Whole liver segmentation based on deep learning and manual adjustment for clinical use in SIRT.

PURPOSE: In selective internal radiation therapy (SIRT), an accurate total liver segmentation is required for activity prescription and absorbed dose calculation. Our goal was to investigate the feasibility of using automatic liver segmentation based on a convolutional neural network (CNN) for CT imaging in SIRT, and the ability of CNN to reduce inter-observer variability of the segmentation. METHODS: A multi-scale CNN was modified for liver segmentation for SIRT patients. The CNN model was trained with 139 datasets from three liver segmentation challenges and 12 SIRT patient datasets from our hospital. Validation was performed on 13 SIRT datasets and 12 challenge datasets. The model was tested on 40 SIRT datasets. One expert manually delineated the livers and adjusted the liver segmentations from CNN for 40 test SIRT datasets. Another expert performed the same tasks for 20 datasets randomly selected from the 40 SIRT datasets. The CNN segmentations were compared with the manual and adjusted segmentations from the experts. The difference between the manual segmentations was compared with the difference between the adjusted segmentations to investigate the inter-observer variability. Segmentation difference was evaluated through dice similarity coefficient (DSC), volume ratio (RV), mean surface distance (MSD), and Hausdorff distance (HD). RESULTS: The CNN segmentation achieved a median DSC of 0.94 with the manual segmentation and of 0.98 with the manually corrected CNN segmentation, respectively. The DSC between the adjusted segmentations is 0.98, which is 0.04 higher than the DSC between the manual segmentations. CONCLUSION: The CNN model achieved good liver segmentations on CT images of good image quality, with relatively normal liver shapes and low tumor burden. 87.5% of the 40 CNN segmentations only needed slight adjustments for clinical use. However, the trained model failed on SIRT data with low dose or contrast, lesions with large density difference from their surroundings, and abnormal liver position and shape. The abovementioned scenarios were not adequately represented in the training data. Despite this limitation, the current CNN is already a useful clinical tool which improves inter-observer agreement and therefore contributes to the standardization of the dosimetry. A further improvement is expected when the CNN will be trained with more data from SIRT patients.

The feasibility of atlas-based automatic segmentation of MRI for H&N radiotherapy planning.

Atlas-based autosegmentation is an established tool for segmenting structures for CT-planned head and neck radiotherapy. MRI is being increasingly integrated into the planning process. The aim of this study is to assess the feasibility of MRI-based, atlas-based autosegmentation for organs at risk (OAR) and lymph node levels, and to compare the segmentation accuracy with CT-based autosegmentation. Fourteen patients with locally advanced head and neck cancer in a prospective imaging study underwent a T1-weighted MRI and a PET-CT (with dedicated contrast-enhanced CT) in an immobilization mask. Organs at risk (orbits, parotids, brainstem, and spinal cord) and the left level II lymph node region were manually delineated on the CT and MRI separately. A 'leave one out' approach was used to automatically segment structures onto the remaining images separately for CT and MRI. Contour comparison was performed using multiple positional metrics: Dice index, mean distance to conformity (MDC), sensitivity index (Se Idx), and inclusion index (Incl Idx). Automatic segmentation using MRI of orbits, parotids, brainstem, and lymph node level was acceptable with a DICE coefficient of 0.73-0.91, MDC 2.0-5.1mm, Se Idx 0.64-0.93, Incl Idx 0.76-0.93. Segmentation of the spinal cord was poor (Dice coefficient 0.37). The process of automatic segmentation was significantly better on MRI compared to CT for orbits, parotid glands, brainstem, and left lymph node level II by multiple positional metrics; spinal cord segmentation based on MRI was inferior compared with CT. Accurate atlas-based automatic segmentation of OAR and lymph node levels is feasible using T1-MRI; segmentation of the spinal cord was found to be poor. Comparison with CT-based automatic segmentation suggests that the process is equally as, or more accurate, using MRI. These results support further translation of MRI-based segmentation methodology into clinicalpractice.

Analytic calculations and synthetic shapes for validation of quantitative contour comparison software.

A high level of variability in reported values was observed in a recent survey of contour similarity measures (CSMs) calculation tools. Such variations in the output measurements prevent meaningful comparison between studies. The purpose of this study was to develop a dataset with analytically calculated gold standard values to facilitate standardization and ensure accuracy of CSM implementations. The dataset was generated in the Digital Imaging and Communications in Medicine (DICOM) format. Both the dataset and the software used for its generation are made publicly available to encourage robust testing of CSM implementations for accuracy, improving consistency between different implementations.

Clinical evaluation of atlas-based auto-segmentation in breast and nodal radiotherapy.

OBJECTIVES: Accurate contouring of anatomical structures allows for high-precision radiotherapy planning, targeting the dose at treatment volumes and avoiding organs at risk. Manual contouring is time-consuming with significant user variability, whereas auto-segmentation (AS) has proven efficiency benefits but requires editing before treatment planning. This study investigated whether atlas-based AS (ABAS) accuracy improves with template atlas group size and character-specific atlas and test case selection. METHODS AND MATERIALS: One clinician retrospectively contoured the breast, nodes, lung, heart, and brachial plexus on 100 CT scans, adhering to peer-reviewed guidelines. Atlases were clustered in group sizes, treatment positions, chest wall separations, and ASs created with Mirada software. The similarity of ASs compared to reference contours was described by the Jaccard similarity coefficient (JSC) and centroid distance variance (CDV). RESULTS: Across group sizes, for all structures combined, the mean JSC was 0.6 (SD 0.3, p = .999). Across atlas-specific groups, 0.6 (SD 0.3, p = 1.000). The correlation between JSC and structure volume was weak in both scenarios (adjusted R2-0.007 and 0.185).Mean CDV was similar across groups but varied up to 1.2 cm for specific structures. CONCLUSIONS: Character-specific atlas groups and test case selection did not improve accuracy outcomes. High-quality ASs were obtained from groups containing as few as ten atlases, subsequently simplifying the application of ABAS. CDV measures indicating auto-segmentation variations on the x, y, and z axes can be utilised to decide on the clinical relevance of variations and reduce AS editing. ADVANCES IN KNOWLEDGE: High-quality ABASs can be obtained from as few as ten template atlases.Atlas and test case selection do not improve AS accuracy.Unlike well-known quantitative similarity indices, volume displacement metrics provide information on the location of segmentation variations, helping assessment of the clinical relevance of variations and reducing clinician editing. Volume displacement metrics combined with the qualitative measure of clinician assessment could reduce user variability.

An investigation into the risk of population bias in deep learning autocontouring.

BACKGROUND AND PURPOSE: To date, data used in the development of Deep Learning-based automatic contouring (DLC) algorithms have been largely sourced from single geographic populations. This study aimed to evaluate the risk of population-based bias by determining whether the performance of an autocontouring system is impacted by geographic population. MATERIALS AND METHODS: 80 Head Neck CT deidentified scans were collected from four clinics in Europe (n = 2) and Asia (n = 2). A single observer manually delineated 16 organs-at-risk in each. Subsequently, the data was contoured using a DLC solution, and trained using single institution (European) data. Autocontours were compared to manual delineations using quantitative measures. A Kruskal-Wallis test was used to test for any difference between populations. Clinical acceptability of automatic and manual contours to observers from each participating institution was assessed using a blinded subjective evaluation. RESULTS: Seven organs showed a significant difference in volume between groups. Four organs showed statistical differences in quantitative similarity measures. The qualitative test showed greater variation in acceptance of contouring between observers than between data from different origins, with greater acceptance by the South Korean observers. CONCLUSION: Much of the statistical difference in quantitative performance could be explained by the difference in organ volume impacting the contour similarity measures and the small sample size. However, the qualitative assessment suggests that observer perception bias has a greater impact on the apparent clinical acceptability than quantitatively observed differences. This investigation of potential geographic bias should extend to more patients, populations, and anatomical regions in the future.

Geometric and dosimetric analysis of CT- and MR-based automatic contouring for the EPTN contouring atlas in neuro-oncology.

PURPOSE: Atlas-based and deep-learning contouring (DLC) are methods for automatic segmentation of organs-at-risk (OARs). The European Particle Therapy Network (EPTN) published a consensus-based atlas for delineation of OARs in neuro-oncology. In this study, geometric and dosimetric evaluation of automatically-segmented neuro-oncological OARs was performed using CT- and MR-models following the EPTN-contouring atlas. METHODS: Image and contouring data from 76 neuro-oncological patients were included. Two atlas-based models (CT-atlas and MR-atlas) and one DLC-model (MR-DLC) were created. Manual contours on registered CT-MR-images were used as ground-truth. Results were analyzed in terms of geometrical (volumetric Dice similarity coefficient (vDSC), surface DSC (sDSC), added path length (APL), and mean slice-wise Hausdorff distance (MSHD)) and dosimetrical accuracy. Distance-to-tumor analysis was performed to analyze to which extent the location of the OAR relative to planning target volume (PTV) has dosimetric impact, using Wilcoxon rank-sum tests. RESULTS: CT-atlas outperformed MR-atlas for 22/26 OARs. MR-DLC outperformed MR-atlas for all OARs. Highest median (95 %CI) vDSC and sDSC were found for the brainstem in MR-DLC: 0.92 (0.88-0.95) and 0.84 (0.77-0.89) respectively, as well as lowest MSHD: 0.27 (0.22-0.39)cm. Median dose differences (ΔD) were within ± 1 Gy for 24/26(92 %) OARs for all three models. Distance-to-tumor showed a significant correlation for ΔDmax,0.03cc-parameters when splitting the data in ≤ 4 cm and > 4 cm OAR-distance (p < 0.001). CONCLUSION: MR-based DLC and CT-based atlas-contouring enable high-quality segmentation. It was shown that a combination of both CT- and MR-autocontouring models results in the best quality.

Multicenter comparison of measures for quantitative evaluation of contouring in radiotherapy.

Background and Purpose: A wide range of quantitative measures are available to facilitate clinical implementation of auto-contouring software, on-going Quality Assurance (QA) and interobserver contouring variation studies. This study aimed to assess the variation in output when applying different implementations of the measures to the same data in order to investigate how consistently such measures are defined and implemented in radiation oncology. Materials and Methods: A survey was conducted to assess if there were any differences in definitions of contouring measures or their implementations that would lead to variation in reported results between institutions. This took two forms: a set of computed tomography (CT) image data with "Test" and "Reference" contours was distributed for participants to process using their preferred tools and report results, and a questionnaire regarding the definition of measures and their implementation was completed by the participants. Results: Thirteen participants completed the survey and submitted results, with one commercial and twelve in-house solutions represented. Excluding outliers, variations of up to 50% in Dice Similarity Coefficient (DSC), 50% in 3D Hausdorff Distance (HD), and 200% in Average Distance (AD) were observed between the participant submitted results. Collaborative investigation with participants revealed a large number of bugs in implementation, confounding the understanding of intentional implementation choices. Conclusion: Care must be taken when comparing quantitative results between different studies. There is a need for a dataset with clearly defined measures and ground truth for validation of such tools prior to their use.

Real-world analysis of manual editing of deep learning contouring in the thorax region.

Background and purpose: User-adjustments after deep-learning (DL) contouring in radiotherapy were evaluated to get insight in real-world editing during clinical practice. This study assessed the amount, type and spatial regions of editing of auto-contouring for organs-at-risk (OARs) in routine clinical workflow for patients in the thorax region. Materials and methods: A total of 350 lung cancer and 362 breast cancer patients, contoured between March 2020 and March 2021 using a commercial DL-contouring method followed by manual adjustments were retrospectively analyzed. Subsampling was performed for some OARs, using an inter-slice gap of 1-3 slices. Commonly-used whole-organ contouring assessment measures were calculated, and all cases were registered to a common reference shape per OAR to identify regions of manual adjustment. Results were expressed as the median, 10th-90th percentile of adjustment and visualized using 3D renderings. Results: Per OAR, the median amount of editing was below 1 mm. However, large adjustments were found in some locations for most OARs. In general, enlarging of the auto-contours was needed. Subsampling DL-contours showed less adjustments were made in the interpolated slices compared to simulated no-subsampling for these OARs. Conclusion: The real-world performance of automatic DL-contouring software was evaluated and proven useful in clinical practice. Specific regions-of-adjustment were identified per OAR in the thorax region, and separate models were found to be necessary for specific clinical indications different from training data. This analysis showed the need to perform routine clinical analysis especially when procedures or acquisition protocols change to have the best configuration of the workflow.

Machine learning-based detection of aberrant deep learning segmentations of target and organs at risk for prostate radiotherapy using a secondary segmentation algorithm.

Objective.The output of a deep learning (DL) auto-segmentation application should be reviewed, corrected if needed and approved before being used clinically. This verification procedure is labour-intensive, time-consuming and user-dependent, which potentially leads to significant errors with impact on the overall treatment quality. Additionally, when the time needed to correct auto-segmentations approaches the time to delineate target and organs at risk from scratch, the usability of the DL model can be questioned. Therefore, an automated quality assurance framework was developed with the aim to detect in advance aberrant auto-segmentations.Approach. Five organs (prostate, bladder, anorectum, femoral head left and right) were auto-delineated on CT acquisitions for 48 prostate patients by an in-house trained primary DL model. An experienced radiation oncologist assessed the correctness of the model output and categorised the auto-segmentations into two classes whether minor or major adaptations were needed. Subsequently, an independent, secondary DL model was implemented to delineate the same structures as the primary model. Quantitative comparison metrics were calculated using both models' segmentations and used as input features for a machine learning classification model to predict the output quality of the primary model.Main results. For every organ, the approach of independent validation by the secondary model was able to detect primary auto-segmentations that needed major adaptation with high sensitivity (recall = 1) based on the calculated quantitative metrics. The surface DSC and APL were found to be the most indicated parameters in comparison to standard quantitative metrics for the time needed to adapt auto-segmentations.Significance. This proposed method includes a proof of concept for the use of an independent DL segmentation model in combination with a ML classifier to improve time saving during QA of auto-segmentations. The integration of such system into current automatic segmentation pipelines can increase the efficiency of the radiotherapy contouring workflow.

Beyond automatic medical image segmentation-the spectrum between fully manual and fully automatic delineation.

Semi-automatic and fully automatic contouring tools have emerged as an alternative to fully manual segmentation to reduce time spent contouring and to increase contour quality and consistency. Particularly, fully automatic segmentation has seen exceptional improvements through the use of deep learning in recent years. These fully automatic methods may not require user interactions, but the resulting contours are often not suitable to be used in clinical practice without a review by the clinician. Furthermore, they need large amounts of labelled data to be available for training. This review presents alternatives to manual or fully automatic segmentation methods along the spectrum of variable user interactivity and data availability. The challenge lies to determine how much user interaction is necessary and how this user interaction can be used most effectively. While deep learning is already widely used for fully automatic tools, interactive methods are just at the starting point to be transformed by it. Interaction between clinician and machine, via artificial intelligence, can go both ways and this review will present the avenues that are being pursued to improve medical image segmentation.

Segmentation-guided multi-modal registration of liver images for dose estimation in SIRT.

PURPOSE: Selective internal radiation therapy (SIRT) requires a good liver registration of multi-modality images to obtain precise dose prediction and measurement. This study investigated the feasibility of liver registration of CT and MR images, guided by segmentation of the liver and its landmarks. The influence of the resulting lesion registration on dose estimation was evaluated. METHODS: The liver segmentation was done with a convolutional neural network (CNN), and the landmarks were segmented manually. Our image-based registration software and its liver-segmentation-guided extension (CNN-guided) were tuned and evaluated with 49 CT and 26 MR images from 20 SIRT patients. Each liver registration was evaluated by the root mean square distance (RMSD) of mean surface distance between manually delineated liver contours and mass center distance between manually delineated landmarks (lesions, clips, etc.). The root mean square of RMSDs (RRMSD) was used to evaluate all liver registrations. The CNN-guided registration was further extended by incorporating landmark segmentations (CNN&LM-guided) to assess the value of additional landmark guidance. To evaluate the influence of segmentation-guided registration on dose estimation, mean dose and volume percentages receiving at least 70 Gy (V70) estimated on the 99mTc-labeled macro-aggregated albumin (99mTc-MAA) SPECT were computed, either based on lesions from the reference 99mTc-MAA CT (reference lesions) or from the registered floating CT or MR images (registered lesions) using the CNN- or CNN&LM-guided algorithms. RESULTS: The RRMSD decreased for the floating CTs and MRs by 1.0 mm (11%) and 3.4 mm (34%) using CNN guidance for the image-based registration and by 2.1 mm (26%) and 1.4 mm (21%) using landmark guidance for the CNN-guided registration. The quartiles for the relative mean dose difference (the V70 difference) between the reference and registered lesions and their correlations [25th, 75th; r] are as follows: [- 5.5% (- 1.3%), 5.6% (3.4%); 0.97 (0.95)] and [- 12.3% (- 2.1%), 14.8% (2.9%); 0.96 (0.97)] for the CNN&LM- and CNN-guided CT to CT registrations, [- 7.7% (- 6.6%), 7.0% (3.1%); 0.97 (0.90)] and [- 15.1% (- 11.3%), 2.4% (2.5%); 0.91 (0.78)] for the CNN&LM- and CNN-guided MR to CT registrations. CONCLUSION: Guidance by CNN liver segmentations and landmarks markedly improves the performance of the image-based registration. The small mean dose change between the reference and registered lesions demonstrates the feasibility of applying the CNN&LM- or CNN-guided registration to volume-level dose prediction. The CNN&LM- and CNN-guided registrations for CTs can be applied to voxel-level dose prediction according to their small V70 change for most lesions. The CNN-guided MR to CT registration still needs to incorporate landmark guidance for smaller change of voxel-level dose estimation.

Interactive contouring through contextual deep learning.

PURPOSE: To investigate a deep learning approach that enables three-dimensional (3D) segmentation of an arbitrary structure of interest given a user provided two-dimensional (2D) contour for context. Such an approach could decrease delineation times and improve contouring consistency, particularly for anatomical structures for which no automatic segmentation tools exist. METHODS: A series of deep learning segmentation models using a Recurrent Residual U-Net with attention gates was trained with a successively expanding training set. Contextual information was provided to the models, using a previously contoured slice as an input, in addition to the slice to be contoured. In total, 6 models were developed, and 19 different anatomical structures were used for training and testing. Each of the models was evaluated for all 19 structures, even if they were excluded from the training set, in order to assess the model's ability to segment unseen structures of interest. Each model's performance was evaluated using the Dice similarity coefficient (DSC), Hausdorff distance, and relative added path length (APL). RESULTS: The segmentation performance for seen and unseen structures improved when the training set was expanded by addition of structures previously excluded from the training set. A model trained exclusively on heart structures achieved a DSC of 0.33, HD of 44 mm, and relative APL of 0.85 when segmenting the spleen, whereas a model trained on a diverse set of structures, but still excluding the spleen, achieved a DSC of 0.80, HD of 13 mm, and relative APL of 0.35. Iterative prediction performed better compared to direct prediction when considering unseen structures. CONCLUSIONS: Training a contextual deep learning model on a diverse set of structures increases the segmentation performance for the structures in the training set, but importantly enables the model to generalize and make predictions even for unseen structures that were not represented in the training set. This shows that user-provided context can be incorporated into deep learning contouring to facilitate semi-automatic segmentation of CT images for any given structure. Such an approach can enable faster de-novo contouring in clinical practice.

Three-dimensional ultrasound imaging of mammary ducts in lactating women: a feasibility study.

OBJECTIVE: The main function of the breast is to produce milk for offspring. As such, the ductal system, which carries milk from the milk-secreting glands (alveoli) to the nipple, is central to the natural function of the breast. The ductal system is also the region in which many malignancies originate and spread. In this study, we aimed to assess the feasibility of manual mapping of ductal systems from 3-dimensional (3D) ultrasound data and to evaluate the structures found with respect to conventional understanding of breast anatomy and physiology. METHODS: Three-dimensional ultrasound data of the breast were acquired using a mechanical system, which captures data in a conical shape covering most of the breast without excessive compression. Manual mapping of the ductal system was performed using custom software for data from 4 lactating volunteers. RESULTS: Observational results are presented for ultrasound data from the 4 lactating volunteers. For all volunteers, only a small number of ductal structures were engorged with milk, suggesting that the lactiferous activity of the breast may be localized. These enlarged ducts were predominantly found in the inferior lateral quadrant of each breast. The observation was also made that the enlarged, milk-storing parts of the duct were spread throughout the ductal system and not directly below the nipple as conventional anatomy suggests. CONCLUSIONS: Ultrasound visualization of the 3D structure of milk-laden ducts in an uncompressed breast has been shown. Using manual tracing, it was possible to track milk-laden ducts of diameters less than 1 mm.

External validation of deep learning-based contouring of head and neck organs at risk.

BACKGROUND AND PURPOSE: Head and neck (HN) radiotherapy can benefit from automatic delineation of tumor and surrounding organs because of the complex anatomy and the regular need for adaptation. The aim of this study was to assess the performance of a commercially available deep learning contouring (DLC) model on an external validation set. MATERIALS AND METHODS: The CT-based DLC model, trained at the University Medical Center Groningen (UMCG), was applied to an independent set of 58 patients from the Radboud University Medical Center (RUMC). DLC results were compared to the RUMC manual reference using the Dice similarity coefficient (DSC) and 95th percentile of Hausdorff distance (HD95). Craniocaudal spatial information was added by calculating binned measures. In addition, a qualitative evaluation compared the acceptance of manual and DLC contours in both groups of observers. RESULTS: Good correspondence was shown for the mandible (DSC 0.90; HD95 3.6 mm). Performance was reasonable for the glandular OARs, brainstem and oral cavity (DSC 0.78-0.85, HD95 3.7-7.3 mm). The other aerodigestive tract OARs showed only moderate agreement (DSC 0.53-0.65, HD95 around 9 mm). The binned measures displayed the largest deviations caudally and/or cranially. CONCLUSIONS: This study demonstrates that the DLC model can provide a reasonable starting point for delineation when applied to an independent patient cohort. The qualitative evaluation did not reveal large differences in the interpretation of contouring guidelines between RUMC and UMCG observers.

Assessment of manual adjustment performed in clinical practice following deep learning contouring for head and neck organs at risk in radiotherapy.

BACKGROUND AND PURPOSE: Auto-contouring performance has been widely studied in development and commissioning studies in radiotherapy, and its impact on clinical workflow assessed in that context. This study aimed to evaluate the manual adjustment of auto-contouring in routine clinical practice and to identify improvements regarding the auto-contouring model and clinical user interaction, to improve the efficiency of auto-contouring. MATERIALS AND METHODS: A total of 103 clinical head and neck cancer cases, contoured using a commercial deep-learning contouring system and subsequently checked and edited for clinical use were retrospectively taken from clinical data over a twelve-month period (April 2019-April 2020). The amount of adjustment performed was calculated, and all cases were registered to a common reference frame for assessment purposes. The median, 10th and 90th percentile of adjustment were calculated and displayed using 3D renderings of structures to visually assess systematic and random adjustment. Results were also compared to inter-observer variation reported previously. Assessment was performed for both the whole structures and for regional sub-structures, and according to the radiation therapy technologist (RTT) who edited the contour. RESULTS: The median amount of adjustment was low for all structures (<2 mm), although large local adjustment was observed for some structures. The median was systematically greater or equal to zero, indicating that the auto-contouring tends to under-segment the desired contour. CONCLUSION: Auto-contouring performance assessment in routine clinical practice has identified systematic improvements required technically, but also highlighted the need for continued RTT training to ensure adherence to guidelines.

Improving automatic delineation for head and neck organs at risk by Deep Learning Contouring.

INTRODUCTION: Adequate head and neck (HN) organ-at-risk (OAR) delineation is crucial for HN radiotherapy and for investigating the relationships between radiation dose to OARs and radiation-induced side effects. The automatic contouring algorithms that are currently in clinical use, such as atlas-based contouring (ABAS), leave room for improvement. The aim of this study was to use a comprehensive evaluation methodology to investigate the performance of HN OAR auto-contouring when using deep learning contouring (DLC), compared to ABAS. METHODS: The DLC neural network was trained on 589 HN cancer patients. DLC was compared to ABAS by providing each method with an independent validation cohort of 104 patients, which had also been manually contoured. For each of the 22 OAR contours - glandular, upper digestive tract and central nervous system (CNS)-related structures - the dice similarity coefficient (DICE), and absolute mean and max dose differences (|Δmean-dose| and |Δmax-dose|) performance measures were obtained. For a subset of 7 OARs, an evaluation of contouring time, inter-observer variation and subjective judgement was performed. RESULTS: DLC resulted in equal or significantly improved quantitative performance measures in 19 out of 22 OARs, compared to the ABAS (DICE/|Δmean dose|/|Δmax dose|: 0.59/4.2/4.1 Gy (ABAS); 0.74/1.1/0.8 Gy (DLC)). The improvements were mainly for the glandular and upper digestive tract OARs. DLC significantly reduced the delineation time for the inexperienced observer. The subjective evaluation showed that DLC contours were more often preferable to the ABAS contours overall, were considered to be more precise, and more often confused with manual contours. Manual contours still outperformed both DLC and ABAS; however, DLC results were within or bordering the inter-observer variability for the manual edited contours in this cohort. CONCLUSION: The DLC, trained on a large HN cancer patient cohort, outperformed the ABAS for the majority of HN OARs.

Applications and limitations of machine learning in radiation oncology.

Machine learning approaches to problem-solving are growing rapidly within healthcare, and radiation oncology is no exception. With the burgeoning interest in machine learning comes the significant risk of misaligned expectations as to what it can and cannot accomplish. This paper evaluates the role of machine learning and the problems it solves within the context of current clinical challenges in radiation oncology. The role of learning algorithms within the workflow for external beam radiation therapy are surveyed, considering simulation imaging, multimodal fusion, image segmentation, treatment planning, quality assurance, and treatment delivery and adaptation. For each aspect, the clinical challenges faced, the learning algorithms proposed, and the successes and limitations of various approaches are analyzed. It is observed that machine learning has largely thrived on reproducibly mimicking conventional human-driven solutions with more efficiency and consistency. On the other hand, since algorithms are generally trained using expert opinion as ground truth, machine learning is of limited utility where problems or ground truths are not well-defined, or if suitable measures of correctness are not available. As a result, machines may excel at replicating, automating and standardizing human behaviour on manual chores, meanwhile the conceptual clinical challenges relating to definition, evaluation, and judgement remain in the realm of human intelligence and insight.

An Evaluation of Atlas Selection Methods for Atlas-Based Automatic Segmentation in Radiotherapy Treatment Planning.

Atlas-based automatic segmentation is used in radiotherapy planning to accelerate the delineation of organs at risk (OARs). Atlas selection has been proposed as a way to improve the accuracy and execution time of segmentation, assuming that, the more similar the atlas is to the patient, the better the results will be. This paper presents an analysis of atlas selection methods in the context of radiotherapy treatment planning. For a range of commonly contoured OARs, a thorough comparison of a large class of typical atlas selection methods has been performed. For this evaluation, clinically contoured CT images of the head and neck ( N=316 ) and thorax ( N=280 ) were used. The state-of-the-art intensity and deformation similarity-based atlas selection methods were found to compare poorly to perfect atlas selection. Counter-intuitively, atlas selection methods based on a fixed set of representative atlases outperformed atlas selection methods based on the patient image. This study suggests that atlas-based segmentation with currently available selection methods compares poorly to the potential best performance, hampering the clinical utility of atlas-based segmentation. Effective atlas selection remains an open challenge in atlas-based segmentation for radiotherapy planning.

Volume segmentation and reconstruction from freehand three-dimensional ultrasound data with application to ovarian follicle measurement.

This article presents a semi-automatic method for segmentation and reconstruction of freehand three-dimensional (3D) ultrasound data. The method incorporates a number of interesting features within the level-set framework: First, segmentation is carried out using region competition, requiring multiple distinct and competing regions to be encoded within the framework. This region competition uses a simple dot-product based similarity measure to compare intensities within each region. In addition, segmentation and surface reconstruction is performed within the 3D domain to take advantage of the additional spatial information available. This means that the method must interpolate the surface where there are gaps in the data, a feature common to freehand 3D ultrasound reconstruction. Finally, although the level-set method is restricted to a voxel grid, no assumption is made that the data being segmented will conform to this grid and may be segmented in its world-reference position. The volume reconstruction method is demonstrated in vivo for the volume measurement of ovarian follicles. The 3D reconstructions produce a lower error variance than the current clinical measurement based on a mean diameter estimated from two-dimensional (2D) images. However, both the clinical measurement and the semi-automatic method appear to underestimate the true follicular volume.

Comparative evaluation of autocontouring in clinical practice: A practical method using the Turing test.

PURPOSE: Automated techniques for estimating the contours of organs and structures in medical images have become more widespread and a variety of measures are available for assessing their quality. Quantitative measures of geometric agreement, for example, overlap with a gold-standard delineation, are popular but may not predict the level of clinical acceptance for the contouring method. Therefore, surrogate measures that relate more directly to the clinical judgment of contours, and to the way they are used in routine workflows, need to be developed. The purpose of this study is to propose a method (inspired by the Turing Test) for providing contour quality measures that directly draw upon practitioners' assessments of manual and automatic contours. This approach assumes that an inability to distinguish automatically produced contours from those of clinical experts would indicate that the contours are of sufficient quality for clinical use. In turn, it is anticipated that such contours would receive less manual editing prior to being accepted for clinical use. In this study, an initial assessment of this approach is performed with radiation oncologists and therapists. METHODS: Eight clinical observers were presented with thoracic organ-at-risk contours through a web interface and were asked to determine if they were automatically generated or manually delineated. The accuracy of the visual determination was assessed, and the proportion of contours for which the source was misclassified recorded. Contours of six different organs in a clinical workflow were for 20 patient cases. The time required to edit autocontours to a clinically acceptable standard was also measured, as a gold standard of clinical utility. Established quantitative measures of autocontouring performance, such as Dice similarity coefficient with respect to the original clinical contour and the misclassification rate accessed with the proposed framework, were evaluated as surrogates of the editing time measured. RESULTS: The misclassification rates for each organ were: esophagus 30.0%, heart 22.9%, left lung 51.2%, right lung 58.5%, mediastinum envelope 43.9%, and spinal cord 46.8%. The time savings resulting from editing the autocontours compared to the standard clinical workflow were 12%, 25%, 43%, 77%, 46%, and 50%, respectively, for these organs. The median Dice similarity coefficients between the clinical contours and the autocontours were 0.46, 0.90, 0.98, 0.98, 0.94, and 0.86, respectively, for these organs. CONCLUSIONS: A better correspondence with time saving was observed for the misclassification rate than the quantitative contour measures explored. From this, we conclude that the inability to accurately judge the source of a contour indicates a reduced need for editing and therefore a greater time saving overall. Hence, task-based assessments of contouring performance may be considered as an additional way of evaluating the clinical utility of autosegmentation methods.