To systematically examine the design, reporting standards, risk of bias, and claims of studies comparing the performance of diagnostic deep learning algorithms for medical imaging with that of expert ...clinicians.
Systematic review.
Medline, Embase, Cochrane Central Register of Controlled Trials, and the World Health Organization trial registry from 2010 to June 2019.
Randomised trial registrations and non-randomised studies comparing the performance of a deep learning algorithm in medical imaging with a contemporary group of one or more expert clinicians. Medical imaging has seen a growing interest in deep learning research. The main distinguishing feature of convolutional neural networks (CNNs) in deep learning is that when CNNs are fed with raw data, they develop their own representations needed for pattern recognition. The algorithm learns for itself the features of an image that are important for classification rather than being told by humans which features to use. The selected studies aimed to use medical imaging for predicting absolute risk of existing disease or classification into diagnostic groups (eg, disease or non-disease). For example, raw chest radiographs tagged with a label such as pneumothorax or no pneumothorax and the CNN learning which pixel patterns suggest pneumothorax.
Adherence to reporting standards was assessed by using CONSORT (consolidated standards of reporting trials) for randomised studies and TRIPOD (transparent reporting of a multivariable prediction model for individual prognosis or diagnosis) for non-randomised studies. Risk of bias was assessed by using the Cochrane risk of bias tool for randomised studies and PROBAST (prediction model risk of bias assessment tool) for non-randomised studies.
Only 10 records were found for deep learning randomised clinical trials, two of which have been published (with low risk of bias, except for lack of blinding, and high adherence to reporting standards) and eight are ongoing. Of 81 non-randomised clinical trials identified, only nine were prospective and just six were tested in a real world clinical setting. The median number of experts in the comparator group was only four (interquartile range 2-9). Full access to all datasets and code was severely limited (unavailable in 95% and 93% of studies, respectively). The overall risk of bias was high in 58 of 81 studies and adherence to reporting standards was suboptimal (<50% adherence for 12 of 29 TRIPOD items). 61 of 81 studies stated in their abstract that performance of artificial intelligence was at least comparable to (or better than) that of clinicians. Only 31 of 81 studies (38%) stated that further prospective studies or trials were required.
Few prospective deep learning studies and randomised trials exist in medical imaging. Most non-randomised trials are not prospective, are at high risk of bias, and deviate from existing reporting standards. Data and code availability are lacking in most studies, and human comparator groups are often small. Future studies should diminish risk of bias, enhance real world clinical relevance, improve reporting and transparency, and appropriately temper conclusions.
PROSPERO CRD42019123605.
AbstractObjectiveTo systematically examine the design, reporting standards, risk of bias, and claims of studies comparing the performance of diagnostic deep learning algorithms for medical imaging ...with that of expert clinicians.DesignSystematic review.Data sourcesMedline, Embase, Cochrane Central Register of Controlled Trials, and the World Health Organization trial registry from 2010 to June 2019.Eligibility criteria for selecting studiesRandomised trial registrations and non-randomised studies comparing the performance of a deep learning algorithm in medical imaging with a contemporary group of one or more expert clinicians. Medical imaging has seen a growing interest in deep learning research. The main distinguishing feature of convolutional neural networks (CNNs) in deep learning is that when CNNs are fed with raw data, they develop their own representations needed for pattern recognition. The algorithm learns for itself the features of an image that are important for classification rather than being told by humans which features to use. The selected studies aimed to use medical imaging for predicting absolute risk of existing disease or classification into diagnostic groups (eg, disease or non-disease). For example, raw chest radiographs tagged with a label such as pneumothorax or no pneumothorax and the CNN learning which pixel patterns suggest pneumothorax.Review methodsAdherence to reporting standards was assessed by using CONSORT (consolidated standards of reporting trials) for randomised studies and TRIPOD (transparent reporting of a multivariable prediction model for individual prognosis or diagnosis) for non-randomised studies. Risk of bias was assessed by using the Cochrane risk of bias tool for randomised studies and PROBAST (prediction model risk of bias assessment tool) for non-randomised studies.ResultsOnly 10 records were found for deep learning randomised clinical trials, two of which have been published (with low risk of bias, except for lack of blinding, and high adherence to reporting standards) and eight are ongoing. Of 81 non-randomised clinical trials identified, only nine were prospective and just six were tested in a real world clinical setting. The median number of experts in the comparator group was only four (interquartile range 2-9). Full access to all datasets and code was severely limited (unavailable in 95% and 93% of studies, respectively). The overall risk of bias was high in 58 of 81 studies and adherence to reporting standards was suboptimal (<50% adherence for 12 of 29 TRIPOD items). 61 of 81 studies stated in their abstract that performance of artificial intelligence was at least comparable to (or better than) that of clinicians. Only 31 of 81 studies (38%) stated that further prospective studies or trials were required.ConclusionsFew prospective deep learning studies and randomised trials exist in medical imaging. Most non-randomised trials are not prospective, are at high risk of bias, and deviate from existing reporting standards. Data and code availability are lacking in most studies, and human comparator groups are often small. Future studies should diminish risk of bias, enhance real world clinical relevance, improve reporting and transparency, and appropriately temper conclusions.Study registrationPROSPERO CRD42019123605.
Virus infection is sensed by pattern recognition receptors (PRRs) detecting virus nucleic acids and initiating an innate immune response. DNA-dependent protein kinase (DNA-PK) is a PRR that binds ...cytosolic DNA and is antagonized by vaccinia virus (VACV) protein C16. Here, VACV protein C4 is also shown to antagonize DNA-PK by binding to Ku and blocking Ku binding to DNA, leading to a reduced production of cytokines and chemokines in vivo and a diminished recruitment of inflammatory cells. C4 and C16 share redundancy in that a double deletion virus has reduced virulence not seen with single deletion viruses following intradermal infection. However, non-redundant functions exist because both single deletion viruses display attenuated virulence compared to wild-type VACV after intranasal infection. It is notable that VACV expresses two proteins to antagonize DNA-PK, but it is not known to target other DNA sensors, emphasizing the importance of this PRR in the response to infection in vivo.
Display omitted
•DNA-PK is a pattern recognition receptor that binds cytosolic DNA•Vaccinia virus proteins C4 and C16 antagonize DNA-PK by blocking DNA binding•C4 and C16 inhibit IRF3 signaling, cytokine production, and immune cell recruitment•C4 and C16 share redundant and non-redundant functions in vivo
DNA-PK is a pattern recognition receptor (PRR) that binds cytosolic DNA and stimulates IRF3 signaling. Scutts et al. show that vaccinia virus antagonizes this DNA sensor with two proteins, C4 and C16, which both block DNA binding.
Chronic kidney disease (CKD) describes a long-term decline in kidney function and has many causes. It affects hundreds of millions of people worldwide every year. It can have a strong negative impact ...on patients, especially when combined with cardiovascular disease (CVD): patients with both conditions have lower survival chances. In this context, computational intelligence applied to electronic health records can provide insights to physicians that can help them make better decisions about prognoses or therapies. In this study we applied machine learning to medical records of patients with CKD and CVD. First, we predicted if patients develop severe CKD, both including and excluding information about the year it occurred or date of the last visit. Our methods achieved top mean Matthews correlation coefficient (MCC) of +0.499 in the former case and a mean MCC of +0.469 in the latter case. Then, we performed a feature ranking analysis to understand which clinical factors are most important: age, eGFR, and creatinine when the temporal component is absent; hypertension, smoking, and diabetes when the year is present. We then compared our results with the current scientific literature, and discussed the different results obtained when the time feature is excluded or included. Our results show that our computational intelligence approach can provide insights about diagnosis and relative important of different clinical variables that otherwise would be impossible to observe.
Abstract
Introduction
The increasing longevity of the Western population means patients with a more advanced age are being diagnosed with resectable disease. With improvements in imaging and ...diagnostic capabilities, this trend is likely to develop further. As a unit operating on a higher proportion of older patients and with limited literature regarding the population of older than 85 years, we retrospectively compared the outcomes of patients older than 85 years in our unit treated with elective lung resection for non-small cell lung cancer (NSCLC) with those between the age of 80 and 84 years inclusive.
Methods
All patients who underwent elective lung cancer resection between the years 2012 and 2015 were identified from the National Thoracic Surgical Database.
Results
A total of 701 elective lung resections were performed during this time frame; 76 patients between the ages of 80 and 84 years and 18 patients older than 85 years. The follow-up period was 3 to 7 years. There was a significant increase in the Thoracic Surgery Scoring System (2.04; 2.96%,
p
= 0.0015) and a significant reduction in the transfer factor (94.7; 69.5%,
p
= 0.0001) between the younger and older groups. There were three (3.9%) in-hospital deaths in the 80 to 84 years age group and no in-hospital deaths in the 85 years and older age group.
Conclusion
This study demonstrates that surgery for early NSCLC can be safely performed in 85 years and older population. This is a higher risk population and parenchymal-sparing procedures should be considered.