Radiomic analysis là gì

 Radiomics

Dr Henry Knipe◉◈ and Dr Muhammad Idris◉ et al. From Radiopaedia

Radiomics (as applied to radiology) is a field of medical study that aims to extract a large number of quantitative features from medical images using data characterization algorithms. The data is assessed for improved decision support. It has the potential to uncover disease characteristics that are difficult to identify by human vision alone.

Process

The process of creating a database of correlative quantitative features, which can be used to analyze subsequent (unknown) cases includes the following steps ³.

Initial image processing

Using a variety of reconstruction algorithms such as contrast, edge enhancement, etc. This influences the quality and usability of the images, which in turn determines how easily and accurately an abnormal characteristic could be detected and characterized.

Image segmentation

Identify/create areas (2D images) or volumes of interest (3D images). Can be done either manually, semi-automated, or fully automated using artificial intelligence.

For large data sets, an automated process is needed because manual techniques are usually very time-consuming and tend to be less accurate, less reproducible and less consistent compared with automated artificial intelligence techniques.

Features extraction and qualification

Features include volume, shape, surface, density, and intensity, texture, location, and relations with the surrounding tissues.

Semantic features are those that are commonly used in the radiology lexicon to describe regions of interest.

Agnostic features are those that attempt to capture lesion heterogeneity through quantitative mathematical descriptors.

Examples of semantic features

·         shape

·         location

·         vascularity

·         spiculation

·         necrosis

·         attachments

Equivalent examples of agnostic features

·         histogram (skewness, kurtosis)

·         Haralick textures

·         Laws textures

·         wavelets

·         Laplacian transforms

·         Minkowski functions

·         fractal dimensions

Uses

Radiomics can be applied to most imaging modalities including radiographs, ultrasound, CT, MRI and PET studies. It can be used to increase the precision in the diagnosis, assessment of prognosis, and prediction of therapy response, particularly in combination with clinical, biochemical, and genetic data. The technique has been used in oncological studies, but potentially can be applied to any disease.

A typical example of radiomics is using texture analysis to correlate molecular and histological features of diffuse high-grade gliomas ². 

The determination of most discriminatory radiomics feature extraction methods varies with the modality of imaging and the pathology studied and is therefore currently (c.2019) the focus of research in the field of radiomics.

Current challenges include the development of a common nomenclature, image data sharing, large computing power and storage requirements, and validating models across different imaging platforms and patient populations.

References

Promoted articles (advertising)

1.    MAPS: A Quantitative Radiomics Approach for Prostate Cancer Detection

EMBS Trans Biomed Eng, 2015

2.    Robust Collaborative Clustering of Subjects and Radiomic Features for Cancer Prognosis

EMBS Trans Biomed Eng, 2020

3.    Pattern Classification for Gastrointestinal Stromal Tumors by Integration of Radiomics and Deep Convolutional Features

EMBS J Biomed Health Inform, 2018

1.    Multi-Objective-Based Radiomic Feature Selection for Lesion Malignancy Classification

EMBS J Biomed Health Inform, 2019

2.    Benefits and limitations of real-world evidence: lessons from EGFR mutation-positive NSCLC

Bassel Nazha et al., Future Oncology, 2020

3.    A Radiomics Approach With CNN for Shear-Wave Elastography Breast Tumor Classification

EMBS Trans Biomed Eng, 2018

 

 

 

NANG GAN, HEMANGIOMA VÀ U GAN QUA THUẬT TOÁN A I VÀ DEEP LEARNING

TRÍ TUỆ THÔNG MINH CHẨN ĐOÁN U GAN

 

...

Focal liver lesion detection 

Deep learning algorithms combined with multiple image modalities have been widely used in the detection of focal liver lesions (Table 2). The combination of deep learning methods with CNNs and CT for liver disease diagnosis has gained wide attention[35] . Compared with the visual assessment, this strategy may capture more detailed lesion features and make more accurate diagnosis. According to Vivantil et al by using deep learning models based on longitudinal liver CT studies, new liver tumors could be detected automatically with a true positive rate of 86%, while the stand-alone detection rate was only 72% and this method achieved a precision of 87% and an improvement of 39% over the traditional SVM mode[36] . Some studies[37-39] have also used CNNs based on CT to detect liver tumors automatically, but these machine learning methods may not reliably detect new tumors because of the insufficient representativeness of small new tumors in the training data. Ben-Cohen et al developed a CNN model predicting the primary origin of liver metastasis among four sites (melanoma, colorectal cancer, pancreatic cancer, and breast cancer) with CT images[40] . In the task of automatic multiclass categorization of liver metastatic lesions, the automated system was able to achieve a 56% accuracy for the primary sites. If the prediction was made as top-2 and top-3 classification tasks, the accuracy could be up to 0.83 and 0.99, respectively. These automated systems may provide favorable decision support for physicians to achieve more efficient treatment. CNN models which use ultrasound images to detect liver lesions were also developed. According to Liu et al by using a CNN model based on liver ultrasound images, the proposed method can effectively extract the liver capsules and accurately diagnose liver cirrhosis, with the diagnostic AUC being able to reach 0.968. Compared with two kinds of low level feature extraction method histogram of oriented gradients (HOG) and local binary pattern (LBP), whose mean accuracy rates were 83.6% and 81.4%, respectively, the deep learning method achieved a better classification accuracy of 86.9%[41] . It was reported that deep learning system using CNN showed a superior performance for fatty liver disease detection and risk stratification compared to conventional machine learning systems with the detection and risk stratification accuracy of 100%[42] . Hassan et al used the sparse auto encoder to access the representation of the liver ultrasound image and utilized the softmax layer to detect and distinguish different focal liver diseases. They found that the deep learning method achieved an overall accuracy of 97.2% compared with the accuracy rates of multi-SVM, KNN(K-Nearest Neighbor), and naive Bayes, which were 96.5, 93.6, and 95.2%, respectively[43] . An ANN based on 18F-FDG PET/CT scan, demographic, and laboratory data showed a high sensitivity and specificity to detect liver malignancy and had a highly significant correlation with MR imaging findings which served as the reference standard[44] . The AUCs of lesion-dependent network and lesion-independent network were 0.905 (standard error, 0.0370) and 0.896 (standard error, 0.0386), respectively. The automated neural network could help identify nonvisually apparent focal FDG uptake in the liver, which was possibly positive for liver malignancy, and serve as a clinical adjunct to aid in interpretation of PET images of the liver.

CHALLENGES AND FUTURE PERSPECTIVES 

There is considerable controversy about the time needed to implement fully automated clinical tasks by deep learning methods[59] . The debated time ranges from a few years to decades. The automated solutions based on deep learning aim to solve the most common clinical problems which demand a lot of long-term accumulation of expertise or are much too complicated for human readers, for example, lung screening CT, mammograms and so on. Next, researchers need to develop more advanced deep learning algorithms to solve more complex medical imaging problems, such as ultrasound or PET. At present, a common shortage of AI tools is that they cannot resolve multiple tasks. There is currently no comprehensive AI system capable of detecting multiple abnormalities throughout the human body. A great amount of medical data which are electronically organized and amassed in a systematic style facilitate access and retrieval by researchers. However, the lack of curation of the training data is a major drawback in learning any AI model. To select relevant patient cohort for specific AI task or make segmentation within images is essential and helpful. Some segmentation algorithms using AI[60] are not perfect to curate data, as they always need human experts to verify accuracy. Unsupervised learning which includes generative adversarial networks[61] and variational autoencoders[62] may achieve automated data curation by learning discriminatory features without explicit labeling. Many studies have explored the possibilities of unsupervised learning application in brain MRI[63] and mammography[64] and more field applications of this state of the art method are needed. It is of great significance to indicate that AI is different from human intelligence in numerous ways. Although various forms of AI have exceeded human performance, they lacked higher-level background knowledge and failed to establish associations like the human brain.

In addition, AI is trained for one task only. The AI field of medical imaging is still in its infancy, especially in the ultrasound field. It is almost impossible for AI to replace radiologists in the coming decades, but radiologists using AI will inevitably replace radiologists who do not. With the advancement of AI technology, radiologists will achieve an increased accuracy with higher efficiency. We also need to call for advocacy for creating interconnected networks of identifying patient data from around the world and training AI on a large scale according to different patient demographics, geographic areas, diseases, etc. Only in this way can we create an AI that is socially responsible and benefits more people.

Ultrasound outperforms x-ray for neonatal pneumothorax

By Theresa Pablos, AuntMinnie staff writer

December 23, 2020 -- Lung ultrasound (LUS) scans outperformed chest x-rays for diagnosing neonatal pneumothorax in a new review that included more than 500 newborns. Ultrasound achieved better sensitivity and specificity and took less time to perform, according to the December 16 study in Ultrasound in Medicine & Biology.

Pneumothorax is a common but life-threatening illness seen in neonatal intensive care units. While CT is the gold standard for diagnosing pneumothorax in adults, chest x-ray is the preferred modality for newborns in order to reduce exposure to ionizing radiation.

Still, chest x-ray has its limitations for neonates. Newborns are especially at risk for latent effects from repeated exposure to ionizing radiation, and it can be difficult to detect pneumothorax with chest x-ray.

Instead, lung ultrasound may be the better first-line imaging modality for diagnosing pneumothorax in infants. Not only does ultrasound not expose infants to ionizing radiation -- it also appears to be more accurate.

"LUS is a new choice for the diagnosis and treatment of neonatal [pneumothorax]," wrote the authors, led by Qiang Fei, PhD, a professor at Zhejiang University School of Medicine in Hangzhou, China. "Compared with [chest x-ray], ultrasound combines the advantages of bedside diagnosis, avoidance of irradiation, cost-effectiveness, high accuracy, and reliability."

For the review, Fei and colleagues reviewed both Chinese- and English-language databases to find prospective studies investigating the diagnostic performance of chest x-ray and lung ultrasound for neonatal pneumothorax. Eight studies with a total of 529 infants met their inclusion criteria.

Lung ultrasound performed better than chest x-ray in the review. Ultrasound netted a sensitivity of 98% and specificity 99%, compared to 82% and 96% for chest radiography. Furthermore, ultrasound achieved an area under the curve of 0.997 and was faster to perform in five out of the eight studies.

The authors also calculated the diagnostic odds ratio (DOR) for both modalities -- a measurement of the effectiveness of a diagnostic test where higher numbers represent more effectiveness. Lung ultrasound achieved a DOR of 920, while chest radiographs had a DOR of 45.

Chest x-ray vs. lung ultrasound for neonatal pneumothorax
	Chest x-ray	Lung ultrasound
Sensitivity	82%	98%
Specificity	96%	99%
Diagnostic odds ratio (DOR)	45	920

"[Chest x-ray] is associated with a certain rate of misdiagnosis and is less sensitive than LUS for the diagnosis of mild-to-moderate [pneumothorax], especially in premature infants," Fei and colleagues wrote.

They theorized that chest x-ray may have limited usefulness in this population because the lesions are small and can be deep in the lungs. Meanwhile, ultrasound may be better suited to imaging the thin chest walls and narrow thorax of newborns.

The authors didn't have enough studies to sufficiently evaluate the accuracy of lung ultrasound features for diagnosing pneumothorax. However, the disappearance of lung sliding and B-lines and the presence of A-lines looked promising as diagnostic markers of the illness. In addition, the presence or absence of lung points -- where the visceral and parietal pleural surfaces meet -- looked useful for helping to determine illness severity.

Based on the findings, the researchers recommended lung ultrasound as a first-line modality for diagnosing pneumothorax in this population.

"[Chest x-ray] could be carried out as the second-line procedure if there are doubts about the findings during LUS examination, such as examination of neonates with large-area atelectasis," they wrote.

Elastography shows promise for rotator cuff tears

By Theresa Pablos, AuntMinnie staff writer

December 4, 2020 -- Ultrasound with a shear-wave elastography (SWE) technique was comparable to MRI for preoperative evaluation of rotator cuff tears, according to a Thursday presentation at the RSNA 2020 virtual meeting. Elastography measurements showed moderate correlation with MR metrics in the 60-patient study.

Assessing muscle quality is critical for planning surgery to repair the supraspinatus muscle in patients with a torn rotator cuff. While MRI has been long used to evaluate rotator cuff muscles, SWE has emerged as a promising new metric that can be performed at the point of care in order to measure muscle elasticity.

"SWE may be useful to predict tendon repairability by evaluating muscle quality," said presenter Dr. Eun Kyung Khil from the radiology department at Hallym University Dongtan Sacred Heart Hospital in Hwaseong, South Korea.

The prospective study compared ultrasound SWE measurements and conventional MRI metrics for predicting whether surgery to repair the supraspinatus muscle would be successful. It included 60 patients with supraspinatus tears who underwent both preoperative MRI and ultrasound scans between May 2019 and August 2020.

One radiologist with five years of musculoskeletal imaging experience performed the ultrasound scans, which the researchers used to calculate the mean elasticity score, median elasticity score, and elasticity ratio.

Elasticity was calculated using a longitudinal ultrasound scan of three regions of interest of the supraspinatus muscle, and the scan was repeated three times in order to have nine total region-of-interest measurements. Meanwhile, the elasticity ratio was calculated by dividing the mean elasticity of the supraspinatus muscle by mean elasticity of the trapezius muscle.

In addition, two radiologists read MRI images, which were acquired with a 3-tesla system. The researchers used the following three standard tools to measure muscle evaluation:

• Goutallier grade system to account for fat-to-muscle ratio
• Occupation ratio of the area of supraspinatus muscle to the supraspinatus fossa
• Warner's muscle atrophy grade

The authors compared the ultrasound and MRI measurements for patients whose surgery was successful, defined as a complete or near-complete repair of the rotator cuff, to patients with an incomplete rotator cuff repair.

MRI and SWE measurements in patients who underwent supraspinatus repair surgery
		Complete repair	Incomplete repair	p-value
MRI	Goutallier grade	1.8	3.78	< 0.001
	Occupation ratio	59.88	31.56	< 0.001
	Muscle atrophy grade	0.39	2.33	< 0.001
SWE	Emean, kPa	31.25	43.84	< 0.001
	Emedian, kPa	29.9	43.54	< 0.001
	Eratio(SST/Tra)	1.84	3.68	< 0.001

Patients with a successful rotator cuff repair surgery had significantly higher scores on both ultrasound and MRI. Both the mean and median elastography measurements were at least 10 kPa higher in the incomplete repair group.

In addition, Khil said the sensitivity and specificity of SWE were high when using a mean elasticity cutoff value of 35.06 kPa and an elasticity ratio cutoff value of 2.61.

Furthermore, the three elastography measurements on ultrasound showed decent correlation with the MRI metrics. The correlation was particularly strong for the elasticity ratio, which had a coefficient agreement of 0.57 with Goutallier grade and 0.66 with muscle atrophy grade on MRI.

"The correlation coefficient was over 0.4, showing a significant moderate correlation, especially in elasticity ratio," Khil said.

The findings were limited by a small number of participants, especially those in the failed repair group. However, it still demonstrated that SWE looks promising for preoperative evaluation of rotator cuff tears.

"In the preoperative evaluation of [supraspinatus] muscle quality using SWE, especially elasticity ratio, showed moderate correlation with existing MR measurements," Khil said.

MRI, U S diagnose post-COVID-19 muscle weakness

By Kate Madden Yee, AuntMinnie.com staff writer

December 1, 2020 -- A combination of MR neurography and ultrasound could help clinicians better diagnose what is causing some recovered COVID-19 patients to continue to experience chronic pain, numbness, or weakness in their hands or limbs, according to commentary published December 1 in Radiology.

The article suggests that combining the two imaging modalities is an effective way to identify nerve damage that results from COVID-19, especially as patients recover and are tracked long-term, lead author Dr. Swati Deshmukh of Northwestern University in Chicago said in a statement released by the university.

"There are physicians out there who are seeing these otherwise young, healthy patients, and they don't know exactly what's wrong and they're thinking, 'What am I supposed to do for patients with post-COVID pain and weakness?' " Deshmukh said. "I want physicians and patients to be aware of the diagnostic options available due to recent innovations in technology and inquire if advanced imaging might be right for them."

As the COVID-19 pandemic has continued, clinicians have observed neuromuscular complications of the illness as those who have recovered participate in rehabilitative care. Imaging is a key tool for evaluating what causes these complications -- from inflammatory neuropathy, to prone positioning-related compression injuries, to nerve entrapment as the result of hematoma, Deshmukh's group noted.

An MR image of a patient in their early 20s shows nerve injury of the left brachial plexus in the neck. The patient experienced left arm weakness and pain after recovering from COVID-19 respiratory illness, which prompted them to see their primary care physician. As a result of the MRI findings, the patient was referred to the COVID-19 neurology clinic for treatment. Image and caption courtesy of Northwestern University.

Determining the cause makes treatment more effective, according to the team. If the cause is due to injury from prone positioning, a patient would be referred for rehabilitation or peripheral nerve surgery. If nerve damage has been caused by inflammatory response, the patient should see a neurologist. And if the damage is caused by hematoma, blood thinner medications should be adjusted and surgery may be necessary, according to the group.

Ultrahigh-resolution ultrasound and MR neurography (which visualizes the peripheral nerves) can localize the problem and assess the severity of nerve damage and whether that damage has affected the muscles, according to the team.

"Peripheral nerve imaging aids diagnosis and may guide management in COVID-19 patients with neuromuscular symptoms arising from the infectious disease, hospitalization course, or secondary to a complication in treatment

US AI solves real-world clinical problems

 

AI helps solve real-world clinical problems

By Adam Davidson, AuntMinnie.com contributing writer

AI can walk ultrasound users through scans and enhance the image, ensuring the acquisition of high-quality and consistent scans that can be easily interpreted.

AI can also identify anatomy and anomalies on scans and make measurements to help readers interpret scans. AI can serve as a safeguard, ensuring that the radiologist does not miss any areas of concern on a scan, and as a second opinion, improving the radiologist's confidence in their diagnosis.

The chart below lists the importance of various AI tasks, as indicated in an Omdia survey of users.

According to the AIUS report, AI for ultrasound is versatile, with 60% of respondents' practices using AI in multiple clinical applications. AI utilization in general imaging was the most frequently reported clinical application, but Omdia expects AI utilization in point-of-care (POC), cardiology, and nontraditional applications to increase during the next few years as image libraries grow, enabling the development of more specialized algorithms.

The development of AI at the edge of care and in POC settings will drive integrated and cloud-based AI deployment in addition to the use of AI with portable medical imaging equipment.

..,

ED labels 84% of its abdominal ultrasound exams as inappropriate, causing downstream problems

ED labels 84% of its abdominal ultrasound exams as inappropriate, causing downstream problems

Matt O'Connor | November 16, 2020 | Diagnostic Imaging

 

Abdominal pain is among the reasons patients visit the emergency department, with CT and ultrasound both serving as front-line tests to assess such complaints. New data suggests, however, that many of these exams are ordered inappropriately, leading to negative downstream consequences. 

That’s what researchers discovered after analyzing more than 250 exams completed at a non-trauma tertiary care hospital over a three-month span. Based on the American College of Radiology appropriateness criteria, 36% of CT scans were inappropriately ordered along with 84% of ultrasound exams. 

Guideline-discordant US images also caused providers to utilize an additional imaging modality in 20% of cases, causing longer ED stays, extra tests, and added costs, the authors wrote Sunday in Current Problems in Diagnostic Radiology. 

The reasons for these high numbers are multifaceted, but malpractice fears and concerns over missing a low-probability diagnosis likely top the list, Martina Zaguini Francisco, MD, with the Federal University of Health Sciences of Porto Alegre in Brazil, and colleagues explained. 

The team also pointed to educational gaps as a sizable problem.

“Although there is wide and ready dissemination of ACR tools, the lack of awareness of existing guidelines remains a major problem,” Zaguini et al. wrote. “This results not only in imaging overuse but also in wrong modalities being requested, leading to additional imaging orders during the same visit.”

More than 85% of emergency physicians admit to ordering too many tests, the researchers explained. And exams may be considered inappropriate for many reasons, including choosing the wrong modality, opting for one test over a more appropriate first-line exam, or if a test doesn’t change therapeutic management. 

With this in mind, Zaguini et al. retrospectively reviewed 135 CT and 143 US exams ordered for abdominal complaints in the ED between January and March 2019. 

They found that appropriately ordered exams were “significantly” more likely to yield findings compatible with clinicians’ initial diagnosis.

“This highlights the high impact that correct exam selection has on finding confirmative or actionable results on imaging,” the authors wrote.

Based on ACR Appropriateness Criteria, inappropriate CT scans were most often ordered for biliary disease, pancreatitis, renal failure, and uncomplicated pyelonephritis. 

For ultrasound, meanwhile, discordant exams were typically requested for acute abdominal pain, uncomplicated pyelonephritis, diverticulitis, and appendicitis. 

The study was limited by its single-center design and may not be generalizable to other organizations, the authors noted.