An adversarial machine learning framework and biomechanical model-guided approach for computing 3D lung tissue elasticity from end-expiration 3DCT

Intraoperative identification of patient-specific elastic modulus of the meniscus during arthroscopy

BACKGROUND AND OBJECTIVE

Degenerative meniscus tissue has been associated with a lower elastic modulus and can lead to the development of arthrosis. Safe intraoperative measurement of in vivo elastic modulus of the human meniscus could contribute to a better understanding of meniscus health, and for developing surgical simulators where novice surgeons can learn to distinguish healthy from degenerative meniscus tissue. Such measurement can also support intraoperative decision-making by providing a quantitative measure of the meniscus health condition. The objective of this study is to demonstrate a method for intraoperative identification of meniscus elastic modulus during arthroscopic probing using an adaptive observer method.

METHODS

Ex vivo arthroscopic examinations were performed on five cadaveric knees to estimate the elastic modulus of the anterior, mid-body, and posterior regions of lateral and medial menisci. Real-time intraoperative force-displacement data was obtained and utilized for modulus estimation through an adaptive observer method. For the validation of arthroscopic elastic moduli, an inverse parameter identification approach using optimization, based on biomechanical indentation tests and finite element analyses, was employed. Experimental force-displacement data in various anatomical locations were measured through indentation. An iterative optimization algorithm was employed to optimize elastic moduli and Poisson's ratios by comparing experimental force values at maximum displacement with the corresponding force values from linear elastic region-specific finite element models. Finally, the estimated elastic modulus values obtained from ex vivo arthroscopy were compared against optimized values using a paired t-test.

RESULTS

The elastic moduli obtained from ex vivo arthroscopy and optimization showcased subject specificity in material properties. Additionally, the results emphasized anatomical and regional specificity within the menisci. The anterior region of the medial menisci exhibited the highest elastic modulus among the anatomical locations studied (9.97±3.20MPa from arthroscopy and 5.05±1.97MPa from finite element-based inverse parameter identification). The paired t-test results indicated no statistically significant difference between the elastic moduli obtained from arthroscopy and inverse parameter identification, suggesting the feasibility of stiffness estimation using arthroscopic examination.

CONCLUSIONS

This study has demonstrated the feasibility of intraoperative identification of patient-specific elastic modulus for meniscus tissue during arthroscopy.

Fusing Diverse Decision Rules in 3D-Radiomics for Assisting Diagnosis of Lung Adenocarcinoma

This study aimed to develop an interpretable diagnostic model for subtyping of pulmonary adenocarcinoma, including minimally invasive adenocarcinoma (MIA), adenocarcinoma in situ (AIS), and invasive adenocarcinoma (IAC), by integrating 3D-radiomic features and clinical data. Data from multiple hospitals were collected, and 10 key features were selected from 1600 3D radiomic signatures and 11 radiological features. Diverse decision rules were extracted using ensemble learning methods (gradient boosting, random forest, and AdaBoost), fused, ranked, and selected via RuleFit and SHAP to construct a rule-based diagnostic model. The model's performance was evaluated using AUC, precision, accuracy, recall, and F1-score and compared with other models. The rule-based diagnostic model exhibited excellent performance in the training, testing, and validation cohorts, with AUC values of 0.9621, 0.9529, and 0.8953, respectively. This model outperformed counterparts relying solely on selected features and previous research models. Specifically, the AUC values for the previous research models in the three cohorts were 0.851, 0.893, and 0.836. It is noteworthy that individual models employing GBDT, random forest, and AdaBoost demonstrated AUC values of 0.9391, 0.8681, and 0.9449 in the training cohort, 0.9093, 0.8722, and 0.9363 in the testing cohort, and 0.8440, 0.8640, and 0.8750 in the validation cohort, respectively. These results highlight the superiority of the rule-based diagnostic model in the assessment of lung adenocarcinoma subtypes, while also providing insights into the performance of individual models. Integrating diverse decision rules enhanced the accuracy and interpretability of the diagnostic model for lung adenocarcinoma subtypes. This approach bridges the gap between complex predictive models and clinical utility, offering valuable support to healthcare professionals and patients.

Deep learning-based lung image registration: A review

Lung image registration can effectively describe the relative motion of lung tissues, thereby helping to solve series problems in clinical applications. Since the lungs are soft and fairly passive organs, they are influenced by respiration and heartbeat, resulting in discontinuity of lung motion and large deformation of anatomic features. This poses great challenges for accurate registration of lung image and its applications. The recent application of deep learning (DL) methods in the field of medical image registration has brought promising results. However, a versatile registration framework has not yet emerged due to diverse challenges of registration for different regions of interest (ROI). DL-based image registration methods used for other ROI cannot achieve satisfactory results in lungs. In addition, there are few review articles available on DL-based lung image registration. In this review, the development of conventional methods for lung image registration is briefly described and a more comprehensive survey of DL-based methods for lung image registration is illustrated. The DL-based methods are classified according to different supervision types, including fully-supervised, weakly-supervised and unsupervised. The contributions of researchers in addressing various challenges are described, as well as the limitations of these approaches. This review also presents a comprehensive statistical analysis of the cited papers in terms of evaluation metrics and loss functions. In addition, publicly available datasets for lung image registration are also summarized. Finally, the remaining challenges and potential trends in DL-based lung image registration are discussed.

An analysis of the regional heterogeneity in tissue elasticity in lung cancer patients with COPD

Purpose

Recent advancements in obtaining image-based biomarkers from CT images have enabled lung function characterization, which could aid in lung interventional planning. However, the regional heterogeneity in these biomarkers has not been well documented, yet it is critical to several procedures for lung cancer and COPD. The purpose of this paper is to analyze the interlobar and intralobar heterogeneity of tissue elasticity and study their relationship with COPD severity.

Methods

We retrospectively analyzed a set of 23 lung cancer patients for this study, 14 of whom had COPD. For each patient, we employed a 5DCT scanning protocol to obtain end-exhalation and end-inhalation images and semi-automatically segmented the lobes. We calculated tissue elasticity using a biomechanical property estimation model. To obtain a measure of lobar elasticity, we calculated the mean of the voxel-wise elasticity values within each lobe. To analyze interlobar heterogeneity, we defined an index that represented the properties of the least elastic lobe as compared to the rest of the lobes, termed the Elasticity Heterogeneity Index (EHI). An index of 0 indicated total homogeneity, and higher indices indicated higher heterogeneity. Additionally, we measured intralobar heterogeneity by calculating the coefficient of variation of elasticity within each lobe.

Results

The mean EHI was 0.223 ± 0.183. The mean coefficient of variation of the elasticity distributions was 51.1% ± 16.6%. For mild COPD patients, the interlobar heterogeneity was low compared to the other categories. For moderate-to-severe COPD patients, the interlobar and intralobar heterogeneities were highest, showing significant differences from the other groups.

Conclusion

We observed a high level of lung tissue heterogeneity to occur between and within the lobes in all COPD severity cases, especially in moderate-to-severe cases. Heterogeneity results demonstrate the value of a regional, function-guided approach like elasticity for procedures such as surgical decision making and treatment planning.

Elastin in healthy and diseased lung

Elastic fibers are an essential part of the pulmonary extracellular matrix (ECM). Intact elastin is required for normal function and its damage contributes profoundly to the etiology and pathology of lung disease. This highlights the need for novel lung-specific imaging methodology that enables high-resolution 3D visualization of the ECM. We consider elastin's involvement in chronic respiratory disease and examine recent methods for imaging and modeling of the lung in the context of advances in lung tissue engineering for research and clinical application.

Automatic liver tumor localization using deep learning-based liver boundary motion estimation and biomechanical modeling (DL-Bio)

PURPOSE

Recently, two-dimensional-to-three-dimensional (2D-3D) deformable registration has been applied to deform liver tumor contours from prior reference images onto estimated cone-beam computed tomography (CBCT) target images to automate on-board tumor localizations. Biomechanical modeling has also been introduced to fine-tune the intra-liver deformation-vector-fields (DVFs) solved by 2D-3D deformable registration, especially at low-contrast regions, using tissue elasticity information and liver boundary DVFs. However, the caudal liver boundary shows low contrast from surrounding tissues in the cone-beam projections, which degrades the accuracy of the intensity-based 2D-3D deformable registration there and results in less accurate boundary conditions for biomechanical modeling. We developed a deep-learning (DL)-based method to optimize the liver boundary DVFs after 2D-3D deformable registration to further improve the accuracy of subsequent biomechanical modeling and liver tumor localization.

METHODS

The DL-based network was built based on the U-Net architecture. The network was trained in a supervised fashion to learn motion correlation between cranial and caudal liver boundaries to optimize the liver boundary DVFs. Inputs of the network had three channels, and each channel featured the 3D DVFs estimated by the 2D-3D deformable registration along one Cartesian direction (x, y, z). To incorporate patient-specific liver boundary information into the DVFs, the DVFs were masked by a liver boundary ring structure generated from the liver contour of the prior reference image. The network outputs were the optimized DVFs along the liver boundary with higher accuracy. From these optimized DVFs, boundary conditions were extracted for biomechanical modeling to further optimize the solution of intra-liver tumor motion. We evaluated the method using 34 liver cancer patient cases, with 24 for training and 10 for testing. We evaluated and compared the performance of three methods: 2D-3D deformable registration, 2D-3D-Bio (2D-3D deformable registration with biomechanical modeling), and DL-Bio (DL model prediction with biomechanical modeling). The tumor localization errors were quantified through calculating the center-of-mass-errors (COMEs), DICE coefficients, and Hausdorff distance between deformed liver tumor contours and manually segmented "gold-standard" contours.

RESULTS

The predicted DVFs by the DL model showed improved accuracy at the liver boundary, which translated into more accurate liver tumor localizations through biomechanical modeling. On a total of 90 evaluated images and tumor contours, the average (± sd) liver tumor COMEs of the 2D-3D, 2D-3D-Bio, and DL-Bio techniques were 4.7 ± 1.9 mm, 2.9 ± 1.0 mm, and 1.7 ± 0.4 mm. The corresponding average (± sd) DICE coefficients were 0.60 ± 0.12, 0.71 ± 0.07, and 0.78 ± 0.03; and the average (± sd) Hausdorff distances were 7.0 ± 2.6 mm, 5.4 ± 1.5 mm, and 4.5 ± 1.3 mm, respectively.

CONCLUSION

DL-Bio solves a general correlation model to improve the accuracy of the DVFs at the liver boundary. With improved boundary conditions, the accuracy of biomechanical modeling can be further increased for accurate intra-liver low-contrast tumor localization.

Scalable quorum-based deep neural networks with adversarial learning for automated lung lobe segmentation in fast helical free-breathing CTs

PURPOSE

Fast helical free-breathing CT (FHFBCT) scans are widely used for 5DCT and 5D Cone Beam imaging protocols. For quantitative analysis of lung physiology and function, it is important to segment the lung lobes in these scans. Since the 5DCT protocols use up to 25 FHFBCT scans, it is important that this segmentation task be automated. In this paper, we present a deep neural network (DNN) framework for segmenting the lung lobes in near real time.

METHODS

A total of 22 patient datasets (550 3D CT scans) were used for the study. Each of the lung lobes was manually segmented and considered ground-truth. A supervised and constrained generative adversarial network (CGAN) was employed for learning each set of lobe segmentations for each patient with 12 patients designated for training data. The resulting generator DNNs represented the lobe segmentations for each training dataset. A quorum-based algorithm was then implemented to test validation data consisting of 10 separate patient datasets (250 3D CTs). Each of the DNNs predicted their corresponding lobes for the validation data, and equal weights were given to the 12 generator CGANs. The quorum process worked by selecting the weighted average result of all 12 CGAN results for each lobe.

RESULTS

When evaluated against ground-truth segmentations, the quorum-based lobe segmentation was observed to have average structural similarity index, normalized cross-correlation coefficient, and dice coefficient values of 0.929, 0.806, and 0.814, respectively, compared to values of 0.911, 0.698, and 0.696, respectively, using a conventional strategy.

CONCLUSION

The proposed quorum-based approach computed segmentations with clinically acceptable accuracy in near real time using a multi-GPU-based computing setup. This method is scalable as more patient-specific CGANs can be added to the quorum over time.