RT @EmmesCRO: In March 2019, @JAMANetworkOpen published “Comparison of Artificial Intelligence Techniques to Evaluate Performance of a Clas…
In March 2019, @JAMANetworkOpen published “Comparison of Artificial Intelligence Techniques to Evaluate Performance of a Classifier for Automatic Grading of Prostate Cancer From Digitized Histopathologic Images” coauthored by Emmes Canada's Darby Thompson:
"This study suggests that patch-based training and evaluation of machine learning models may be flawed and should be avoided, and multiexpert data should be used to obtain a more realistic evaluation of model performance" #healthcare #MachineLearning #ML #
This study shows the importance of cross-validating your AI algorithm against multiple experts. https://t.co/FftCIumO5H #ai #artificialintelligence #fda #medicaldevices #samd
RT @RogerBohn: A careful study showing that “AI for X is better than human doctors” is probably wrong. It’s easy to set up experiments to…
A careful study showing that “AI for X is better than human doctors” is probably wrong. It’s easy to set up experiments to make #ai look good. Much harder for AI to work better in real settings. X = prostate cancer in this case. https://t.co/vvT4r7pdOr
Comparison of Artificial Intelligence Techniques to Evaluate Performance of a Classifier for Automatic Grading of Prostate Cancer From Digitized Histopathologic Images https://t.co/dGfX9sGixy
Comparison of Artificial Intelligence Techniques to Evaluate Performance of a Classifier for Automatic Grading of Prostate Cancer From Digitized Histopathologic Images. | Oncology | JAMA Network Open | JAMA Network https://t.co/N4qJuvgfFX
RT @JAMANetworkOpen: Quality improvement study comparing several AI cross-validation approaches for grading prostate cancer suggests patien…
Quality improvement study comparing several AI cross-validation approaches for grading prostate cancer suggests patient-based cross-validation and multiexpert data are still needed https://t.co/iI33tFaChZ
RT @VanProstateCtr: A comparison of AI techniques to evaluate automatic grading of prostate cancer from digitized histopathologic images, a…
RT @VanProstateCtr: A comparison of AI techniques to evaluate automatic grading of prostate cancer from digitized histopathologic images, a…
RT @VanProstateCtr: A comparison of AI techniques to evaluate automatic grading of prostate cancer from digitized histopathologic images, a…
RT @VanProstateCtr: A comparison of AI techniques to evaluate automatic grading of prostate cancer from digitized histopathologic images, a…
RT @VanProstateCtr: A comparison of AI techniques to evaluate automatic grading of prostate cancer from digitized histopathologic images, a…
Researchers @VanProstateCtr show that patch-based training and evaluation could lead to significant overestimation of a model’s predictive accuracy in medical image analysis: https://t.co/077wkicbDP
A comparison of AI techniques to evaluate automatic grading of prostate cancer from digitized histopathologic images, a multi-site study by researchers including Drs @larrygoldenberg and @pcvblack ,was published in JAMA Network Open https://t.co/hfPQ4BYyq
Study using #machinelearning for automated grading of #prostatecancer highlights potential pitfalls of common model training and performance evaluation methods such as those based on single-expert input https://t.co/vKIdac3wzz
Quality improvement study comparing several AI cross-validation approaches for grading #prostatecancer suggests patient-based cross-validation and multiexpert data are still needed https://t.co/PMHDiC3gus @larrygoldenberg https://t.co/yx0F48ui4v