The system, developed by digital health startup Cognoa, uses information from questionnaires and videos to help pediatricians diagnose autism. It received marketing authorization from the FDA on Wednesday.

In a first, the Food and Drug Administration gave the green light to an algorithm designed to help clinicians diagnose autism in young children. Developed by Palo Alto-based startup Cognoa, the software uses questionnaires from parents, clinicians, and home videos to make a recommendation to assist pediatricians with diagnosis. 

The goal is to identify autism spectrum disorder (ASD) earlier. On average, most kids in the U.S. are diagnosed around age 4. 

“Many of these children are waiting for long periods of time before they get in (to a specialist),” Cognoa CMO Dr. Sharief Taraman, a pediatric neurologist, said in a Zoom interview. “This is a really big deal. We have not had a diagnostic of this kind getting market authorization.” 

Taraman said the software uses machine learning to identify “maximally predictive” features from the questionnaires and two short home videos 

Of course, asking people to provide videos of their kids is very personal. He said families have to give permission for videos to be reviewed by video analysts and the physicians involved in their care.

The FDA’s authorization was based on results from a prospective, double-blinded study that compared how well the software performed in helping diagnose autism compared to a panel of clinicians making a diagnosis based on DSM-5 criteria. Cognoa went through the FDA’s de novo pathway for low- or moderate-risk devices that don’t have a predicate. 

It was evaluated on 425 kids ages 18 months through five years, across 14 different sites. Taraman said the company also made a point to recruit a diverse group of patients for the trial, in terms of race, ethnicity, gender, education and socioeconomic status. Currently, girls and minorities are often diagnosed with ASD at a later age. 

According to the FDA, Cognoa’s test yielded a false positive result in 15 out of 303 kids in the trial without ASD. Meanwhile, it yielded a false negative in just one of the 122 kids with ASD. 

In cases where there wasn’t a clear diagnosis or a rule-out, the algorithm gave an indeterminate result. In total, it provided a diagnosis for about 32% of patients in the trial. 

Having the ability to give an indeterminate result was important, Taraman said, that way the algorithm wouldn’t yield too many false positives, or overlook kids who have other neurodevelopmental conditions that need to be addressed. 

“Technology’s always a tool. It should never be a replacement for a clinician,” he said. “The test is not meant to be a standalone.”

 Cognoa plans to begin marketing the software, called Canvas Dx, later this year. 

“Autism actually is a beautiful thing,” Taraman said. “Our goal is not to ‘turn off’ autism; our goal is to address challenges that come with autism.” 

The post FDA authorizes machine learning software to help diagnose autism appeared first on Artificial Intelligence.

Deep learning can segment white brain matter in PET/CT scans to help clinicians diagnose a variety of neurodegenerative disorders, according to a new study.

A team of South Korean researchers created a type of artificial intelligence known as a generative adversarial network using an existing model called pix2pix, they said Feb. 10 in the Journal of Digital Imaging. And when applied to 18F-FDG PET/CT and MRI scan data from hundreds of patients, it accurately segmented and assessed the volume of white matter in the brain.

Quantifying these changes may help radiologists identify and diagnose a number of different brain diseases, first author Kyeong Taek Oh, with Yonsei University College of Medicine, and colleagues wrote.

“The volume change of white matter has been reported in aging, psychosis and multiple sclerosis. Also, white matter changes were observed in patients with Alzheimer’s disease with extensive gray matter atrophy,” according to Taek et al.

“It (white matter volume change) provides the severity, extent, and location of disease which are important clues for the identification of subtypes, staging and prognostication of neurodegenerative diseases,” the authors added.

To reach their conclusions, the researchers included data from 192 patients who had both PET/CT and MRI scans performed. They used 154 individuals’ data for training, 19 for validation and the remainder for testing their platform. Five experts graded the network-produced images using a segmentation quality score.

Their adversarial network performed well, with 78% of segmentation results notching “adequate” scores.

With this tool, Taek and colleagues wrote, clinicians could not only diagnose the brain diseases mentioned earlier, but more importantly, identify white matter “hypersensitivities,” which have been associated with an increased risk for certain dementias as well as impaired cognitive abilities.

“The segmentation results of the proposed method showed excellent performance mimicking the ground truth images of MRI compared with several commonly used deep learning methods,” the group concluded. “Further studies are needed to elucidate the clinical implications of FDG PET/CT based white matter segmentation in brain research.”

The post Applying deep learning to PET/CT scans helps clinicians diagnose neurodegenerative disorders appeared first on Artificial Intelligence.

A team of researchers at the James A. Haley Veterans’ Hospital in Tampa, Florida, is revolutionizing the way cancer is documented by enlisting the help of a computer to diagnose the disease in one of the largest patient populations in the nation: veterans.

Sophisticated artificial intelligence is capable of drastically altering how cancer is diagnosed and treated by learning to distinguish imagery of tissue containing cancerous cells from pictures of healthy tissue, a recent study in the Federal Practitioner journal claims.

“Based on a set of images selected to represent a specific tissue or disease process, the computer can be trained to evaluate and recognize new and unique images from patients and render a diagnosis,” the study’s authors wrote.

To test machine learning software, researchers uploaded hundreds of microscopic images of commonly diagnosed forms of the disease, such as lung or colon cancer, along with pictures of non-cancerous cells. At the conclusion of the test, the software — both Google- and Apple-based versions were tested — not only distinguished cancerous cells from non-cancerous tissue with a success rate of better than 90 percent, but it also indicated the exact form of cancer it was analyzing.

The ability of user-friendly machine learning software to learn and efficiently perform traditional human tasks in less time will alleviate some of the demand on medical practitioners who are already being stretched thin, the authors claim.

By coupling AI with a growing list of telehealth options, specialists have the potential to reach patients from anywhere in the world. Greater accessibility would especially benefit the millions of patients in the VA’s healthcare system, many of whom live in remote, rural areas where specialists or facilities needed to treat unique diseases are scarce at best.

A collaborative doctor-AI system can also diminish patient wait times and effectively eliminate the time-consuming paperwork analysis that has always bogged down practitioners. What it won’t do, according to one of the study’s authors, is replace its Homo sapien counterparts.

“Our ultimate goal would be to create programs that can be rolled out in the entire VA system so that pathologists who are working solo, or maybe there are two pathologists in some small VAs, would have the benefit of having something that is helping them become more productive, help them prioritize the workload and improve quality,” Dr. Andrew Borkowski said in a VA release.

And while the hope of machine learning enthusiasts is to eventually apply AI-assisted healthcare on a global scale, early testing using the VA’s expansive patient base allows for the mining of data from a seemingly limitless source.

The myriad imagery generated from the nearly 50,000 cancer diagnoses of veterans each year, for example, will enable AI software to analyze more data, learn faster, and expand application to other demographics and diseases at a pace other healthcare systems cannot match.

All this is not to say there won’t be obstacles to overcome before AI can be considered entirely viable — ensuring the impeccable accuracy of its decision-making paramount among them.

In 2019, Google-run AI software was fed hundreds of images and tested to determine whether it could predict the early onset of a deadly kidney disease. Two of every three AI-generated results yielded false positives. Significant diagnostic errors like that can be detrimental to practitioners who then follow up on phantom diseases using valuable time that could be spent treating patients in dire need, Mildred Cho, associate director of the Stanford Center for Biomedical Ethics, told WUSF News.

Continued success in machine learning trials like the one at the James A. Haley Veterans’ Hospital, however, bode well for AI’s future implementation into healthcare. Researchers hope continuously evolving software, such as Apple-produced AI that is now capable of recognizing images that have been rotated, flipped, or cropped, will help alleviate a glaring industry-wide trend.

The “number of pathologists in the U.S. is dramatically decreasing, and many other countries have marked physician shortages, especially in fields of specialized training such as pathology,” the study’s authors wrote. “These models could readily assist physicians in underserved countries and impact shortages of pathologists elsewhere by providing more specific diagnoses in an expedited manner.”

Future application of machine learning AI, the study concluded, will be immeasurably beneficial in diagnosing and documenting everything from various forms of cancer to non-cancerous diseases, brain hemorrhages, blood disorders, infections, and inflammatory issues.

“The potential of these technologies to improve health care delivery to veteran patients seems to be limited only by the imagination of the user.”

The post VA doctors are using artificial intelligence to diagnose cancer appeared first on Artificial Intelligence.

Data science and artificial intelligence are no longer buzz words in the biomedical research community. Physicians and other caregivers are increasingly being encouraged by hospitals and health insurance companies to utilize low-resolution dense biometric data captured using wearable medical devices. However classical healthcare heavily relies on high accuracy sparse datasets, i.e. patients are expected to get a thorough medical checkup once in a while, as opposed to continuous monitoring of a handful of vital parameters. The most significant impact of data science will be in helping physicians extract clinically relevant information from such dense low-quality data sets.

Radiologists and cardiologists are increasingly relying on automated high dimensional image processing algorithms to detect the likelihood of coronary artery disease from non-contrast chest CT scans. Similarly, radiologists and pulmonologists are using similar artificial intelligence based technology to identify clinically relevant structural and functional parameters of the lungs from chest CT scans. Understanding the basics of artificial intelligence will empower physicians to go beyond using these tools as black-boxes and deliver maximum impact for care pathways.

In this article, we have listed five such reasons why physicians and caregivers should learn about emerging technology such as data science and artificial intelligence

1. Diagnose using large volumes of data generated from continuous monitoring

With the advent of wearable medical device companies such as CloudDX, Snap40 and QuasaR clinicians can now look at continuous daily biometric data collected over months. Both primary and advanced data science techniques can be used to derive medically relevant outcomes from these dense data. Basic descriptive statistical results like the average resting heart rate could give you a quick understanding of the overall cardiac health of the patient. More advanced indicators such as stress index or LF/HF ratio of RR distance could be used to predict chances of heart arrhythmias more accurately. Data science will allow physicians to analyze these data sets both at local (days or weeks) and global (months or years) timescales, using a combination of both early warning scores and visual inspection of the data.

The post 5 Reasons Why Doctors Should Learn Data Science appeared first on Artificial Intelligence.