Can AI reliably measure carotid intima-media thickness?

By Erik L. Ridley, AuntMinnie staff writer

May 25, 2018 -- Artificial intelligence (AI) software that combines deep-learning and machine-learning techniques can measure carotid intima-media thickness (CIMT) more accurately than sonographers can, according to research in the July 1 issue of Computers in Biology and Medicine.

In testing, a multi-institutional and multinational team led by Mainak Biswas of the National Institute of Technology Goa in India found that its deep learning-based model outperformed previous automated methods. The model was also up to 20% more accurate than sonographers in measuring CIMT, an important biomarker for cardiovascular disease and stroke monitoring.

"The results showed that the performance of the [deep learning]-based approach was superior to the nonintelligence-based conventional methods that use spatial intensities alone," the authors wrote. "The [deep-learning] system can be used for stroke risk assessment during routine or clinical trial modes."

An important biomarker

An increase in CIMT -- the mean perpendicular distance between the lumen-intima (LI) and the media-adventitia (MA) interfaces -- has been associated with an increased risk of cardiovascular events and stroke. However, the current process of measuring CIMT suffers from accuracy and reproducibility issues due to factors such as variability in patient nationality, ethnicity, disease, and age group. Technical challenges also play a role; traditional manual segmentation of these regions is slow, error-prone, and subject to intraobserver and interobserver variability, according to the researchers (Comput Biol Med, July 1, 2018, Vol. 98, pp. 100-117).

As a result, a number of automated techniques have been developed for predicting CIMT, using various spatial features such as grayscale median, pixel classification, gradient edges, space-scale, or a combination of these features, the researchers explained.

"Despite their strong contributions, these external factors make the spatial-based methods prone to variability and a lack of robustness when it comes to completely automated designs," they wrote.

Biswas and colleagues hypothesized that a deep-learning system would be more reliable and accurate than previous methods, thanks to its inherent ability to provide better regional segmentation output. To train and test a deep-learning model, the researchers used 396 high-resolution B-mode ultrasound images of the right and left common carotid artery from 203 patients at Toho University in Japan. The ultrasound scans were obtained on one of three ultrasound systems (Aplio XV, Aplio XG, and Xario) from Canon Medical Systems. Of the 396 images, 90% were used for training the deep-learning model and 10% were set aside for testing.

Manual tracing of the lumen and adventitia borders was performed using ImgTracer software (AtheroPoint). Dr. Jasjit Suri, PhD, from AtheroPoint served as senior author on the study.

A 2-stage system

The researchers developed a two-stage system that made use of both deep and machine learning. In the first stage, a convolution layer-based encoder was used to extract image features, while a decoder based on a fully convolutional neural network (CNN) performed image segmentation. The raw inner lumen borders and raw outer interadventitial borders generated during this process were then smoothed with a machine learning-based method. The model utilized these final borders to calculate CIMT values from the LI and MA far walls using the standardized polyline distance metric method.

As two different sets of gold standards -- lumen regional information and interadventitial regional information -- were used during the design of the deep-learning model, the researchers trained and evaluated two different algorithms. Compared with the gold standard, the two deep-learning algorithms yielded CIMT error rates of 0.126 ± 0.134 mm and 0.124 ± 1.0 mm. They also significantly outperformed previously developed systems for measuring CIMT, according to the researchers.

Biswas and colleagues also compared the performance of the deep-learning algorithms with mean far-wall CIMT measurements calculated by sonographers in 346 images. Both models correlated better with the ground truth than the sonographer measurements, which had been performed in real-time in the institution's vascular ultrasound laboratory.

Coefficient of correlation with ground truth
	Manual sonographer measurement of CIMT	Deep-learning method	Improvement of deep-learning method over sonographer measurement
Ground truth 1	0.80	Model 1: 0.96	20%
Ground truth 2	0.83	Model 2: 0.95	14.5%

The deep-learning method takes only a few milliseconds to perform, according to the researchers. They acknowledged, though, that the system relied on a dataset limited to a Japanese diabetic cohort, and it has not been tested on a wide variety of datasets.

As a result, it needs to be evaluated further in a multiethnic patient population with subclinical atherosclerosis and low-, moderate-, and high-risk scenarios, they wrote. The approach must also be evaluated on ultrasound images from different equipment vendors, and it should also be extended from a desktop PC-based application to a web-based version, according to the group.

KIDNEY STONE AND ULTRASOUND

Which modality works best for diagnosing kidney stones?

By Kate Madden Yee, AuntMinnie.com staff writer

May 24, 2018 -- Which modality works best for diagnosing kidney stone disease, also known as urolithiasis: digital tomosynthesis, ultrasound, or the current reference standard of multidetector CT (MDCT)? It depends, according to a study published online May 17 in the European Journal of Radiology.

Many imaging modalities can be used to diagnose the disease, wrote a team led by Dr. Manavjit Singh Sandhu of the Postgraduate Institute of Medical Education and Research (PGIMER) in Chandigarh, India. But it can be challenging to determine which one is best in a given clinical situation.

"With numerous technological advancements in the field of radiology, many imaging modalities can be employed for the diagnosis of urolithiasis and it becomes confusing and at times difficult to decide which one to choose and when," Sandhu and colleagues wrote. "Clinicians [must] be aware of the potential benefits and relative strength of each imaging modality [to balance its use with] healthcare costs, radiation burden, and contrast patient safety in a given clinical scenario."

Imaging is key

Urolithiasis affects a wide range of patients, and imaging is a key part of both diagnosing the condition and following patients after diagnosis, according to the group. MDCT is the current gold standard for detecting the disease, with a sensitivity of 97% and a specificity of 98%. But it also has the highest radiation dose among the modalities used for this purpose.

"CT cannot be used too frequently in patients with recurrent calculi, or in post-treatment patients on follow-up," Sandhu and colleagues noted.

As an alternative to MDCT, digital tomosynthesis overcomes limitations found in conventional tomography, imparting minimal radiation dose and removing overlying structures that can confuse diagnosis. And ultrasound offers benefits such as convenience, low cost, and lack of radiation. So which of these two modalities should clinicians use to diagnose kidney stone disease, and when?

Sandhu and colleagues compared the diagnostic performance of digital tomosynthesis with that of ultrasound, using MDCT as the reference standard. The study included 66 patients who were either suspected of having kidney stone disease or had a history of recurrent disease; of these, 41 had urolithiasis and 25 had nonrenal causes of abdominal pain.

All patients underwent digital tomosynthesis, ultrasound, and MDCT within a 24-hour period. Two radiologists categorized the calculi, or stones, according to location and size. Sandhu's group then examined the sensitivity, specificity, and positive and negative predictive values for tomosynthesis and ultrasound.

In the 41 patients with urolithiasis, MDCT found 121 stones (105 renal, 14 ureteric, and two vesical), most of which were smaller than 5 mm.

No. of calculi found with MDCT, digital tomosynthesis, and ultrasound
Category	MDCT	Digital tomosynthesis		Ultrasound
		Reader 1	Reader 2	Reader 1	Reader 2
Location
Kidney	105	51	47	56	51
Ureter	14	11	10	6	5
Urinary bladder	2	2	2	2	2
Size
< 5 mm	52	9	7	9	7
5-10 mm	32	28	25	22	19
> 10 mm	37	27	26	33	32

The average overall sensitivity of digital tomosynthesis for identifying kidney stone disease was 50% (p < 0.001), while the sensitivity of ultrasound was 50.4% (p = 0.005). As for identifying renal stones, digital tomosynthesis had a sensitivity of 47.1% and ultrasound a sensitivity of 50.9%; for ureteric stones, sensitivity was 74.9% with digital tomosynthesis and 39.2% with ultrasound.

"The disappointingly low sensitivity [of digital tomosynthesis] may be attributed to the fact that [most] of the calculi in our study group were smaller than 5 mm ... [which decreases] the overall sensitivity for digital tomosynthesis," the authors noted.

Digital tomosynthesis vs. ultrasound for kidney stone disease
Performance measure	Digital tomosynthesis	Ultrasound
Sensitivity	50%	50.4%
Specificity	89.8%	89.8%
Positive predictive value	96%	96%
Negative predictive value	26.5%	26.5%
p-value	< 0.001	< 0.005

Although the results did not show a statistically significant difference between digital tomosynthesis and ultrasound for diagnosing urolithiasis, the researchers did find that digital tomosynthesis performed better than ultrasound when it came to ureteric stones, suggesting that the modality may be preferred for the initial evaluation of these patients.

"Among the 14 ureteric calculi, the majority were in the mid ureter, which is technically the most difficult part of the ureter to examine on ultrasound as it is generally obscured by overlying bowel gas shadows," the group wrote. "The diagnostic accuracy of digital tomosynthesis in detecting ureteric calculi was relatively improved as it removes the overlying structures [and] enhances local tissue separation ... which allows for better visualization of calculi in the ureter."

Ultrasound, however, is effective in identifying hydroureteronephrosis, a condition in which the kidney and ureter swell because of obstructed urine flow. Because hydroureteronephrosis needs to be ruled out for a urolithiasis diagnosis, the researchers concluded that both modalities have a place in the radiology toolkit for diagnosing kidney stone disease.

"In this study, we found no statistically significant difference between the performance of ultrasound and digital tomosynthesis in diagnosis of urolithiasis," the authors concluded. "Digital tomosynthesis performed significantly better than ultrasound in detecting ureteric calculi ... and therefore may be preferred in this subset of patients ... [but] in clinical practice, ultrasound still would remain the preferred modality in the initial workup of patients, especially those presenting acutely."