Basic of machine learning and deep learning in imaging for medical physicists

Development of AI-Based Diagnostic Algorithm for Nasal Bone Fracture Using Deep Learning

Facial bone fractures are relatively common, with the nasal bone the most frequently fractured facial bone. Computed tomography is the gold standard for diagnosing such fractures. Most nasal bone fractures can be treated using a closed reduction. However, delayed diagnosis may cause nasal deformity or other complications that are difficult and expensive to treat. In this study, the authors developed an algorithm for diagnosing nasal fractures by learning computed tomography images of facial bones with artificial intelligence through deep learning. A significant concordance with human doctors' reading results of 100% sensitivity and 77% specificity was achieved. Herein, the authors report the results of a pilot study on the first stage of developing an algorithm for analyzing fractures in the facial bone.

Positron Emission Tomography-Derived Radiomics and Artificial Intelligence in Multiple Myeloma: State-of-the-Art

Multiple myeloma (MM) is a heterogeneous neoplasm accounting for the second most prevalent hematologic disorder. The identification of noninvasive, valuable biomarkers is of utmost importance for the best patient treatment selection, especially in heterogeneous diseases like MM. Despite molecular imaging with positron emission tomography (PET) has achieved a primary role in the characterization of MM, it is not free from shortcomings. In recent years, radiomics and artificial intelligence (AI), which includes machine learning (ML) and deep learning (DL) algorithms, have played an important role in mining additional information from medical images beyond human eyes' resolving power. Our review provides a summary of the current status of radiomics and AI in different clinical contexts of MM. A systematic search of PubMed, Web of Science, and Scopus was conducted, including all the articles published in English that explored radiomics and AI analyses of PET/CT images in MM. The initial results have highlighted the potential role of such new features in order to improve the clinical stratification of MM patients, as well as to increase their clinical benefits. However, more studies are warranted before these approaches can be implemented in clinical routines.

Predicting clinical outcome of sulfur mustard induced ocular injury using machine learning model

The sight-threatening sulfur mustard (SM) induced ocular injury presents specific symptoms in each clinical stage. The acute injury develops in all exposed eyes and may heal or deteriorate into chronic late pathology. Early detection of eyes at risk of developing late pathology may assist in providing unique monitoring and specific treatments only to relevant cases. In this study, we evaluated a machine-learning (ML) model for predicting the development of SM-induced late pathology based on clinical data of the acute phase in the rabbit model. Clinical data from 166 rabbit eyes exposed to SM vapor was used retrospectively. The data included a comprehensive clinical evaluation of the cornea, eyelids and conjunctiva using a semi-quantitative clinical score. A random forest classifier ML model, was trained to predict the development of corneal neovascularization four weeks post-ocular exposure to SM vapor using clinical scores recorded three weeks earlier. The overall accuracy in predicting the clinical outcome of SM-induced ocular injury was 73%. The accuracy in identifying eyes at risk of developing corneal neovascularization and future healed eyes was 75% and 59%, respectively. The most important parameters for accurate prediction were conjunctival secretion and corneal opacity at 1w and corneal erosions at 72 h post-exposure. Predicting the clinical outcome of SM-induced ocular injury based on the acute injury parameters using ML is demonstrated for the first time. Although the prediction accuracy was limited, probably due to the small dataset, it pointed out towards various parameters during the acute injury that are important for predicting SM-induced late pathology and revealing possible pathological mechanisms.

Recent advancements in machine learning and deep learning-based breast cancer detection using mammograms

OBJECTIVE

Mammogram-based automatic breast cancer detection has a primary role in accurate cancer diagnosis and treatment planning to save valuable lives. Mammography is one basic yet efficient test for screening breast cancer. Very few comprehensive surveys have been presented to briefly analyze methods for detecting breast cancer with mammograms. In this article, our objective is to give an overview of recent advancements in machine learning (ML) and deep learning (DL)-based breast cancer detection systems.

METHODS

We give a structured framework to categorize mammogram-based breast cancer detection techniques. Several publicly available mammogram databases and different performance measures are also mentioned.

RESULTS

After deliberate investigation, we find most of the works classify breast tumors either as normal-abnormal or malignant-benign rather than classifying them into three classes. Furthermore, DL-based features are more significant than hand-crafted features. However, transfer learning is preferred over others as it yields better performance in small datasets, unlike classical DL techniques.

SIGNIFICANCE AND CONCLUSION

In this article, we have made an attempt to give recent advancements in artificial intelligence (AI)-based breast cancer detection systems. Furthermore, a number of challenging issues and possible research directions are mentioned, which will help researchers in further scopes of research in this field.

PET/CT and SPECT/CT Imaging of HER2-Positive Breast Cancer

HER2 (Human Epidermal Growth Factor Receptor 2)-positive breast cancer is characterized by amplification of the HER2 gene and is associated with more aggressive tumor growth, increased risk of metastasis, and poorer prognosis when compared to other subtypes of breast cancer. HER2 expression is therefore a critical tumor feature that can be used to diagnose and treat breast cancer. Moving forward, advances in HER2 in vivo imaging, involving the use of techniques such as positron emission tomography (PET) and single-photon emission computed tomography (SPECT), may allow for a greater role for HER2 status in guiding the management of breast cancer patients. This will apply both to patients who are HER2-positive and those who have limited-to-minimal immunohistochemical HER2 expression (HER2-low), with imaging ultimately helping clinicians determine the size and location of tumors. Additionally, PET and SPECT could help evaluate effectiveness of HER2-targeted therapies, such as trastuzumab or pertuzumab for HER2-positive cancers, and specially modified antibody drug conjugates (ADC), such as trastuzumab-deruxtecan, for HER2-low variants. This review will explore the current and future role of HER2 imaging in personalizing the care of patients diagnosed with breast cancer.

Diagnostic Accuracy of Machine Learning AI Architectures in Detection and Classification of Lung Cancer: A Systematic Review

The application of artificial intelligence (AI) in diagnostic imaging has gained significant interest in recent years, particularly in lung cancer detection. This systematic review aims to assess the accuracy of machine learning (ML) AI algorithms in lung cancer detection, identify the ML architectures currently in use, and evaluate the clinical relevance of these diagnostic imaging methods. A systematic search of PubMed, Web of Science, Cochrane, and Scopus databases was conducted in February 2023, encompassing the literature published up until December 2022. The review included nine studies, comprising five case-control studies, three retrospective cohort studies, and one prospective cohort study. Various ML architectures were analyzed, including artificial neural network (ANN), entropy degradation method (EDM), probabilistic neural network (PNN), support vector machine (SVM), partially observable Markov decision process (POMDP), and random forest neural network (RFNN). The ML architectures demonstrated promising results in detecting and classifying lung cancer across different lesion types. The sensitivity of the ML algorithms ranged from 0.81 to 0.99, while the specificity varied from 0.46 to 1.00. The accuracy of the ML algorithms ranged from 77.8% to 100%. The AI architectures were successful in differentiating between malignant and benign lesions and detecting small-cell lung cancer (SCLC) and non-small-cell lung cancer (NSCLC). This systematic review highlights the potential of ML AI architectures in the detection and classification of lung cancer, with varying levels of diagnostic accuracy. Further studies are needed to optimize and validate these AI algorithms, as well as to determine their clinical relevance and applicability in routine practice.

An interpretable feature-learned model for overall survival classification of High-Grade Gliomas

PURPOSE

An accurate and well-defined survival prediction of High Grade Gliomas (HGGs) is indispensable because of its high incidence and aggressiveness. Therefore, this paper presents a unified framework for fully automatic overall survival classification and its interpretation.

METHODS AND MATERIALS

Initially, a glioma detection model is utilized to detect the tumorous images. A pre-processing module is designed for extracting 2D slices and creating a survival data array for the classification network. Then, the classification pipeline is integrated with two separate pathways: a modality-specific and a modality-concatenated pathway. The modality-specific pathway runs three separate CNNs for extracting rich predictive features from three sub-regions of HGGs (peritumoral edema, enhancing tumor and necrosis) by using three neuro-imaging modalities. In these pathways, the image vectors of the different modalities are also concatenated to the final fusion layer to overcome the loss of lower-level tumor features. Furthermore, to exploit the intra-modality correlations, a modality-concatenated pathway is also added to the classification pipeline. The experiments are conducted on BraTS 2018 and BraTS 2019 benchmarks, demonstrating that the proposed approach performs competitively in classifying HGG patients into three survival groups, namely, short, mid, and long survivors.

RESULTS

The proposed approach achieves an overall classification accuracy, sensitivity, and specificity of about 0.998, 0.997, and 0.999, respectively, for the BraTS 2018 dataset, and for BraTS 2019, these values correspond to 1.000, 0.999, and 0.999.

CONCLUSIONS

The results indicate that the proposed model achieves the highest values of the evaluation metrics for the overall survival classification of HGG.

Deep image and feature prior algorithm based on U-ConformerNet structure

PURPOSE

The reconstruction performance of the deep image prior (DIP) approach is limited by the conventional convolutional layer structure and it is difficult to enhance its potential. In order to improve the quality of image reconstruction and suppress artifacts, we propose a DIP algorithm with better performance, and verify its superiority in the latest case.

METHODS

We construct a new U-ConformerNet structure as the DIP algorithm's network, replacing the traditional convolutional layer-based U-net structure, and introduce the 'lpips' deep network based feature distance regularization method. Our algorithm can switch between supervised and unsupervised modes at will to meet different needs.

RESULTS

The reconstruction was performed on the low dose CT dataset (LoDoPaB). Our algorithm attained a PSNR of more than 35 dB under unsupervised conditions, and the PSNR under the supervised condition is greater than 36 dB. Both of which are better than the performance of the DIP-TV. Furthermore, the accuracy of this method is positively connected with the quality of the a priori image with the help of deep networks. In terms of noise eradication and artifact suppression, the DIP algorithm with U-ConformerNet structure outperforms the standard DIP method based on convolutional structure.

CONCLUSIONS

It is known by experimental verification that, in unsupervised mode, the algorithm improves the output PSNR by at least 2-3 dB when compared to the DIP-TV algorithm (proposed in 2020). In supervised mode, our algorithm approaches that of the state-of-the-art end-to-end deep learning algorithms.

EBHI: A new Enteroscope Biopsy Histopathological H&E Image Dataset for image classification evaluation

BACKGROUND AND PURPOSE

Colorectal cancer has become the third most common cancer worldwide, accounting for approximately 10% of cancer patients. Early detection of the disease is important for the treatment of colorectal cancer patients. Histopathological examination is the gold standard for screening colorectal cancer. However, the current lack of histopathological image datasets of colorectal cancer, especially enteroscope biopsies, hinders the accurate evaluation of computer-aided diagnosis techniques. Therefore, a multi-category colorectal cancer dataset is needed to test various medical image classification methods to find high classification accuracy and strong robustness.

METHODS

A new publicly available Enteroscope Biopsy Histopathological H&E Image Dataset (EBHI) is published in this paper. To demonstrate the effectiveness of the EBHI dataset, we have utilized several machine learning, convolutional neural networks and novel transformer-based classifiers for experimentation and evaluation, using an image with a magnification of 200×.

RESULTS

Experimental results show that the deep learning method performs well on the EBHI dataset. Classical machine learning methods achieve maximum accuracy of 76.02% and deep learning method achieves a maximum accuracy of 95.37%.

CONCLUSION

To the best of our knowledge, EBHI is the first publicly available colorectal histopathology enteroscope biopsy dataset with four magnifications and five types of images of tumor differentiation stages, totaling 5532 images. We believe that EBHI could attract researchers to explore new classification algorithms for the automated diagnosis of colorectal cancer, which could help physicians and patients in clinical settings.

Clinical prototype implementation enabling an improved day-to-day mammography compression

PURPOSE

In mammography, breast compression is achieved by lowering a compression paddle on the breast. Despite the directive that compression is needed, there is no concrete guideline on its execution. To estimate the degree of compression, current mammography units only provide compression force and breast thickness as parameters. Therefore, radiographers could be induced to mainly determine the level of compression based on compression force and apply the same value to all breast sizes. In this case, smaller breast sizes are exposed to higher pressure. This results in a highly varying perception of discomfort or even pain during the procedure, depending on the breast size.

METHODS

To overcome this imbalance, current research results suggest that pressure might be a more qualified parameter for a more uniform compression among all breast sizes. To utilize pressure, the contact area between breast and compression paddle must be determined. In this paper, we present an easy-to-implement prototype enabling a real-time pressure-based measure without the need of direct patient contact. Using an optical camera, the contact area between the breast and the compression paddle is automatically segmented by a deep learning model.

RESULTS

The model provides a mean pixel accuracy of 96.7% (SD: 2.3%), mean frequency-weighted intersection over union of 88.5% (SD: 6.3%), and a Dice score of 93.6% (SD: 2.2%). The subsequent pressure display is updated more than five times per second which enables the use in clinical routines to set the compression level.

CONCLUSION

This prototype could help guiding to an improved breast compression routine in mammography procedures.

Diagnostic test accuracy of artificial intelligence-based imaging for lung cancer screening: A systematic review and meta-analysis

BACKGROUND

Lung cancer is the principal cause of cancer-related deaths worldwide. Early detection of lung cancer with screening is indispensable to reduce the high morbidity and mortality rates. Artificial intelligence (AI) is widely utilised in healthcare, including in the assessment of medical images. A growing number of reviews studied the application of AI in lung cancer screening, but no overarching meta-analysis has examined the diagnostic test accuracy (DTA) of AI-based imaging for lung cancer screening.

OBJECTIVE

To systematically review the DTA of AI-based imaging for lung cancer screening.

METHODS

PubMed, EMBASE, Cochrane Library, CINAHL, IEEE Xplore, Web of Science, ACM Digital Library, Scopus, PsycINFO, and ProQuest Dissertations and Theses were searched from inception to date. Studies that were published in English and that evaluated the performance of AI-based imaging for lung cancer screening were included. Two independent reviewers screened titles and abstracts and used the Quality Assessment of Diagnostic Accuracy Studies-2 tool to appraise the quality of selected studies. Grading of Recommendations Assessment, Development, and Evaluation to diagnostic tests was used to assess the certainty of evidence.

RESULTS

Twenty-six studies with 150,721 imaging data were included. Hierarchical summary receiver-operating characteristic model used for meta-analysis demonstrated that the pooled sensitivity for AI-based imaging for lung cancer screening was 94.6 % (95 % CI: 91.4 % to 96.7 %) and specificity was 93.6 % (95 % CI: 88.5 % to 96.6 %). Subgroup analyses revealed that similar results were found among different types of AI, region, data source, and year of publication, but the overall quality of evidence was very low.

CONCLUSION

AI-based imaging could effectively detect lung cancer and be incorporated into lung cancer screening programs. Further high-quality DTA studies on large lung cancer screening populations are required to validate AI's role in early lung cancer detection.

Image-based motion artifact reduction on liver dynamic contrast enhanced MRI

Liver MRI images often suffer from degraded quality due to ghosting or blurring artifacts caused by patient respiratory or bulk motion. In this study, we developed a two-stage deep learning model to reduce motion artifact on dynamic contrast enhanced (DCE) liver MRIs. The stage-I network utilized a deep residual network with a densely connected multi-resolution block (DRN-DCMB) network to remove most motion artifacts. The stage-II network applied the generative adversarial network (GAN) and perceptual loss compensation to preserve image structural features. The stage-I network served as the generator of GAN and its pretrained parameters in stage-I were further updated via backpropagation during stage-II training. The stage-I network was trained using small image patches with simulated motion artifacts including image-space rotational and translational motion, and K-space based centric and interleaved linear motion, sinusoidal, and rotational motion to mimic liver motion patterns. The stage-II network training used full-size images with the same types of simulated motion. The liver DCE-MRI image volumes without obvious motion artifacts in 10 patients were used for the training process, of which 1020 images of 8 patients were used for training and 240 images of 2 patients for validation. Finally, the whole two-stage deep learning model was tested with simulated motion images (312 clean images from 5 test patients) and patient images with real motion artifacts (28 motion images from 12 patients). The resulted images after two-stage processing demonstrated reduced motion artifacts while preserved anatomic details without image blurriness, with SSIM of 0.935 ± 0.092, MSE of 60.7 ± 9.0 × 10^-3, and PSNR of 32.054 ± 2.219.

Beyond Imaging and Genetic Signature in Glioblastoma: Radiogenomic Holistic Approach in Neuro-Oncology

Glioblastoma (GBM) is a malignant brain tumor exhibiting rapid and infiltrative growth, with less than 10% of patients surviving over 5 years, despite aggressive and multimodal treatments. The poor prognosis and the lack of effective pharmacological treatments are imputable to a remarkable histological and molecular heterogeneity of GBM, which has led, to date, to the failure of precision oncology and targeted therapies. Identification of molecular biomarkers is a paradigm for comprehensive and tailored treatments; nevertheless, biopsy sampling has proved to be invasive and limited. Radiogenomics is an emerging translational field of research aiming to study the correlation between radiographic signature and underlying gene expression. Although a research field still under development, not yet incorporated into routine clinical practice, it promises to be a useful non-invasive tool for future personalized/adaptive neuro-oncology. This review provides an up-to-date summary of the recent advancements in the use of magnetic resonance imaging (MRI) radiogenomics for the assessment of molecular markers of interest in GBM regarding prognosis and response to treatments, for monitoring recurrence, also providing insights into the potential efficacy of such an approach for survival prognostication. Despite a high sensitivity and specificity in almost all studies, accuracy, reproducibility and clinical value of radiomic features are the Achilles heel of this newborn tool. Looking into the future, investigators' efforts should be directed towards standardization and a disciplined approach to data collection, algorithms, and statistical analysis.

PET-Derived Radiomics and Artificial Intelligence in Breast Cancer: A Systematic Review

Breast cancer (BC) is a heterogeneous malignancy that still represents the second cause of cancer-related death among women worldwide. Due to the heterogeneity of BC, the correct identification of valuable biomarkers able to predict tumor biology and the best treatment approaches are still far from clear. Although molecular imaging with positron emission tomography/computed tomography (PET/CT) has improved the characterization of BC, these methods are not free from drawbacks. In recent years, radiomics and artificial intelligence (AI) have been playing an important role in the detection of several features normally unseen by the human eye in medical images. The present review provides a summary of the current status of radiomics and AI in different clinical settings of BC. A systematic search of PubMed, Web of Science and Scopus was conducted, including all articles published in English that explored radiomics and AI analyses of PET/CT images in BC. Several studies have demonstrated the potential role of such new features for the staging and prognosis as well as the assessment of biological characteristics. Radiomics and AI features appear to be promising in different clinical settings of BC, although larger prospective trials are needed to confirm and to standardize this evidence.

Artificial intelligence performance in image-based ovarian cancer identification: A systematic review and meta-analysis

BACKGROUND

Accurate identification of ovarian cancer (OC) is of paramount importance in clinical treatment success. Artificial intelligence (AI) is a potentially reliable assistant for the medical imaging recognition. We systematically review articles on the diagnostic performance of AI in OC from medical imaging for the first time.

METHODS

The Medline, Embase, IEEE, PubMed, Web of Science, and the Cochrane library databases were searched for related studies published until August 1, 2022. Inclusion criteria were studies that developed or used AI algorithms in the diagnosis of OC from medical images. The binary diagnostic accuracy data were extracted to derive the outcomes of interest: sensitivity (SE), specificity (SP), and Area Under the Curve (AUC). The study was registered with the PROSPERO, CRD42022324611.

FINDINGS

Thirty-four eligible studies were identified, of which twenty-eight studies were included in the meta-analysis with a pooled SE of 88% (95%CI: 85-90%), SP of 85% (82-88%), and AUC of 0.93 (0.91-0.95). Analysis for different algorithms revealed a pooled SE of 89% (85-92%) and SP of 88% (82-92%) for machine learning; and a pooled SE of 88% (84-91%) and SP of 84% (80-87%) for deep learning. Acceptable diagnostic performance was demonstrated in subgroup analyses stratified by imaging modalities (Ultrasound, Magnetic Resonance Imaging, or Computed Tomography), sample size (≤300 or >300), AI algorithms versus clinicians, year of publication (before or after 2020), geographical distribution (Asia or non Asia), and the different risk of bias levels (≥3 domain low risk or < 3 domain low risk).

INTERPRETATION

AI algorithms exhibited favorable performance for the diagnosis of OC through medical imaging. More rigorous reporting standards that address specific challenges of AI research could improve future studies.

FUNDING

This work was supported by the Natural Science Foundation of China (No. 82073647 to Q-JW and No. 82103914 to T-TG), LiaoNing Revitalization Talents Program (No. XLYC1907102 to Q-JW), and 345 Talent Project of Shengjing Hospital of China Medical University (No. M0268 to Q-JW and No. M0952 to T-TG).

Machine learning framework for automatic image quality evaluation involving a mammographic American College of Radiology phantom

PURPOSE

The image quality (IQ) of mammographic images is essential when making a diagnosis, but the quality assurance process for radiological equipment is subjective. We therefore aimed to design an automatic IQ evaluation architecture based on a support vector machine (SVM) dedicated to evaluating images taken of mammography American College of Radiology (ACR) phantom.

METHODS

A total of 461 phantom images were acquired using mammographic equipment from 10 vendors. Two experienced medical physicists scored the images by consensus. The phantom datasets were randomly divided into training (80%) and testing (20%) sets. Each phantom image (with 6 fibers, 5 specks, and 5 masses) was detected by using bounding boxes, then cropped and divided into 16 pattern images. We identified 159 features for each pattern image. Manual scores were used to assign 3 labels (visible, invisible, and semivisible) to each pattern image. Multiclass-SVM models were trained with 3 types of patterns. Sub-datasets were randomly selected at 10% increments of the total dataset to determine a minimal effective training subset size for the automatic framework. A feature combination test and an analysis of variance were performed to identify the most influential features.

RESULTS

The accuracy of the model in evaluating fiber, speck, and mass patterns was 90.2%, 98.2%, and 88.9%, respectively. The performance was equivalent when the sample size was at least 138 (30% of 461) phantom images. The most influential feature was the position feature.

CONCLUSIONS

The proposed SVM-based automatic IQ evaluation framework applied to a mammographic ACR phantom accurately matched manual evaluations.

Current challenges of implementing artificial intelligence in medical imaging

The idea of using artificial intelligence (AI) in medical practice has gained vast interest due to its potential to revolutionise healthcare systems. However, only some AI algorithms are utilised due to systems' uncertainties, besides the never-ending list of ethical and legal concerns. This paper intends to provide an overview of current AI challenges in medical imaging with an ultimate aim to foster better and effective communication among various stakeholders to encourage AI technology development. We identify four main challenges in implementing AI in medical imaging, supported with consequences and past events when these problems fail to mitigate. Among them is the creation of a robust AI algorithm that is fair, trustable and transparent. Another issue is on data governance, in which best practices in data sharing must be established to promote trust and protect the patients' privacy. Next, stakeholders, such as the government, technology companies and hospital management, should come to a consensus in creating trustworthy AI policies and regulatory frameworks, which is the fourth challenge, to support, encourage and spur innovation in digital AI healthcare technology. Lastly, we discussed the efforts of various organizations such as the World Health Organisation (WHO), American College of Radiology (ACR), European Society of Radiology (ESR) and Radiological Society of North America (RSNA), who are already actively pursuing ethical developments in AI. The efforts by various stakeholders will eventually overcome hurdles and the deployment of AI-driven healthcare applications in clinical practice will become a reality and hence lead to better healthcare services and outcomes.

Intracerebral hemorrhage detection on computed tomography images using a residual neural network

Intracerebral hemorrhage (ICH) is a high mortality rate, critical medical injury, produced by the rupture of a blood vessel of the vascular system inside the skull. ICH can lead to paralysis and even death. Therefore, it is considered a clinically dangerous disease that needs to be treated quickly. Thanks to the advancement in machine learning and the computing power of today's microprocessors, deep learning has become an unbelievably valuable tool for detecting diseases, in particular from medical images. In this work, we are interested in differentiating computer tomography (CT) images of healthy brains and ICH using a ResNet-18, a deep residual convolutional neural network. In addition, the gradient-weighted class activation mapping (Grad-CAM) technique was employed to visually explore and understand the network's decisions. The generalizability of the detector was assessed through a 100-iteration Monte Carlo cross-validation (80% of the data for training and 20% for test). In a database with 200 CT images of brains (100 with ICH and 100 without ICH), the detector yielded, on average, 95.93%accuracy, 96.20% specificity, 95.65% sensitivity, 96.40% precision, and 95.91% F1-core, with an average computing time of 165.90 s to train the network (on 160 images) and 1.17 s to test it with 40 CT images. These results are comparable with the state of the art with a simpler and lower computational load approach. Our detector could assist physicians in their medical decision, in resource optimization and in reducing the time and error in the diagnosis of ICH.

Automatic head computed tomography image noise quantification with deep learning

PURPOSE

Computed tomography (CT) image noise is usually determined by standard deviation (SD) of pixel values from uniform image regions. This study investigates how deep learning (DL) could be applied in head CT image noise estimation.

METHODS

Two approaches were investigated for noise image estimation of a single acquisition image: direct noise image estimation using supervised DnCNN convolutional neural network (CNN) architecture, and subtraction of a denoised image estimated with denoising UNet-CNN experimented with supervised and unsupervised noise2noise training approaches. Noise was assessed with local SD maps using 3D- and 2D-CNN architectures. Anthropomorphic phantom CT image dataset (N = 9 scans, 3 repetitions) was used for DL-model comparisons. Mean square error (MSE) and mean absolute percentage errors (MAPE) of SD values were determined using the SD values of subtraction images as ground truth. Open-source clinical head CT low-dose dataset (N_train = 37, N_test = 10 subjects) were used to demonstrate DL applicability in noise estimation from manually labeled uniform regions and in automated noise and contrast assessment.

RESULTS

The direct SD estimation using 3D-CNN was the most accurate assessment method when comparing in phantom dataset (MAPE = 15.5%, MSE = 6.3HU). Unsupervised noise2noise approach provided only slightly inferior results (MAPE = 20.2%, MSE = 13.7HU). 2DCNN and unsupervised UNet models provided the smallest MSE on clinical labeled uniform regions.

CONCLUSIONS

DL-based clinical image assessment is feasible and provides acceptable accuracy as compared to true image noise. Noise2noise approach may be feasible in clinical use where no ground truth data is available. Noise estimation combined with tissue segmentation may enable more comprehensive image quality characterization.

Validation of deep learning-based nonspecific estimates for amyloid burden quantification with longitudinal data

PURPOSE

To validate our previously proposed method of quantifying amyloid-beta (Aβ) load using nonspecific (NS) estimates generated with convolutional neural networks (CNNs) using [¹⁸F]Florbetapir scans from longitudinal and multicenter ADNI data.

METHODS

188 paired MR (T1-weighted and T2-weighted) and PET images were downloaded from the ADNI3 dataset, of which 49 subjects had 2 time-point scans. 40 Aβ- subjects with low specific uptake were selected for training. Multimodal ScaleNet (SN) and monomodal HighRes3DNet (HRN), using either T1-weighted or T2-weighted MR images as inputs) were trained to map structural MR to NS-PET images. The optimized SN and HRN networks were used to estimate the NS for all scans and then subtracted from SUVr images to determine the specific amyloid load (SAβ_L) images. The association of SAβ_L with various cognitive and functional test scores was evaluated using Spearman analysis, as well as the differences in SAβ_L with cognitive test scores for 49 subjects with 2 time-point scans and sensitivity analysis.

RESULTS

SAβ_L derived from both SN and HRN showed higher association with memory-related cognitive test scores compared to SUVr. However, for longitudinal scans, only SAβ_L estimated from multimodal SN consistently performed better than SUVr for all memory-related cognitive test scores.

CONCLUSIONS

Our proposed method of quantifying Aβ load using NS estimated from CNN correlated better than SUVr with cognitive decline for both static and longitudinal data, and was able to estimate NS of [¹⁸F]Florbetapir. We suggest employing multimodal networks with both T1-weighted and T2-weighted MR images for better NS estimation.

Outcome Prediction for SARS-CoV-2 Patients Using Machine Learning Modeling of Clinical, Radiological, and Radiomic Features Derived from Chest CT Images

(1) Background: Chest Computed Tomography (CT) has been proposed as a non-invasive method for confirming the diagnosis of SARS-CoV-2 patients using radiomic features (RFs) and baseline clinical data. The performance of Machine Learning (ML) methods using RFs derived from semi-automatically segmented lungs in chest CT images was investigated regarding the ability to predict the mortality of SARS-CoV-2 patients. (2) Methods: A total of 179 RFs extracted from 436 chest CT images of SARS-CoV-2 patients, and 8 clinical and 6 radiological variables, were used to train and evaluate three ML methods (Least Absolute Shrinkage and Selection Operator [LASSO] regularized regression, Random Forest Classifier [RFC], and the Fully connected Neural Network [FcNN]) for their ability to predict mortality using the Area Under the Curve (AUC) of Receiver Operator characteristic (ROC) Curves. These three groups of variables were used separately and together as input for constructing and comparing the final performance of ML models. (3) Results: All the ML models using only RFs achieved an informative level regarding predictive ability, outperforming radiological assessment, without however reaching the performance obtained with ML based on clinical variables. The LASSO regularized regression and the FcNN performed equally, both being superior to the RFC. (4) Conclusions: Radiomic features based on semi-automatically segmented CT images and ML approaches can aid in identifying patients with a high risk of mortality, allowing a fast, objective, and generalizable method for improving prognostic assessment by providing a second expert opinion that outperforms human evaluation.

Recent Applications of Artificial Intelligence in Radiotherapy: Where We Are and Beyond

In recent decades, artificial intelligence (AI) tools have been applied in many medical fields, opening the possibility of finding novel solutions for managing very complex and multifactorial problems, such as those commonly encountered in radiotherapy (RT). We conducted a PubMed and Scopus search to identify the AI application field in RT limited to the last four years. In total, 1824 original papers were identified, and 921 were analyzed by considering the phase of the RT workflow according to the applied AI approaches. AI permits the processing of large quantities of information, data, and images stored in RT oncology information systems, a process that is not manageable for individuals or groups. AI allows the iterative application of complex tasks in large datasets (e.g., delineating normal tissues or finding optimal planning solutions) and might support the entire community working in the various sectors of RT, as summarized in this overview. AI-based tools are now on the roadmap for RT and have been applied to the entire workflow, mainly for segmentation, the generation of synthetic images, and outcome prediction. Several concerns were raised, including the need for harmonization while overcoming ethical, legal, and skill barriers.

Enhancing the impact of Artificial Intelligence in Medicine: A joint AIFM-INFN Italian initiative for a dedicated cloud-based computing infrastructure

Artificial Intelligence (AI) techniques have been implemented in the field of Medical Imaging for more than forty years. Medical Physicists, Clinicians and Computer Scientists have been collaborating since the beginning to realize software solutions to enhance the informative content of medical images, including AI-based support systems for image interpretation. Despite the recent massive progress in this field due to the current emphasis on Radiomics, Machine Learning and Deep Learning, there are still some barriers to overcome before these tools are fully integrated into the clinical workflows to finally enable a precision medicine approach to patients' care. Nowadays, as Medical Imaging has entered the Big Data era, innovative solutions to efficiently deal with huge amounts of data and to exploit large and distributed computing resources are urgently needed. In the framework of a collaboration agreement between the Italian Association of Medical Physicists (AIFM) and the National Institute for Nuclear Physics (INFN), we propose a model of an intensive computing infrastructure, especially suited for training AI models, equipped with secure storage systems, compliant with data protection regulation, which will accelerate the development and extensive validation of AI-based solutions in the Medical Imaging field of research. This solution can be developed and made operational by Physicists and Computer Scientists working on complementary fields of research in Physics, such as High Energy Physics and Medical Physics, who have all the necessary skills to tailor the AI-technology to the needs of the Medical Imaging community and to shorten the pathway towards the clinical applicability of AI-based decision support systems.

Strategies to develop radiomics and machine learning models for lung cancer stage and histology prediction using small data samples

Predictive models based on radiomics and machine-learning (ML) need large and annotated datasets for training, often difficult to collect. We designed an operative pipeline for model training to exploit data already available to the scientific community. The aim of this work was to explore the capability of radiomic features in predicting tumor histology and stage in patients with non-small cell lung cancer (NSCLC). We analyzed the radiotherapy planning thoracic CT scans of a proprietary sample of 47 subjects (L-RT) and integrated this dataset with a publicly available set of 130 patients from the MAASTRO NSCLC collection (Lung1). We implemented intra- and inter-sample cross-validation strategies (CV) for evaluating the ML predictive model performances with not so large datasets. We carried out two classification tasks: histology classification (3 classes) and overall stage classification (two classes: stage I and II). In the first task, the best performance was obtained by a Random Forest classifier, once the analysis has been restricted to stage I and II tumors of the Lung1 and L-RT merged dataset (AUC = 0.72 ± 0.11). For the overall stage classification, the best results were obtained when training on Lung1 and testing of L-RT dataset (AUC = 0.72 ± 0.04 for Random Forest and AUC = 0.84 ± 0.03 for linear-kernel Support Vector Machine). According to the classification task to be accomplished and to the heterogeneity of the available dataset(s), different CV strategies have to be explored and compared to make a robust assessment of the potential of a predictive model based on radiomics and ML.