial (κ=0.78) and per-lobe agreement was moderate (κ=0.40). For characterization, there was fair to substantial agreement (κ=0.36-0.74).

CONCLUSION

Small pulmonary nodules on chest CT are a common finding in otherwise healthy children, but detection and characterization have only moderate interobserver agreement.

Evaluation of pediatric thoracic disorders: comparison of unenhanced fast-imaging-sequence 1.5-T MRI and contrast-enhanced MDCT

OBJECTIVE

The purpose of this study was to investigate the efficacy of thoracic MRI with fast imaging sequences without contrast administration at 1.5 T for evaluating thoracic abnormalities by comparing MRI findings with contrast-enhanced MDCT findings.

SUBJECTS AND METHODS

A prospective study included consecutively registered pediatric patients who from December 2009 to January 2012 underwent thoracic MDCT followed within 2 days by MRI for evaluation of thoracic abnormalities. The final study sample consisted of 71 children (36 boys, 35 girls; mean age, 8.6 ± 4.5 years; range, 2 months-16 years) and 71 paired thoracic MRI and MDCT studies. Thoracic MRI was performed in the axial and coronal planes with the following fast imaging sequences: T1-weighted fast-field echo inversion prepulse, T2-weighted balanced fast-field echo multiple 2D, T1- and T2-weighted turbo spin-echo cardiac-triggering parallel imaging technique without cardiac monitoring, and STIR. Thoracic MDCT was performed with i.v. contrast administration. Two pediatric radiologists independently reviewed each MRI and MDCT study for abnormalities in the lung, large airways, and mediastinal, pleural, and musculoskeletal structures. The sensitivity, specificity, and overall accuracy of MRI were calculated. Interobserver agreement was measured with the kappa coefficient.

RESULTS

With MDCT as the reference standard, 51 of 71 (72%) patients had abnormal findings on MDCT studies, including infections in 21 (42%) cases, neoplasms in 19 (37%) cases, interstitial lung disease in seven (14%) cases, pleural effusion in three (6%) cases, and congenital bronchogenic cyst in one (2%) case. The overall diagnostic accuracy, sensitivity, and specificity of MRI for detecting thoracic abnormalities were 69 of 71 (97%), 49 of 51 (96%), and 20 of 20 (100%). Two undiagnosed findings with MRI that were detected with MDCT were mild bronchiectasis and small pulmonary nodule (3 mm). Almost perfect interobserver agreement was found between two reviewers with 70 of 71 agreements (κ = 0.97; 95% CI, 0.92-1.00; p < 0.001). CONCLUSION; MRI with fast imaging sequences without contrast administration is comparable to contrast-enhanced MDCT for detecting thoracic abnormalities in pediatric patients. Use of MRI with fast imaging sequences without contrast administration as a first-line cross-sectional imaging study in lieu of contrast-enhanced MDCT has the potential to benefit this patient population owing to reduced radiation exposure and i.v. contrast administration.

Advanced functional thoracic imaging in children: from basic concepts to clinical applications

The lungs and airways are organs involved in fairly complex body functions, including ventilation, perfusion, respiratory motion and gas exchange. Imaging evaluation of the pediatric thorax is challenging because involuntary, nonsynchronous respiratory motions and cardiac pulsations degrade image quality appreciably. The extraction of clinically useful functional information from noninvasive imaging methods has been realized even in children thanks to recent technical advancements in thoracic imaging modalities. In this article, advanced functional thoracic imaging techniques in children, focusing on CT and MRI, will be explored from basic concepts to clinical applications.

Newer CT applications and their alternatives: what is appropriate in children?

Innovations in image acquisition and reconstruction technologies have greatly expanded the range of CT applications available in the routine clinical setting. CT images of sub-millimeter resolution can now be acquired of entire body regions in a few seconds or even sub-second time, allowing depiction of fine anatomical detail uncompromised by motion artifact. With sophisticated visualization software, image data can be processed into multiplanar, volume-rendered, cine and other formats to better display anatomical abnormalities and facilitate newer applications such as CT angiography, enterography, urography, tracheobronchography and cardiac CT. Newer applications including dual-energy material decomposition CT are furthering the transition of CT from a purely morphological to a combined anatomical, functional and metabolic imaging technique. These newer applications have largely been pioneered in adult populations, and heightened concern of the risk of carcinogenesis from ionizing radiation tempers dissemination of their use in children. Similar information can often be gleaned from alternative imaging modalities without ionizing radiation exposure, such as MRI and US, and what is most appropriate in children will depend on relative diagnostic efficacy, cost, availability and local expertise.

Lung nodules in pediatric oncology patients: a prediction rule for when to biopsy

PURPOSE

The purpose of the study was to develop a prediction rule regarding the factors that most accurately predict the diagnosis of a malignancy in a lung nodule in the pediatric oncology patient.

METHODS

A retrospective review of pediatric oncology patients that underwent lung nodule resection between 1998 and 2007 was performed. Multivariable logistic regression was used to create a prediction rule.

RESULTS

Fifty pediatric oncology patients underwent 21 thoracotomies and 48 thoracoscopies to resect discrete lung nodules seen on computed tomographic scans during workup for metastasis or routine surveillance. The mean nodule size was 10.43 ± 7.08 mm. The most significant predictors of malignancy in a nodule were peripheral location (odds ratio [OR], 9.1); size between 5 and 10 mm (OR, 2.78); location within the right lower lobe (OR, 2.43); and patients with osteosarcoma (OR, 10.8), Ewing sarcoma (OR, 3.05), or hepatocellular carcinoma (OR, 2.38).

CONCLUSIONS

Lesions that are between 5 and 10 mm in size and peripherally located in patients with osteosarcoma, Ewing sarcoma, or hepatocellular carcinoma are most likely to be malignant. Use of a prediction rule can help guide clinical practice by determining which patients should undergo surgical resection of lung nodules and which patients may be closely observed with continued radiologic studies.