scholarly journals Deep learning from “passive feeding” to “selective eating” of real-world data

2020 ◽  
Vol 3 (1) ◽  
Author(s):  
Zhongwen Li ◽  
Chong Guo ◽  
Danyao Nie ◽  
Duoru Lin ◽  
Yi Zhu ◽  
...  
Keyword(s):  
Deep Learning ◽  
Real World ◽  
Diagnostic System ◽  
Standard Of Care ◽  
Poor Quality ◽  
Diagnostic Systems ◽  
Real World Data ◽  
Clinical Images ◽  
World Data ◽  
Selective Eating

Abstract Artificial intelligence (AI) based on deep learning has shown excellent diagnostic performance in detecting various diseases with good-quality clinical images. Recently, AI diagnostic systems developed from ultra-widefield fundus (UWF) images have become popular standard-of-care tools in screening for ocular fundus diseases. However, in real-world settings, these systems must base their diagnoses on images with uncontrolled quality (“passive feeding”), leading to uncertainty about their performance. Here, using 40,562 UWF images, we develop a deep learning–based image filtering system (DLIFS) for detecting and filtering out poor-quality images in an automated fashion such that only good-quality images are transferred to the subsequent AI diagnostic system (“selective eating”). In three independent datasets from different clinical institutions, the DLIFS performed well with sensitivities of 96.9%, 95.6% and 96.6%, and specificities of 96.6%, 97.9% and 98.8%, respectively. Furthermore, we show that the application of our DLIFS significantly improves the performance of established AI diagnostic systems in real-world settings. Our work demonstrates that “selective eating” of real-world data is necessary and needs to be considered in the development of image-based AI systems.

P1-666: REAL WORLD DATA IN AUSTRALIA DEMONSTRATES POOR QUALITY OF LIFE IS LINKED TO DETERIORATION IN COGNITIVE FUNCTION

2006 ◽  
Vol 14 (7S_Part_11) ◽  
pp. P598-P598
Author(s):  
Eddie Jones ◽  
Rezaul Karim Khandker ◽  
Christopher M. Black ◽  
James Pike ◽  
Joseph Husbands ◽  
...  
Keyword(s):  
Quality Of Life ◽  
Cognitive Function ◽  
Real World ◽  
Poor Quality ◽  
Real World Data ◽  
World Data

scholarly journals From Reflex to Reflection: Two Tricks AI Could Learn from Us

Philosophies ◽  
2019 ◽  
Vol 4 (2) ◽  
pp. 27
Author(s):  
Jean-Louis Dessalles
Keyword(s):  
Machine Learning ◽  
Neural Networks ◽  
Deep Learning ◽  
Cognitive Processes ◽  
Real World ◽  
Machine Learning Techniques ◽  
Human Intervention ◽  
Real World Data ◽  
World Data ◽  
Learning Techniques

Deep learning and other similar machine learning techniques have a huge advantage over other AI methods: they do function when applied to real-world data, ideally from scratch, without human intervention. However, they have several shortcomings that mere quantitative progress is unlikely to overcome. The paper analyses these shortcomings as resulting from the type of compression achieved by these techniques, which is limited to statistical compression. Two directions for qualitative improvement, inspired by comparison with cognitive processes, are proposed here, in the form of two mechanisms: complexity drop and contrast. These mechanisms are supposed to operate dynamically and not through pre-processing as in neural networks. Their introduction may bring the functioning of AI away from mere reflex and closer to reflection.

scholarly journals L*a*b*Fruits: A Rapid and Robust Outdoor Fruit Detection System Combining Bio-Inspired Features with One-Stage Deep Learning Networks

Sensors ◽  
2020 ◽  
Vol 20 (1) ◽  
pp. 275 ◽  
Author(s):  
Raymond Kirk ◽  
Grzegorz Cielniak ◽  
Michael Mangan
Keyword(s):  
Deep Learning ◽  
Real World ◽  
Model Complexity ◽  
Detection Accuracy ◽  
Learning Networks ◽  
Real World Data ◽  
Data Set ◽  
World Data ◽  
One Stage ◽  
Controlled Conditions

Automation of agricultural processes requires systems that can accurately detect and classify produce in real industrial environments that include variation in fruit appearance due to illumination, occlusion, seasons, weather conditions, etc. In this paper we combine a visual processing approach inspired by colour-opponent theory in humans with recent advancements in one-stage deep learning networks to accurately, rapidly and robustly detect ripe soft fruits (strawberries) in real industrial settings and using standard (RGB) camera input. The resultant system was tested on an existent data-set captured in controlled conditions as well our new real-world data-set captured on a real strawberry farm over two months. We utilise F 1 score, the harmonic mean of precision and recall, to show our system matches the state-of-the-art detection accuracy ( F 1 : 0.793 vs. 0.799) in controlled conditions; has greater generalisation and robustness to variation of spatial parameters (camera viewpoint) in the real-world data-set ( F 1 : 0.744); and at a fraction of the computational cost allowing classification at almost 30fps. We propose that the L*a*b*Fruits system addresses some of the most pressing limitations of current fruit detection systems and is well-suited to application in areas such as yield forecasting and harvesting. Beyond the target application in agriculture this work also provides a proof-of-principle whereby increased performance is achieved through analysis of the domain data, capturing features at the input level rather than simply increasing model complexity.

scholarly journals Causal Deep Learning on Real-world Data Reveals the Comparative Effectiveness of Anti-hyperglycemic Treatments

2021 ◽  
Author(s):  
Chinmay Belthangady ◽  
Stefanos Giampanis ◽  
Will Stedden ◽  
Paula Alves ◽  
Stephanie Chong ◽  
...  
Keyword(s):  
Deep Learning ◽  
Real World ◽  
Comparative Effectiveness ◽  
Chronic Conditions ◽  
Treatment Guidelines ◽  
Treatment Options ◽  
Treatment Selection ◽  
Treatment Optimization ◽  
Real World Data ◽  
World Data

Abstract Type 2 Diabetes is associated with severe health outcomes, the effects of which are responsible for approximately 1/4 of total U.S. healthcare spending. Current treatment guidelines endorse a massive number of potential anti-hyperglycemic treatment options in various permutations and combinations. Personalized strategies for optimizing treatment selection are lacking. We analyzed real-world data from a nationwide population of over one million diabetics to evaluate the comparative effectiveness of more than 80 different treatment strategies ranging from monotherapy up to combinations of five concomitant classes of drugs across each of 10 clinical subgroups defined by age, insulin dependence, and number of other chronic conditions. Our causal deep learning approach developed on such data allows for more personalized recommendations of treatment selection. We observe significant differences in blood sugar reduction between patients receiving high vs low ranked treatment options and that less than 2% of the population is on a highly ranked treatment. This method can be extended to explore treatment optimization of other chronic conditions.

scholarly journals Causal Deep Learning on Real-world Data Reveals the Comparative Effectiveness of Anti-hyperglycemic Treatments

2021 ◽  
Author(s):  
Chinmay Belthangady ◽  
Stefanos Giampanis ◽  
Will Stedden ◽  
Paula Alves ◽  
Stephanie Chong ◽  
...  
Keyword(s):  
Deep Learning ◽  
Real World ◽  
Comparative Effectiveness ◽  
Chronic Conditions ◽  
Treatment Guidelines ◽  
Treatment Options ◽  
Treatment Selection ◽  
Treatment Optimization ◽  
Real World Data ◽  
World Data

Abstract Type 2 Diabetes is associated with severe health outcomes, the effects of which are responsible for approximately 1/4 of total U.S. healthcare spending. Current treatment guidelines endorse a massive number of potential anti-hyperglycemic treatment options in various permutations and combinations. Personalized strategies for optimizing treatment selection are lacking. Real-world data from a nationwide population of over one million diabetics was analyzed to evaluate the comparative effectiveness of more than 80 different treatment strategies ranging from monotherapy up to combinations of five concomitant classes of drugs across each of 10 clinical subgroups defined by age, insulin dependence, and number of other chronic conditions. A causal deep learning approach developed on such data allows for more personalized recommendations of treatment selection. Significant differences were observed in blood sugar reduction between patients receiving high vs low ranked treatment options and that less than 2% of the population is on a highly ranked treatment. This method can be extended to explore treatment optimization of other chronic conditions.

scholarly journals Real-World Data on Frontline (FL) Treatments in Multiple Myeloma (MM) Patients across EU5 Countries

Blood ◽  
2018 ◽  
Vol 132 (Supplement 1) ◽  
pp. 5858-5858
Author(s):  
Joana Anjo ◽  
Alex Rider ◽  
Abigail Bailey ◽  
Maren Gaudig
Keyword(s):  
Data Collection ◽  
Real World ◽  
Stem Cell Transplant ◽  
Standard Of Care ◽  
Cell Transplant ◽  
Summary Statistics ◽  
Real World Data ◽  
World Data ◽  
Induction Regimen ◽  
Number Of Cycles

Abstract Objective This analysis was conducted to understand the clinical practice on FL treatment in patients with MM across EU5 countries. Methods Real world data were collected through Adelphi's Disease-Specific ProgrammeTM - a point in time survey administered to physicians (n=241) in EU5 countries between Nov 2017 - Feb 2018. Stem Cell Transplant eligibility and treatments, including number of cycles and dosage, were collected from patient record forms (n=1952). Summary statistics were reported and analysed descriptively. Results Data on FL treatment was collected for 1952 patients; 988 (51%) were still on FL treatment at the time of data collection. Bortezomib-based regimens were used in more than 70% of patients - in both transplanted/transplant eligible (TE, n=572) and non-transplanted/transplant ineligible patients (TIE, n= 1380). In TIE patients, bortezomib, melphalan and dexamethasone (VMP) was the most commonly used regimen, covering almost one third of the patients (31%), followed by bortezomib, either in combination with cyclophosphamide and dexamethasone (VCD, 10%) or thalidomide and dexamethasone (VTD, 10%). The other two FL regimens currently approved in Europe - thalidomide, melphalan and dexamethasone (MPT) and lenalidomide and dexamethasone (Rd) have patient shares of 9% each. When analyzing the TIE patients undergoing FL treatment at time of data collection (n=606), VMP remained the most used regimen (29%) and Rd the second (15%), with VTD and MPT being used in 8% of patients. In TE patients, VTD was the most commonly used induction regimen, being used in 50% of patients, followed by VCD (21%). The numbers remained the same when analyzing the TE patients in FL treatment at time of data collection (n=382), with 54% and 22% using VTD and VCD, respectively. Conclusions Collectively these results indicate that bortezomib-based regimens remain the standard of care in FL treatment of MM in EU5, in both transplant and non-transplant settings. Disclosures Anjo: Janssen: Employment. Rider:Adelphi Real World: Employment. Bailey:Adelphi Real World: Employment. Gaudig:Janssen: Employment.

Using real world data to evaluate overall survival in PD1-treated vs. standard of care metastatic melanoma patients.

2018 ◽  
Vol 36 (15_suppl) ◽  
pp. e21606-e21606 ◽  
Author(s):  
Purvi Dev-Vartak ◽  
Xinyan Yu ◽  
Fa-Qiang Liu ◽  
Paul Cariola ◽  
Jeffrey Hodge ◽  
...  
Keyword(s):  
Overall Survival ◽  
Metastatic Melanoma ◽  
Real World ◽  
Standard Of Care ◽  
Real World Data ◽  
World Data ◽  
Melanoma Patients

Using real-world cohorts to assess the generalizability and relevance of randomized clinical trials (RCTs).

2019 ◽  
Vol 37 (15_suppl) ◽  
pp. 6540-6540 ◽  
Author(s):  
Caroline Savage Bennette ◽  
Nathan Coleman Nussbaum ◽  
Melissa D. Curtis ◽  
Neal J. Meropol
Keyword(s):  
Clinical Trials ◽  
Real World ◽  
Randomized Clinical Trials ◽  
Treatment Options ◽  
Metastatic Breast ◽  
Standard Of Care ◽  
Real World Data ◽  
Eligibility Criteria ◽  
Relevance Score ◽  
World Data

6540 Background: RCTs are the gold standard for understanding the efficacy of new treatments, however, patients (pts) in RCTs often differ from those treated in the real-world. Further, selecting a standard of care (SOC) arm is challenging as treatment options may evolve during the course of a RCT. Our objective was to assess the generalizability and relevance of RCTs supporting recent FDA approvals of anticancer therapies. Methods: RCTs were identified that supported FDA approvals of anticancer therapies (1/1/2016 - 4/30/2018). Relevant pts were selected from the Flatiron Health longitudinal, EHR-derived database, where available. Two metrics were calculated: 1) a trial’s pt generalizability score (% of real-world pts receiving treatment consistent with the control arm therapy for the relevant indication who actually met the trial's eligibility criteria) and 2) a trial’s SOC relevance score (% of real-world pts with the relevant indication and meeting the trial's eligibility criteria who actually received treatment consistent with the control arm therapy). All analyses excluded real-world pts treated after the relevant trial’s enrollment ended. Results: 14 RCTs across 5 cancer types (metastatic breast, advanced non-small cell lung cancer, metastatic renal cell carcinoma, multiple myeloma, and advanced urothelial) were included. There was wide variation in the SOC relevance and pt generalizability scores. The median pt generalizability score was 63% (range 35% - 88%), indicating that most real-world pts would have met the RCT eligibility criteria. The median SOC relevance score was 37% (range 15% - 74%), indicating that most RCT control arms did not reflect the way trial-eligible real-world pts in the US were actually treated. Conclusions: There is great variability across recent RCTs in terms of pt generalizability and relevance of SOC arms. Real-world data can be used to inform selection of control arms, predict impact of inclusion/exclusion criteria, and also assess the generalizability of the results of completed trials. Incorporating real-world data in planning and interpretation of prospective clinical trials could improve accrual and enhance relevance of RCT outcomes.

Sign in / Sign up

Export Citation Format

Share Document