Deep learning approaches for automated classification and segmentation of head and neck cancers and brain tumors in magnetic resonance images: a meta-analysis study

Auto-segmentation of Adult-Type Diffuse Gliomas: Comparison of Transfer Learning-Based Convolutional Neural Network Model vs. Radiologists

Segmentation of glioma is crucial for quantitative brain tumor assessment, to guide therapeutic research and clinical management, but very time-consuming. Fully automated tools for the segmentation of multi-sequence MRI are needed. We developed and pretrained a deep learning (DL) model using publicly available datasets A (n = 210) and B (n = 369) containing FLAIR, T2WI, and contrast-enhanced (CE)-T1WI. This was then fine-tuned with our institutional dataset (n = 197) containing ADC, T2WI, and CE-T1WI, manually annotated by radiologists, and split into training (n = 100) and testing (n = 97) sets. The Dice similarity coefficient (DSC) was used to compare model outputs and manual labels. A third independent radiologist assessed segmentation quality on a semi-quantitative 5-scale score. Differences in DSC between new and recurrent gliomas, and between uni or multifocal gliomas were analyzed using the Mann-Whitney test. Semi-quantitative analyses were compared using the chi-square test. We found that there was good agreement between segmentations from the fine-tuned DL model and ground truth manual segmentations (median DSC: 0.729, std-dev: 0.134). DSC was higher for newly diagnosed (0.807) than recurrent (0.698) (p < 0.001), and higher for unifocal (0.747) than multi-focal (0.613) cases (p = 0.001). Semi-quantitative scores of DL and manual segmentation were not significantly different (mean: 3.567 vs. 3.639; 93.8% vs. 97.9% scoring ≥ 3, p = 0.107). In conclusion, the proposed transfer learning DL performed similarly to human radiologists in glioma segmentation on both structural and ADC sequences. Further improvement in segmenting challenging postoperative and multifocal glioma cases is needed.

Application of Machine Learning for Classification of Brain Tumors: A Systematic Review and Meta-Analysis

BACKGROUND

Classifying brain tumors accurately is crucial for treatment and prognosis. Machine learning (ML) shows great promise in improving tumor classification accuracy. This study evaluates ML algorithms for differentiating various brain tumor types.

METHODS

A systematic review and meta-analysis were conducted, searching PubMed, Embase, and Web of Science up to March 14, 2023. Studies that only investigated image segmentation accuracy or brain tumor detection instead of classification were excluded. We extracted binary diagnostic accuracy data, constructing contingency tables to derive sensitivity and specificity.

RESULTS

Fifty-one studies were included. The pooled area under the curve for glioblastoma versus lymphoma and low-grade versus high-grade gliomas were 0.99 (95% confidence interval [CI]: 0.98-1.00) and 0.89, respectively. The pooled sensitivity and specificity for benign versus malignant tumors were 0.90 (95% CI: 0.85-0.93) and 0.93 (95% CI: 0.90-0.95), respectively. The pooled sensitivity and specificity for low-grade versus high-grade gliomas were 0.99 (95% CI: 0.97-1.00) and 0.94, (95% CI: 0.79-0.99), respectively. Primary versus metastatic tumor identification yields sensitivity and specificity of 0.89, (95% CI: 0.83-0.93) and 0.87 (95% CI: 0.82-0.91), correspondingly. The differentiation of gliomas from pituitary tumors yielded the highest results among primary brain tumor classifications: sensitivity of 0.99 (95% CI: 0.99-1.00) and specificity of 0.99 (95% CI: 0.98-1.00).

CONCLUSIONS

ML demonstrated excellent performance in classifying brain tumor images, with near-maximum area under the curves, sensitivity, and specificity.

Deep Learning for Nasopharyngeal Carcinoma Segmentation in Magnetic Resonance Imaging: A Systematic Review and Meta-Analysis

Nasopharyngeal carcinoma is a significant health challenge that is particularly prevalent in Southeast Asia and North Africa. MRI is the preferred diagnostic tool for NPC due to its superior soft tissue contrast. The accurate segmentation of NPC in MRI is crucial for effective treatment planning and prognosis. We conducted a search across PubMed, Embase, and Web of Science from inception up to 20 March 2024, adhering to the PRISMA 2020 guidelines. Eligibility criteria focused on studies utilizing DL for NPC segmentation in adults via MRI. Data extraction and meta-analysis were conducted to evaluate the performance of DL models, primarily measured by Dice scores. We assessed methodological quality using the CLAIM and QUADAS-2 tools, and statistical analysis was performed using random effects models. The analysis incorporated 17 studies, demonstrating a pooled Dice score of 78% for DL models (95% confidence interval: 74% to 83%), indicating a moderate to high segmentation accuracy by DL models. Significant heterogeneity and publication bias were observed among the included studies. Our findings reveal that DL models, particularly convolutional neural networks, offer moderately accurate NPC segmentation in MRI. This advancement holds the potential for enhancing NPC management, necessitating further research toward integration into clinical practice.

Evaluation of a nnU-Net type automated clinical volumetric tumor segmentation tool for diffuse low-grade glioma follow-up

BACKGROUND AND PURPOSE

Diffuse low-grade gliomas (DLGG) are characterized by a slow and continuous growth and always evolve towards an aggressive grade. Accurate prediction of the malignant transformation is essential as it requires immediate therapeutic intervention. One of its most precise predictors is the velocity of diameter expansion (VDE). Currently, the VDE is estimated either by linear measurements or by manual delineation of the DLGG on T2 FLAIR acquisitions. However, because of the DLGG's infiltrative nature and its blurred contours, manual measures are challenging and variable, even for experts. Therefore we propose an automated segmentation algorithm using a 2D nnU-Net, to 1) gain time and 2) standardize VDE assessment.

MATERIALS AND METHODS

The 2D nnU-Net was trained on 318 acquisitions (T2 FLAIR & 3DT1 longitudinal follow-up of 30 patients, including pre- & post-surgery acquisitions, different scanners, vendors, imaging parameters…). Automated vs. manual segmentation performance was evaluated on 167 acquisitions, and its clinical interest was validated by quantifying the amount of manual correction required after automated segmentation of 98 novel acquisitions.

RESULTS

Automated segmentation showed a good performance with a mean Dice Similarity Coefficient (DSC) of 0.82±0.13 with manual segmentation and a substantial concordance between VDE calculations. Major manual corrections (i.e., DSC<0.7) were necessary only in 3/98 cases and 81% of the cases had a DSC>0.9.

CONCLUSION

The proposed automated segmentation algorithm can successfully segment DLGG on highly variable MRI data. Although manual corrections are sometimes necessary, it provides a reliable, standardized and time-winning support for VDE extraction to asses DLGG growth.

A Fully Automated Deep-Learning Model for Predicting the Molecular Subtypes of Posterior Fossa Ependymomas Using T2-Weighted Images

PURPOSE

We aimed to develop and validate a deep learning (DL) model to automatically segment posterior fossa ependymoma (PF-EPN) and predict its molecular subtypes [Group A (PFA) and Group B (PFB)] from preoperative MR images.

EXPERIMENTAL DESIGN

We retrospectively identified 227 PF-EPNs (development and internal test sets) with available preoperative T2-weighted (T2w) MR images and molecular status to develop and test a 3D nnU-Net (referred to as T2-nnU-Net) for tumor segmentation and molecular subtype prediction. The network was externally tested using an external independent set [n = 40; subset-1 (n = 31) and subset-2 (n =9)] and prospectively enrolled cases [prospective validation set (n = 27)]. The Dice similarity coefficient was used to evaluate the segmentation performance. Receiver operating characteristic analysis for molecular subtype prediction was performed.

RESULTS

For tumor segmentation, the T2-nnU-Net achieved a Dice score of 0.94 ± 0.02 in the internal test set. For molecular subtype prediction, the T2-nnU-Net achieved an AUC of 0.93 and accuracy of 0.89 in the internal test set, an AUC of 0.99 and accuracy of 0.93 in the external test set. In the prospective validation set, the model achieved an AUC of 0.93 and an accuracy of 0.89. The predictive performance of T2-nnU-Net was superior or comparable to that of demographic and multiple radiologic features (AUCs ranging from 0.87 to 0.95).

CONCLUSIONS

A fully automated DL model was developed and validated to accurately segment PF-EPNs and predict molecular subtypes using only T2w MR images, which could help in clinical decision-making.

Automatic Segmentation with Deep Learning in Radiotherapy

This review provides a formal overview of current automatic segmentation studies that use deep learning in radiotherapy. It covers 807 published papers and includes multiple cancer sites, image types (CT/MRI/PET), and segmentation methods. We collect key statistics about the papers to uncover commonalities, trends, and methods, and identify areas where more research might be needed. Moreover, we analyzed the corpus by posing explicit questions aimed at providing high-quality and actionable insights, including: "What should researchers think about when starting a segmentation study?", "How can research practices in medical image segmentation be improved?", "What is missing from the current corpus?", and more. This allowed us to provide practical guidelines on how to conduct a good segmentation study in today's competitive environment that will be useful for future research within the field, regardless of the specific radiotherapeutic subfield. To aid in our analysis, we used the large language model ChatGPT to condense information.

Artificial Intelligence in Neuroradiology: A Review of Current Topics and Competition Challenges

There is an expanding body of literature that describes the application of deep learning and other machine learning and artificial intelligence methods with potential relevance to neuroradiology practice. In this article, we performed a literature review to identify recent developments on the topics of artificial intelligence in neuroradiology, with particular emphasis on large datasets and large-scale algorithm assessments, such as those used in imaging AI competition challenges. Numerous applications relevant to ischemic stroke, intracranial hemorrhage, brain tumors, demyelinating disease, and neurodegenerative/neurocognitive disorders were discussed. The potential applications of these methods to spinal fractures, scoliosis grading, head and neck oncology, and vascular imaging were also reviewed. The AI applications examined perform a variety of tasks, including localization, segmentation, longitudinal monitoring, diagnostic classification, and prognostication. While research on this topic is ongoing, several applications have been cleared for clinical use and have the potential to augment the accuracy or efficiency of neuroradiologists.

A convolutional neural network with self-attention for fully automated metabolic tumor volume delineation of head and neck cancer in [Formula: see text]F]FDG PET/CT

PURPOSE

PET-derived metabolic tumor volume (MTV) and total lesion glycolysis of the primary tumor are known to be prognostic of clinical outcome in head and neck cancer (HNC). Including evaluation of lymph node metastases can further increase the prognostic value of PET but accurate manual delineation and classification of all lesions is time-consuming and prone to interobserver variability. Our goal, therefore, was development and evaluation of an automated tool for MTV delineation/classification of primary tumor and lymph node metastases in PET/CT investigations of HNC patients.

METHODS

Automated lesion delineation was performed with a residual 3D U-Net convolutional neural network (CNN) incorporating a multi-head self-attention block. 698 [Formula: see text]F]FDG PET/CT scans from 3 different sites and 5 public databases were used for network training and testing. An external dataset of 181 [Formula: see text]F]FDG PET/CT scans from 2 additional sites was employed to assess the generalizability of the network. In these data, primary tumor and lymph node (LN) metastases were interactively delineated and labeled by two experienced physicians. Performance of the trained network models was assessed by 5-fold cross-validation in the main dataset and by pooling results from the 5 developed models in the external dataset. The Dice similarity coefficient (DSC) for individual delineation tasks and the primary tumor/metastasis classification accuracy were used as evaluation metrics. Additionally, a survival analysis using univariate Cox regression was performed comparing achieved group separation for manual and automated delineation, respectively.

RESULTS

In the cross-validation experiment, delineation of all malignant lesions with the trained U-Net models achieves DSC of 0.885, 0.805, and 0.870 for primary tumor, LN metastases, and the union of both, respectively. In external testing, the DSC reaches 0.850, 0.724, and 0.823 for primary tumor, LN metastases, and the union of both, respectively. The voxel classification accuracy was 98.0% and 97.9% in cross-validation and external data, respectively. Univariate Cox analysis in the cross-validation and the external testing reveals that manually and automatically derived total MTVs are both highly prognostic with respect to overall survival, yielding essentially identical hazard ratios (HR) ([Formula: see text]; [Formula: see text] vs. [Formula: see text]; [Formula: see text] in cross-validation and [Formula: see text]; [Formula: see text] vs. [Formula: see text]; [Formula: see text] in external testing).

CONCLUSION

To the best of our knowledge, this work presents the first CNN model for successful MTV delineation and lesion classification in HNC. In the vast majority of patients, the network performs satisfactory delineation and classification of primary tumor and lymph node metastases and only rarely requires more than minimal manual correction. It is thus able to massively facilitate study data evaluation in large patient groups and also does have clear potential for supervised clinical application.

Deep learning algorithm for predicting subacromial motion trajectory: Dynamic shoulder ultrasound analysis

Subacromial motion metrics can be extracted from dynamic shoulder ultrasonography, which is useful for identifying abnormal motion patterns in painful shoulders. However, frame-by-frame manual labeling of anatomical landmarks in ultrasound images is time consuming. The present study aims to investigate the feasibility of a deep learning algorithm for extracting subacromial motion metrics from dynamic ultrasonography. Dynamic ultrasound imaging was retrieved by asking 17 participants to perform cyclic shoulder abduction and adduction along the scapular plane, whereby the trajectory of the humeral greater tubercle (in relation to the lateral acromion) was depicted by the deep learning algorithm. Extraction of the subacromial motion metrics was conducted using a convolutional neural network (CNN) or a self-transfer learning-based (STL)-CNN with or without an autoencoder (AE). The mean absolute error (MAE) compared with the manually-labeled data (ground truth) served as the main outcome variable. Using eight-fold cross-validation, the average MAE was proven to be significantly higher in the group using CNN than in those using STL-CNN or STL-CNN+AE for the relative difference between the greater tubercle and lateral acromion on the horizontal axis. The MAE for the localization of the two aforementioned landmarks on the vertical axis also seemed to be enlarged in those using CNN compared with those using STL-CNN. In the testing dataset, the errors in relation to the ground truth for the minimal vertical acromiohumeral distance were 0.081-0.333 cm using CNN, compared with 0.002-0.007 cm using STL-CNN. We successfully demonstrated the feasibility of a deep learning algorithm for automatic detection of the greater tubercle and lateral acromion during dynamic shoulder ultrasonography. Our framework also demonstrated the capability of capturing the minimal vertical acromiohumeral distance, which is the most important indicator of subacromial motion metrics in daily clinical practice.

Deep learning in CT image segmentation of cervical cancer: a systematic review and meta-analysis

Background

This paper attempts to conduct a systematic review and meta-analysis of deep learning (DLs) models for cervical cancer CT image segmentation.

Methods

Relevant studies were systematically searched in PubMed, Embase, The Cochrane Library, and Web of science. The literature on DLs for cervical cancer CT image segmentation were included, a meta-analysis was performed on the dice similarity coefficient (DSC) of the segmentation results of the included DLs models. We also did subgroup analyses according to the size of the sample, type of segmentation (i.e., two dimensions and three dimensions), and three organs at risk (i.e., bladder, rectum, and femur). This study was registered in PROSPERO prior to initiation (CRD42022307071).

Results

A total of 1893 articles were retrieved and 14 articles were included in the meta-analysis. The pooled effect of DSC score of clinical target volume (CTV), bladder, rectum, femoral head were 0.86(95%CI 0.84 to 0.87), 0.91(95%CI 0.89 to 0.93), 0.83(95%CI 0.79 to 0.88), and 0.92(95%CI 0.91to 0.94), respectively. For the performance of segmented CTV by two dimensions (2D) and three dimensions (3D) model, the DSC score value for 2D model was 0.87 (95%CI 0.85 to 0.90), while the DSC score for 3D model was 0.85 (95%CI 0.82 to 0.87). As for the effect of the capacity of sample on segmentation performance, no matter whether the sample size is divided into two groups: greater than 100 and less than 100, or greater than 150 and less than 150, the results show no difference (P > 0.05). Four papers reported the time for segmentation from 15 s to 2 min.

Conclusion

DLs have good accuracy in automatic segmentation of CT images of cervical cancer with a less time consuming and have good prospects for future radiotherapy applications, but still need public high-quality databases and large-scale research verification.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13014-022-02148-6.

Gaps and Opportunities of Artificial Intelligence Applications for Pediatric Oncology in European Research: A Systematic Review of Reviews and a Bibliometric Analysis

The application of artificial intelligence (AI) systems is emerging in many fields in recent years, due to the increased computing power available at lower cost. Although its applications in various branches of medicine, such as pediatric oncology, are many and promising, its use is still in an embryonic stage. The aim of this paper is to provide an overview of the state of the art regarding the AI application in pediatric oncology, through a systematic review of systematic reviews, and to analyze current trends in Europe, through a bibliometric analysis of publications written by European authors. Among 330 records found, 25 were included in the systematic review. All papers have been published since 2017, demonstrating only recent attention to this field. The total number of studies included in the selected reviews was 674, with a third including an author with a European affiliation. In bibliometric analysis, 304 out of the 978 records found were included. Similarly, the number of publications began to dramatically increase from 2017. Most explored AI applications regard the use of diagnostic images, particularly radiomics, as well as the group of neoplasms most involved are the central nervous system tumors. No evidence was found regarding the use of AI for process mining, clinical pathway modeling, or computer interpreted guidelines to improve the healthcare process. No robust evidence is yet available in any of the domains investigated by systematic reviews. However, the scientific production in Europe is significant and consistent with the topics covered in systematic reviews at the global level. The use of AI in pediatric oncology is developing rapidly with promising results, but numerous gaps and challenges persist to validate its utilization in clinical practice. An important limitation is the need for large datasets for training algorithms, calling for international collaborative studies.

Methods of 3D printing models of pituitary tumors

Background

Pituitary adenomas can give rise to a variety of clinical disorders and surgery is often the primary treatment option. However, preoperative magnetic resonance imaging (MRI) does not always reliably identify the site of an adenoma. In this setting molecular (functional) imaging (e.g. ¹¹C-methionine PET/CT) may help with tumor localisation, although interpretation of these 2D images can be challenging. 3D printing of anatomicalal models for other indications has been shown to aid surgical planning and improve patient understanding of the planned procedure. Here, we explore the potential utility of four types of 3D printing using PET/CT and co-registered MRI for visualising pituitary adenomas.

Methods

A 3D patient-specific model based on a challenging clinical case was created by segmenting the pituitary gland, pituitary adenoma, carotid arteries and bone using contemporary PET/CT and MR images. The 3D anatomical models were printed using VP, MEX, MJ and PBF 3D printing methods. Different anatomicalal structures were printed in color with the exception of the PBF anatomical model where a single color was used. The anatomical models were compared against the computer model to assess printing accuracy. Three groups of clinicians (endocrinologists, neurosurgeons and ENT surgeons) assessed the anatomical models for their potential clinical utility.

Results

All of the printing techniques produced anatomical models which were spatially accurate, with the commercial printing techniques (MJ and PBF) and the consumer printing techniques (VP and MEX) demonstrating comparable findings (all techniques had mean spatial differences from the computer model of < 0.6 mm). The MJ, VP and MEX printing techniques yielded multicolored anatomical models, which the clinicians unanimously agreed would be preferable to use when talking to a patient; in contrast, 50%, 40% and 0% of endocrinologists, neurosurgeons and ENT surgeons respectively would consider using the PBF model.

Conclusion

3D anatomical models of pituitary tumors were successfully created from PET/CT and MRI using four different 3D printing techniques. However, the expert reviewers unanimously preferred the multicolor prints. Importantly, the consumer printers performed comparably to the commercial MJ printing technique, opening the possibility that these methods can be adopted into routine clinical practice with only a modest investment.

Fully automatic brain tumor segmentation for 3D evaluation in augmented reality

OBJECTIVE

For currently available augmented reality workflows, 3D models need to be created with manual or semiautomatic segmentation, which is a time-consuming process. The authors created an automatic segmentation algorithm that generates 3D models of skin, brain, ventricles, and contrast-enhancing tumor from a single T1-weighted MR sequence and embedded this model into an automatic workflow for 3D evaluation of anatomical structures with augmented reality in a cloud environment. In this study, the authors validate the accuracy and efficiency of this automatic segmentation algorithm for brain tumors and compared it with a manually segmented ground truth set.

METHODS

Fifty contrast-enhanced T1-weighted sequences of patients with contrast-enhancing lesions measuring at least 5 cm3 were included. All slices of the ground truth set were manually segmented. The same scans were subsequently run in the cloud environment for automatic segmentation. Segmentation times were recorded. The accuracy of the algorithm was compared with that of manual segmentation and evaluated in terms of Sørensen-Dice similarity coefficient (DSC), average symmetric surface distance (ASSD), and 95th percentile of Hausdorff distance (HD95).

RESULTS

The mean ± SD computation time of the automatic segmentation algorithm was 753 ± 128 seconds. The mean ± SD DSC was 0.868 ± 0.07, ASSD was 1.31 ± 0.63 mm, and HD95 was 4.80 ± 3.18 mm. Meningioma (mean 0.89 and median 0.92) showed greater DSC than metastasis (mean 0.84 and median 0.85). Automatic segmentation had greater accuracy for measuring DSC (mean 0.86 and median 0.87) and HD95 (mean 3.62 mm and median 3.11 mm) of supratentorial metastasis than those of infratentorial metastasis (mean 0.82 and median 0.81 for DSC; mean 5.26 mm and median 4.72 mm for HD95).

CONCLUSIONS

The automatic cloud-based segmentation algorithm is reliable, accurate, and fast enough to aid neurosurgeons in everyday clinical practice by providing 3D augmented reality visualization of contrast-enhancing intracranial lesions measuring at least 5 cm3. The next steps involve incorporation of other sequences and improving accuracy with 3D fine-tuning in order to expand the scope of augmented reality workflow.