In Quantitative imaging in medicine and surgery

BACKGROUND : The cerebellum and cerebral cortex form the most important cortico-cerebellar system in the brain. However, diffusion magnetic resonance imaging (MRI)-based tractography of the connecting white matter between the cerebellum and cerebral cortex, which support language function, has not been extensively reported on. This work aims to serve as a guideline for facilitating the analysis of white matter tracts along the language-related cerebro-cerebellar pathway (LRCCP), which includes the corticopontine, pontocerebellar, corticorubral, rubroolivary, olivocerebellar, and dentatorubrothalamic tracts.

METHODS : The LRCCP templates were developed via processing the high-resolution, population-averaged atlas available in the Human Connectome Project (HCP)-1065 dataset (2017 Q4, 1,200-subject release) in DSI Studio. The deterministic tracking was performed with the manually selected regions of interest (ROIs) on this atlas according to prior anatomic knowledge. Templates were then applied to the MRI datasets of 30 health participants acquired from a single hospital to verify the practicability of the tracking. The diffusion tensor and shape analysis metrics were calculated for all LRCCP tracts. Differences in the tracking metrics between the left and right hemispheres were compared, and the related white matter asymmetry was discussed.

RESULTS : The LRCCP templates were successfully created and applied to healthy participants for quantitative analysis. Significantly higher mean fractional anisotropy (FA) values were discovered on the left (L) corticorubral tract [L, 0.43±0.02 vs. right (R), 0.41±0.02; P<0.01] and left dentatorubrothalamic tract (L, 0.47±0.02 vs. R, 0.46±0.02; P<0.01). Significant differences in tract volume and streamline number were observed between the corticopontine, corticorubral, and dentatorubrothalamic tracts. The size of the right corticopontine and corticorubral tracts were smaller, and both had smaller streamline numbers and innervation areas when compared with the contralateral sides. The R dentatorubrothalamic tract showed a larger volume (R, 23,582.47±4,160.71 mm³ vs. L, 19,821.27±2,983.91 mm³; P<0.01) and innervation area (R, 2,117.37±433.98 mm² vs. L, 1,610.00±356.19 mm²; P<0.01) than did the L side. No significant differences were observed in the rubroolivary tracts.

CONCLUSIONS : This work suggests the feasibility of applying tractography templates of the LRCCP to quantitatively evaluate white matter properties associated with language function. Lateralized diffusion metrics were observed in preliminary experiments. LRCCP tractography-based research may provide a potential quantitative method to better understanding neuroplasticity.

Yin Hu, Zong Fangrong, Deng Xiaofeng, Zhang Dong, Zhang Yan, Wang Shuo, Wang Yu, Zhao Jizong

2023-Mar-01

Diffusion magnetic resonance imaging, cerebellum, language, shape analysis, tractography, white matter asymmetry

In Quantitative imaging in medicine and surgery

BACKGROUND : Quantitative muscle and fat data obtained through body composition analysis are expected to be a new stable biomarker for the early and accurate prediction of treatment-related toxicity, treatment response, and prognosis in patients with lung cancer. The use of these biomarkers can enable the adjustment of individualized treatment regimens in a timely manner, which is critical to further improving patient prognosis and quality of life. We aimed to develop a deep learning model based on attention for fully automated segmentation of the abdomen from computed tomography (CT) to quantify body composition.

METHODS : A fully automatic segmentation deep learning model was designed based on the attention mechanism and using U-Net as the framework. Subcutaneous fat, skeletal muscle, and visceral fat were manually segmented by two experts to serve as ground truth labels. The performance of the model was evaluated using Dice similarity coefficients (DSCs) and Hausdorff distance at 95th percentile (HD95).

RESULTS : The mean DSC for subcutaneous fat and skeletal muscle were high for both the enhanced CT test set (0.93±0.06 and 0.96±0.02, respectively) and the plain CT test set (0.90±0.09 and 0.95±0.01, respectively). Nevertheless, the model did not perform well in the segmentation performance of visceral fat, especially for the enhanced CT test set. The mean DSC for the enhanced CT test set was 0.87±0.11, while the mean DSC for the plain CT test set was 0.92±0.03. We discuss the reasons for this result.

CONCLUSIONS : This work demonstrates a method for the automatic outlining of subcutaneous fat, skeletal muscle, and visceral fat areas at L3.

Shen Hao, He Pin, Ren Ya, Huang Zhengyong, Li Shuluan, Wang Guoshuai, Cong Minghua, Luo Dehong, Shao Dan, Lee Elaine Yuen-Phin, Cui Ruixue, Huo Li, Qin Jing, Liu Jun, Hu Zhanli, Liu Zhou, Zhang Na

2023-Mar-01

CT scan, deep learning, fat segmentation, muscle segmentation, sarcopenia

In Quantitative imaging in medicine and surgery

BACKGROUND : Image segmentation is an important step during the processing of medical images. For example, for the computer aid diagnostic systems for lung cancer image analysis, the segmented regions of tumors would help doctors in early diagnosis to determine timely and appropriate treatment possibilities and thereby improve the survival rate of the patients. However, general clinical routines of manual segmentation for large number of medical images are very difficult and time consuming, which is the challenge we aim to tackle using our proposed method.

METHODS : A novel image segmentation method with evolutionary learning technique named Group Theoretic Particle Swarm Optimization is proposed. It can tackle multi-level thresholding optimization problem during the segmentation process and rebuild the search paradigm according to the solid mathematical foundation of symmetric group from four designable aspects, which are particle encoding, solution landscape, neighborhood movement and swarm topology, respectively. The Kapur's entropy of multi-level thresholds is assessed as the objective function.

RESULTS : In contrast to those conventional metaheuristics methods for lung cancer image segmentation, this newly presented method generates the best performance result among them. Experimental results show that its Kapur's entropy has the value of 9.07, which is 16% higher than the worst case. Computational time is acceptable at the cost of 173.730 seconds, average level of evaluation metrics [Kappa, Precision, Recall, F1-measure, intersection over union (IoU) and receiver operating characteristic (ROC)] is over 90%, and search process of multi-level threshold combination would finally converge in the later phase of iterations after 700. The ablation study indicates that all components are significant to the contributions of our proposed method.

CONCLUSIONS : Group Theoretic Particle Swarm Optimization for multi-level threshold segmentation is an efficient way to split a medical image into distinct regions and extract tumor tissues regions from the background. It maintains the balanced relationship between diversification and intensification during the search process and helps clinicians to make the diagnosis more accurately. Our proposed method processes potential medical value and clinical meanings.

Lan Kun, Zhou Jianqiang, Jiang Xiaoliang, Wang Jun, Huang Shigao, Yang Jie, Song Qun, Tang Rui, Gong Xueyuan, Liu Kexing, Wu Yaoyang, Li Tengyue

2023-Mar-01

Lung cancer detection, evolutionary computation, group theory, medical image segmentation, metaheuristic

In Quantitative imaging in medicine and surgery

BACKGROUND : It is clinically important to accurately assess the pulmonary function of patients with lung cancer, especially before surgery. This knowledge can help clinicians to monitor patients pre- and post-surgery, predict the impact of surgery on pulmonary function, and help to optimize postsurgical recovery. We used a deep learning approach for assessing pulmonary function on computed tomography (CT) scans in patients with lung cancer before they underwent surgery.

METHODS : A total of 188 patients with lung cancer whose diagnoses had been pathologically confirmed were enrolled in this study. We used a software to automatically delineate regions of interest (ROIs) throughout the airways, lobes, and the whole lungs. We then used AK software to extract radiomics features of the 3 types of ROIs. We randomly separated these cases into a training cohort and a test cohort at a ratio of 7:3. We next constructed a logistic regression model to assess pulmonary function from the radiomics features. The machine learning outcomes were compared with established clinical criteria for pulmonary function. including forced expiratory volume in the first second/forced vital capacity (FEV1/FVC), FVC, and maximum vital capacity (VCmax) to evaluate the accuracy of the machine learning model.

RESULTS : In the ROIs of the lobes, our results showed that the machine learning model had good performance in predicting FVC and VCmax, attaining a Spearman correlation r value of 0.714 with P<0.001 for FVC and a r value of 0.687 with P<0.001 for VCmax. Using the airway ROIs, our model achieved a r of 0.603 with P=0.001 for VCmax. Using the whole lung ROIs, our model achieved a r of 0.704 with P<0.001 for FVC and a r of 0.693 with P<0.001 for VCmax.

CONCLUSIONS : Preoperative CT may provide a means for evaluating pulmonary function in patients with lung cancer. With radiomics features extracted from the airway, lobes, and the whole lung region, and a properly trained machine learning model, it is possible to obtain accurate estimation for metrics used in clinical criteria and to offer clinicians imaging-based indicators for the status of pulmonary functions.

Meng Hongjia, Liu Yun, Xu Xiaoyin, Liao Yuting, Liang Hengrui, Chen Huai

2023-Mar-01

Machine learning, assessment, computed tomography (CT), lung cancer, pulmonary function

In Quantitative imaging in medicine and surgery

BACKGROUND : Early identification of severe acute pancreatitis (SAP) is key to reducing mortality and improving prognosis. We aimed to establish a radiomics model and nomogram for early prediction of acute pancreatitis (AP) severity based on contrast-enhanced computed tomography (CT) images.

METHODS : We retrospectively analyzed 215 patients with first-episode AP, including 141 in the training cohort (87 men and 54 women, mean age 51.37±16.09 years) and 74 in the test cohort (40 men and 34 women, mean age 55.49±17.83 years). Radiomics features were extracted from portal venous phase images based on pancreatic and peripancreatic regions. The light gradient boosting machine (LightGBM) algorithm was used for feature selection, a logistic regression (LR) model was established and trained by 10-fold cross-validation, and a nomogram was established based on the best features. The model's predictive performance was evaluated according to the area under the curve (AUC) of the receiver operating characteristic (ROC) curve, sensitivity, specificity, and accuracy.

RESULTS : A total of 13 optimal radiomics features were selected by LightGBM for LR model building. The AUC of the radiomics (LR) model was 0.992 [95% confidence interval (CI): 0.963-0.996] in the training cohort, 0.965 (95% CI: 0.924-0.981) in the validation cohort, and 0.894 (95% CI: 0.789-0.966) in the test cohort. The sensitivity was 0.862 (95% CI: 0.674-0.954), the specificity was 0.800 (95% CI: 0.649-0.899), and the accuracy was 0.824 (95% CI: 0.720-0.919). The nomogram based on the 13 radiomics features showed that SAP would be predicted when the total score was greater than 124.

CONCLUSIONS : The radiomics model based on enhanced-CT images of pancreatic and peripancreatic regions performed well in the early prediction of AP severity. The nomogram based on selected radiomics features could provide a reference for AP clinical assessment.

Zhao Yanmei, Wei Jiayi, Xiao Bo, Wang Liu, Jiang Xian, Zhu Yuanzhong, He Wenjing

2023-Mar-01

Radiomics, acute pancreatitis (AP), enhanced computed tomography (enhanced CT), machine learning, nomogram

In Quantitative imaging in medicine and surgery

BACKGROUND : The aim of this study was to compare the dose reduction potential and image quality of deep learning-based image reconstruction (DLIR) with those of filtered back-projection (FBP) and iterative reconstruction (IR) and to determine the clinically usable dose of DLIR for low-dose chest computed tomography (LDCT) scans.

METHODS : Multi-slice computed tomography (CT) scans of a chest phantom were performed with various tube voltages and tube currents, and the images were reconstructed using seven methods to control the amount of noise reduction: FBP, three stages of IR, and three stages of DLIR. For subjective image analysis, four radiologists compared 48 image data sets with reference images and rated on a 5-point scale. For quantitative image analysis, the signal to noise ratio (SNR), contrast to noise ratio (CNR), nodule volume, and nodule diameter were measured.

RESULTS : In the subjective analysis, DLIR-Low (0.46 mGy), DLIR-Medium (0.31 mGy), and DLIR-High (0.18 mGy) images showed similar quality to the FBP (2.47 mGy) image. Under the same dose conditions, the SNR and CNR were higher with DLIR-High than with FBP and all the IR methods (all P<0.05). The nodule volume and size with DLIR-High were significantly closer to the real volume than with FBP and all the IR methods (all P<0.001).

CONCLUSIONS : DLIR can improve the image quality of LDCT compared to FBP and IR. In addition, the appropriate effective dose for LDCT would be 0.24 mGy with DLIR-High.

Jung Yunsub, Hur Jin, Han Kyunghwa, Imai Yasuhiro, Hong Yoo Jin, Im Dong Jin, Lee Kye Ho, Desnoyers Melissa, Thomsen Brian, Shigemasa Risa, Um Kyounga, Jang Kyungeun

2023-Mar-01

Low-dose chest computed tomography (LDCT), chest phantom, deep learning-based image reconstruction (DLIR)