Accuracy, uncertainty, and adaptability of automatic myocardial ASL segmentation using deep CNN

Automatic gross tumor volume segmentation with failure detection for safe implementation in locally advanced cervical cancer

Background and Purpose

Automatic segmentation methods have greatly changed the RadioTherapy (RT) workflow, but still need to be extended to target volumes. In this paper, Deep Learning (DL) models were compared for Gross Tumor Volume (GTV) segmentation in locally advanced cervical cancer, and a novel investigation into failure detection was introduced by utilizing radiomic features.

Methods and materials

We trained eight DL models (UNet, VNet, SegResNet, SegResNetVAE) for 2D and 3D segmentation. Ensembling individually trained models during cross-validation generated the final segmentation. To detect failures, binary classifiers were trained using radiomic features extracted from segmented GTVs as inputs, aiming to classify contours based on whether their Dice Similarity Coefficient ( DSC ) < T and DSC ⩾ T . Two distinct cohorts of T2-Weighted (T2W) pre-RT MR images captured in 2D sequences were used: one retrospective cohort consisting of 115 LACC patients from 30 scanners, and the other prospective cohort, comprising 51 patients from 7 scanners, used for testing.

Results

Segmentation by 2D-SegResNet achieved the best DSC, Surface DSC (SDSC 3 mm ), and 95th Hausdorff Distance (95HD): DSC = 0.72 ± 0.16,SDSC 3 mm =0.66 ± 0.17, and 95HD = 14.6 ± 9.0 mm without missing segmentation ( M =0) on the test cohort. Failure detection could generate precision ( P = 0.88 ), recall ( R = 0.75 ), F1-score ( F = 0.81 ), and accuracy ( A = 0.86 ) using Logistic Regression (LR) classifier on the test cohort with a threshold T = 0.67 on DSC values.

Conclusions

Our study revealed that segmentation accuracy varies slightly among different DL methods, with 2D networks outperforming 3D networks in 2D MRI sequences. Doctors found the time-saving aspect advantageous. The proposed failure detection could guide doctors in sensitive cases.

Update on state-of-the-art for arterial spin labeling (ASL) human perfusion imaging outside of the brain

This review article provides an overview of developments for arterial spin labeling (ASL) perfusion imaging in the body (i.e., outside of the brain). It is part of a series of review/recommendation papers from the International Society for Magnetic Resonance in Medicine (ISMRM) Perfusion Study Group. In this review, we focus on specific challenges and developments tailored for ASL in a variety of body locations. After presenting common challenges, organ-specific reviews of challenges and developments are presented, including kidneys, lungs, heart (myocardium), placenta, eye (retina), liver, pancreas, and muscle, which are regions that have seen the most developments outside of the brain. Summaries and recommendations of acquisition parameters (when appropriate) are provided for each organ. We then explore the possibilities for wider adoption of body ASL based on large standardization efforts, as well as the potential opportunities based on recent advances in high/low-field systems and machine-learning. This review seeks to provide an overview of the current state-of-the-art of ASL for applications in the body, highlighting ongoing challenges and solutions that aim to enable more widespread use of the technique in clinical practice.

Neural Network in the Analysis of the MR Signal as an Image Segmentation Tool for the Determination of T1 and T2 Relaxation Times with Application to Cancer Cell Culture

Artificial intelligence has been entering medical research. Today, manufacturers of diagnostic instruments are including algorithms based on neural networks. Neural networks are quickly entering all branches of medical research and beyond. Analyzing the PubMed database from the last 5 years (2017 to 2021), we see that the number of responses to the query "neural network in medicine" exceeds 10,500 papers. Deep learning algorithms are of particular importance in oncology. This paper presents the use of neural networks to analyze the magnetic resonance imaging (MRI) images used to determine MRI relaxometry of the samples. Relaxometry is becoming an increasingly common tool in diagnostics. The aim of this work was to optimize the processing time of DICOM images by using a neural network implemented in the MATLAB package by The MathWorks with the patternnet function. The application of a neural network helps to eliminate spaces in which there are no objects with characteristics matching the phenomenon of longitudinal or transverse MRI relaxation. The result of this work is the elimination of aerated spaces in MRI images. The whole algorithm was implemented as an application in the MATLAB package.

Deep Learning-Based Segmentation and Uncertainty Assessment for Automated Analysis of Myocardial Perfusion MRI Datasets Using Patch-Level Training and Advanced Data Augmentation

In this work, we develop a patch-level training approach and a task-driven intensity-based augmentation method for deep-learning-based segmentation of motion-corrected perfusion cardiac magnetic resonance imaging (MRI) datasets. Further, the proposed method generates an image-based uncertainty map thanks to a novel spatial sliding-window approach used during patch-level training, hence allowing for uncertainty quantification. Using the quantified uncertainty, we detect the out-of-distribution test data instances so that the end-user can be alerted that the test data is not suitable for the trained network. This feature has the potential to enable a more reliable integration of the proposed deep learning-based framework into clinical practice. We test our approach on external MRI data acquired using a different acquisition protocol to demonstrate the robustness of our performance to variations in pulse-sequence parameters. The presented results further demonstrate that our deep-learning image segmentation approach trained with the proposed data-augmentation technique incorporating spatiotemporal (2D+time) patches is superior to the state-of-the-art 2D approach in terms of generalization performance.

Reduction of motion effects in myocardial arterial spin labeling

PURPOSE

To evaluate the accuracy and reproducibility of myocardial blood flow measurements obtained under different breathing strategies and motion correction techniques with arterial spin labeling.

METHODS

A prospective cardiac arterial spin labeling study was performed in 12 volunteers at 3 Tesla. Perfusion images were acquired twice under breath-hold, synchronized-breathing, and free-breathing. Motion detection based on the temporal intensity variation of a myocardial voxel, as well as image registration based on pairwise and groupwise approaches, were applied and evaluated in synthetic and in vivo data. A region of interest was drawn over the mean perfusion-weighted image for quantification. Original breath-hold datasets, analyzed with individual regions of interest for each perfusion-weighted image, were considered as reference values.

RESULTS

Perfusion measurements in the reference breath-hold datasets were in line with those reported in literature. In original datasets, prior to motion correction, myocardial blood flow quantification was significantly overestimated due to contamination of the myocardial perfusion with the high intensity signal of blood pool. These effects were minimized with motion detection or registration. Synthetic data showed that accuracy of the perfusion measurements was higher with the use of registration, in particular after the pairwise approach, which probed to be more robust to motion.

CONCLUSION

Satisfactory results were obtained for the free-breathing strategy after pairwise registration, with higher accuracy and robustness (in synthetic datasets) and higher intrasession reproducibility together with lower myocardial blood flow variability across subjects (in in vivo datasets). Breath-hold and synchronized-breathing after motion correction provided similar results, but these breathing strategies can be difficult to perform by patients.