scholarly journals A Comparative Study of Radiomics and Deep-Learning Based Methods for Pulmonary Nodule Malignancy Prediction in Low Dose CT Images

2021 ◽  
Vol 11 ◽  
Author(s):  
Mehdi Astaraki ◽  
Guang Yang ◽  
Yousuf Zakko ◽  
Iuliana Toma-Dasu ◽  
Örjan Smedby ◽  
...  
Keyword(s):  
Feature Selection ◽  
Deep Learning ◽  
Comparative Study ◽  
Pulmonary Nodule ◽  
Lung Nodule ◽  
Fine Tuning ◽  
Deep Feature ◽  
Feature Based ◽  
End To End ◽  
Deep Learning Model

ObjectivesBoth radiomics and deep learning methods have shown great promise in predicting lesion malignancy in various image-based oncology studies. However, it is still unclear which method to choose for a specific clinical problem given the access to the same amount of training data. In this study, we try to compare the performance of a series of carefully selected conventional radiomics methods, end-to-end deep learning models, and deep-feature based radiomics pipelines for pulmonary nodule malignancy prediction on an open database that consists of 1297 manually delineated lung nodules.MethodsConventional radiomics analysis was conducted by extracting standard handcrafted features from target nodule images. Several end-to-end deep classifier networks, including VGG, ResNet, DenseNet, and EfficientNet were employed to identify lung nodule malignancy as well. In addition to the baseline implementations, we also investigated the importance of feature selection and class balancing, as well as separating the features learned in the nodule target region and the background/context region. By pooling the radiomics and deep features together in a hybrid feature set, we investigated the compatibility of these two sets with respect to malignancy prediction.ResultsThe best baseline conventional radiomics model, deep learning model, and deep-feature based radiomics model achieved AUROC values (mean ± standard deviations) of 0.792 ± 0.025, 0.801 ± 0.018, and 0.817 ± 0.032, respectively through 5-fold cross-validation analyses. However, after trying out several optimization techniques, such as feature selection and data balancing, as well as adding context features, the corresponding best radiomics, end-to-end deep learning, and deep-feature based models achieved AUROC values of 0.921 ± 0.010, 0.824 ± 0.021, and 0.936 ± 0.011, respectively. We achieved the best prediction accuracy from the hybrid feature set (AUROC: 0.938 ± 0.010).ConclusionThe end-to-end deep-learning model outperforms conventional radiomics out of the box without much fine-tuning. On the other hand, fine-tuning the models lead to significant improvements in the prediction performance where the conventional and deep-feature based radiomics models achieved comparable results. The hybrid radiomics method seems to be the most promising model for lung nodule malignancy prediction in this comparative study.

The Efforts of Deep Learning Approaches for Breast Cancer Detection Based on X-Ray Images

2021 ◽  
pp. 290-309
Author(s):  
Aras Masood Ismael ◽  
Juliana Carneiro Gomes
Keyword(s):  
Breast Cancer ◽  
Feature Extraction ◽  
Deep Learning ◽  
Fine Tuning ◽  
Svm Classifier ◽  
X Ray ◽  
Deep Feature ◽  
Deep Cnn ◽  
End To End ◽  
Deep Feature Extraction

In this chapter, deep learning-based approaches, namely deep feature extraction, fine-tuning of pre-trained convolutional neural networks (CNN), and end-to-end training of a developed CNN model, are used to classify the malignant and normal breast X-ray images. For deep feature extraction, pre-trained deep CNN models such as ResNet18, ResNet50, ResNet101, VGG16, and VGG19 are used. For classification of the deep features, the support vector machines (SVM) classifier is used with various kernel functions namely linear, quadratic, cubic, and Gaussian, respectively. The aforementioned pre-trained deep CNN models are also used in fine-tuning procedure. A new CNN model is also proposed in end-to-end training fashion. The classification accuracy is used as performance measurements. The experimental works show that the deep learning has potential in detection of the breast cancer from the X-ray images. The deep features that are extracted from the ResNet50 model and SVM classifier with linear kernel function produced 94.7% accuracy score which the highest among all obtained.

scholarly journals Interpretable deep learning for the remote characterisation of ambulation in multiple sclerosis using smartphones

2021 ◽  
Vol 11 (1) ◽  
Author(s):  
Andrew P. Creagh ◽  
Florian Lipsmeier ◽  
Michael Lindemann ◽  
Maarten De Vos
Keyword(s):  
Multiple Sclerosis ◽  
Deep Learning ◽  
Inertial Sensor ◽  
Heterogeneous Data ◽  
Fine Tuning ◽  
Sensor Data ◽  
Support Vector ◽  
Deep Convolutional Neural Networks ◽  
Healthcare Applications ◽  
Feature Based

AbstractThe emergence of digital technologies such as smartphones in healthcare applications have demonstrated the possibility of developing rich, continuous, and objective measures of multiple sclerosis (MS) disability that can be administered remotely and out-of-clinic. Deep Convolutional Neural Networks (DCNN) may capture a richer representation of healthy and MS-related ambulatory characteristics from the raw smartphone-based inertial sensor data than standard feature-based methodologies. To overcome the typical limitations associated with remotely generated health data, such as low subject numbers, sparsity, and heterogeneous data, a transfer learning (TL) model from similar large open-source datasets was proposed. Our TL framework leveraged the ambulatory information learned on human activity recognition (HAR) tasks collected from wearable smartphone sensor data. It was demonstrated that fine-tuning TL DCNN HAR models towards MS disease recognition tasks outperformed previous Support Vector Machine (SVM) feature-based methods, as well as DCNN models trained end-to-end, by upwards of 8–15%. A lack of transparency of “black-box” deep networks remains one of the largest stumbling blocks to the wider acceptance of deep learning for clinical applications. Ensuing work therefore aimed to visualise DCNN decisions attributed by relevance heatmaps using Layer-Wise Relevance Propagation (LRP). Through the LRP framework, the patterns captured from smartphone-based inertial sensor data that were reflective of those who are healthy versus people with MS (PwMS) could begin to be established and understood. Interpretations suggested that cadence-based measures, gait speed, and ambulation-related signal perturbations were distinct characteristics that distinguished MS disability from healthy participants. Robust and interpretable outcomes, generated from high-frequency out-of-clinic assessments, could greatly augment the current in-clinic assessment picture for PwMS, to inform better disease management techniques, and enable the development of better therapeutic interventions.

scholarly journals End-to-End, Pixel-Wise Vessel-Specific Coronary and Aortic Calcium Detection and Scoring Using Deep Learning

Diagnostics ◽  
2021 ◽  
Vol 11 (2) ◽  
pp. 215
Author(s):  
Gurpreet Singh ◽  
Subhi Al’Aref ◽  
Benjamin Lee ◽  
Jing Lee ◽  
Swee Tan ◽  
...  
Keyword(s):  
Deep Learning ◽  
Coronary Artery ◽  
Clinical Importance ◽  
Learning Model ◽  
Mean Difference ◽  
Interquartile Range ◽  
History Of ◽  
End To End ◽  
Artery Disease ◽  
Deep Learning Model

Conventional scoring and identification methods for coronary artery calcium (CAC) and aortic calcium (AC) result in information loss from the original image and can be time-consuming. In this study, we sought to demonstrate an end-to-end deep learning model as an alternative to the conventional methods. Scans of 377 patients with no history of coronary artery disease (CAD) were obtained and annotated. A deep learning model was trained, tested and validated in a 60:20:20 split. Within the cohort, mean age was 64.2 ± 9.8 years, and 33% were female. Left anterior descending, right coronary artery, left circumflex, triple vessel, and aortic calcifications were present in 74.87%, 55.82%, 57.41%, 46.03%, and 85.41% of patients respectively. An overall Dice score of 0.952 (interquartile range 0.921, 0.981) was achieved. Stratified by subgroups, there was no difference between male (0.948, interquartile range 0.920, 0.981) and female (0.965, interquartile range 0.933, 0.980) patients (p = 0.350), or, between age <65 (0.950, interquartile range 0.913, 0.981) and age ≥65 (0.957, interquartile range 0.930, 0.9778) (p = 0.742). There was good correlation and agreement for CAC prediction (rho = 0.876, p < 0.001), with a mean difference of 11.2% (p = 0.100). AC correlated well (rho = 0.947, p < 0.001), with a mean difference of 9% (p = 0.070). Automated segmentation took approximately 4 s per patient. Taken together, the deep-end learning model was able to robustly identify vessel-specific CAC and AC with high accuracy, and predict Agatston scores that correlated well with manual annotation, facilitating application into areas of research and clinical importance.

Abstract P319: Can Deep Learning Find the Ischemic Core on CT? Transfer Learning From Pre-Trained MRI-Based Networks

Stroke ◽  
2021 ◽  
Vol 52 (Suppl_1) ◽  
Author(s):  
Yannan Yu ◽  
Soren Christensen ◽  
Yuan Xie ◽  
Enhao Gong ◽  
Maarten G Lansberg ◽  
...  
Keyword(s):  
Deep Learning ◽  
Ground Truth ◽  
Learning Model ◽  
Fine Tuning ◽  
Learning Models ◽  
Starting Point ◽  
Stroke Lesion ◽  
Ischemic Core ◽  
Deep Learning Model

Objective: Ischemic core prediction from CT perfusion (CTP) remains inaccurate compared with gold standard diffusion-weighted imaging (DWI). We evaluated if a deep learning model to predict the DWI lesion from MR perfusion (MRP) could facilitate ischemic core prediction on CTP. Method: Using the multi-center CRISP cohort of acute ischemic stroke patient with CTP before thrombectomy, we included patients with major reperfusion (TICI score≥2b), adequate image quality, and follow-up MRI at 3-7 days. Perfusion parameters including Tmax, mean transient time, cerebral blood flow (CBF), and cerebral blood volume were reconstructed by RAPID software. Core lab experts outlined the stroke lesion on the follow-up MRI. A previously trained MRI model in a separate group of patients was used as a starting point, which used MRP parameters as input and RAPID ischemic core on DWI as ground truth. We fine-tuned this model, using CTP parameters as input, and follow-up MRI as ground truth. Another model was also trained from scratch with only CTP data. 5-fold cross validation was used. Performance of the models was compared with ischemic core (rCBF≤30%) from RAPID software to identify the presence of a large infarct (volume>70 or >100ml). Results: 94 patients in the CRISP trial met the inclusion criteria (mean age 67±15 years, 52% male, median baseline NIHSS 18, median 90-day mRS 2). Without fine-tuning, the MRI model had an agreement of 73% in infarct >70ml, and 69% in >100ml; the MRI model fine-tuned on CT improved the agreement to 77% and 73%; The CT model trained from scratch had agreements of 73% and 71%; All of the deep learning models outperformed the rCBF segmentation from RAPID, which had agreements of 51% and 64%. See Table and figure. Conclusions: It is feasible to apply MRP-based deep learning model to CT. Fine-tuning with CTP data further improves the predictions. All deep learning models predict the stroke lesion after major recanalization better than thresholding approaches based on rCBF.

Sign in / Sign up

Export Citation Format

Share Document