ほー。ガンのCNN https://t.co/ihf0143G5a
RT @josephofiowa: The 2017 Nature paper on using machine learning to identify cancerous skin lesions from Stanford PhDs is now a high schoo…
RT @josephofiowa: The 2017 Nature paper on using machine learning to identify cancerous skin lesions from Stanford PhDs is now a high schoo…
The 2017 Nature paper on using machine learning to identify cancerous skin lesions from Stanford PhDs is now a high school hackathon project with @replit and @roboflow. Started: https://t.co/f6rS1S7ZRk Going: https://t.co/D3h6xwTiS2 https://t.co/iuZD29sLS
Una sola neuronal convolucional profunda, entrenada a partir de imágenes, utilizando solo píxeles y etiquetas de enfermedades como inputs, fue ser capaz de clasificar el cáncer de piel con un nivel de competencia de los dermatólogos certificados. https:/
Dermatologist-level classification of skin cancer with deep neural networks | Nature https://t.co/cK2yn2QBls
@Emaperidol Derm has started doing it. They even got a nature paper with it. Definitely not broad practice though. https://t.co/UdH0WQ01mk
Esteva, A., Kuprel, B., Novoa, R. A., Ko, J., Swetter, S. M., Blau, H. M., & Thrun, S. (2017). Dermatologist-level classification of skin cancer with deep neural networks. Nature, 542(7639), 115–118. https://t.co/U0vubtGDGv
A neural network can detect skin cancer as well as a trained dermatologist! #HealthCareAwarenessMonth #healthcareIT https://t.co/iQGUxCGteR
@tunguz But so called SOTA models can fail miserably when tested with real world data. There was a famous dermatology paper on lesion classification, it turns out the model is using rulers to make predictions: https://t.co/c3KCTUkfCJ
5/ ¿Es la certificación una garantía? No. Ha habido grandes fiascos en propuestas de análisis de imágenes médicas incluso "validadas" por grandes revistas: https://t.co/nCoIQZRGYW
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
@endrift @hoalycu @ShriramKMurthi A good analysis is in the 2019 thread from @EricTopol https://t.co/ZttCRw3wlY referencing Winkler, JAMA Dermatol. 2019. Also this Daily Beast story from 2017 https://t.co/nDyVEmWBQ6 referencing Novoa 2017.
Can #DeepLearning help dermatologists with the classification of skin cancer? In our latest Research Tutorial session we looked into the study by Esteva et al., which discusses an algorithm that can classify skin lesions from photographic images. 👉 http
@RealWellAI @Tennesseine Scary this research is pushing 5 years old now. https://t.co/risEeasrZz
@cospiratori 1) Ok. Famose a capi' https://t.co/N4INk4ARHy qui trovi un paper molto interessante sulla classificazione di immagini dermatalogiche attraverso un algoritmo di intelligenza artificiale.
Great initiative: #AI categorization of skin spot photos to help self-diagnosis. Google derm assist: https://t.co/BKJyijM4S2 Unclear how related it is to previous initiatives https://t.co/IWjdVHbIrr
RT @keefer007: Dermatologist-level classification of skin cancer with deep neural networks https://t.co/9e53FlpnM5 #ccme2021 #ai #artificia…
RT @keefer007: Dermatologist-level classification of skin cancer with deep neural networks https://t.co/9e53FlpnM5 #ccme2021 #ai #artificia…
RT @keefer007: Dermatologist-level classification of skin cancer with deep neural networks https://t.co/9e53FlpnM5 #ccme2021 #ai #artificia…
Dermatologist-level classification of skin cancer with deep neural networks https://t.co/9e53FlpnM5 #ccme2021 #ai #artificialintelligence
RT @YoshimotoSho: ディープラーニングが皮膚科医と同等のレベルで皮膚がんを分類できたとの報告。この技術によりスマホで写真を撮って、在宅で皮膚がんかどうかを見極めることができる可能性があるそうです。獣医療でも普及してほしいですが、129,450枚の画像で学習させて…
ディープラーニングが皮膚科医と同等のレベルで皮膚がんを分類できたとの報告。この技術によりスマホで写真を撮って、在宅で皮膚がんかどうかを見極めることができる可能性があるそうです。獣医療でも普及してほしいですが、129,450枚の画像で学習させてるみたい・・・。 https://t.co/ZeCL1jAPvf
@fatihdin4 Fakat ML alanında da overfitting hala bir sorun. Veriyi train/validation/test diye bölmek sorunu çözmüyor. Şöyle ünlü bir çalışma vardı, ortaya çıktı ki overfit etmişler: https://t.co/7jf1kyAwod
e.g. this popular study reported a CNN for skin lesion classification. It turns out that if there is a ruler in the image, the algorithm tends to call it cancer because on average, images in the dataset that have rulers are more likely to be malignant. h
@docprimum @Izno91 @eelte747 @topcao Pour les nævus, l’article compare dermatologues et IA avec comme étalon la biopsie (ce qui est bien). La question est de savoir s’il faut aller à la biopsie ou non. L’IA semble faire au moins aussi bien, peut-être un ch
現実的に、 皮膚がんではAiと皮膚科専門医と 同等の結果を報告されており、 スマートフォンのカメラとの連携で、 さらなる向上が期待できます。 https://t.co/9UI6hyrVmg
RT @edaaydinea: 𝐃𝐞𝐫𝐦𝐚𝐭𝐨𝐥𝐨𝐠𝐢𝐬𝐭-𝐥𝐞𝐯𝐞𝐥 𝐜𝐥𝐚𝐬𝐬𝐢𝐟𝐢𝐜𝐚𝐭𝐢𝐨𝐧 𝐨𝐟 𝐬𝐤𝐢𝐧 𝐜𝐚𝐧𝐜𝐞𝐫 𝐰𝐢𝐭𝐡 𝐝𝐞𝐞𝐩 𝐧𝐞𝐮𝐫𝐚𝐥 𝐧𝐞𝐭𝐰𝐨𝐫𝐤𝐬 This article is a very important source for the…
𝐃𝐞𝐫𝐦𝐚𝐭𝐨𝐥𝐨𝐠𝐢𝐬𝐭-𝐥𝐞𝐯𝐞𝐥 𝐜𝐥𝐚𝐬𝐬𝐢𝐟𝐢𝐜𝐚𝐭𝐢𝐨𝐧 𝐨𝐟 𝐬𝐤𝐢𝐧 𝐜𝐚𝐧𝐜𝐞𝐫 𝐰𝐢𝐭𝐡 𝐝𝐞𝐞𝐩 𝐧𝐞𝐮𝐫𝐚𝐥 𝐧𝐞𝐭𝐰𝐨𝐫𝐤𝐬 This article is a very important source for the skin cancer classification by deep neural network. https://t.co/OZTLrhZcbz #neuralnetworks
RT @rbukralia: Dermatologist-level classification of skin cancer with deep neural networks https://t.co/V2hl0iJ2Mx #AI #ML #DeepLearning #D…
Dermatologist-level classification of skin cancer with deep neural networks https://t.co/V2hl0iJ2Mx #AI #ML #DeepLearning #DataScience #cancer
The most famous example of this (I think, since I can't immediately find that paper she's referring to) is this study - where researchers thought they had trained ai to differentiate cancer from birthmarks. 1/2 https://t.co/zbd9PX6HI3
AI皮膚画像診断は一番入りやすい。スマホ診断できる。2017年Natureで、皮膚科医と同精度のAI検出という論文を見たとき、もっと早く実用化されるのかと思った ただ、どこで競争優位性を構築するかは難しい。お金あればデータ集めてアノテーションしてアプリは作れる https://t.co/u5gyVvhkQI
A few news articles about the impact of the ‘ruler’ in this study, but had anyone done a proper analysis? https://t.co/WVnQELZRPv
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @drdangayach: #CHEST2019 #CHESTchat T3: #AI could help us combat #implicitbias with diverse, #bigdata there is a pitfall bec these algor…
#CHEST2019 #CHESTchat T3: #AI could help us combat #implicitbias with diverse, #bigdata there is a pitfall bec these algorithms can be widely applicable without bias if they are trained on unbiased data. #dermatology &lesion classification https://t.co
RT @Age_Matters: Dermatologist-level classification of skin cancer with deep neural networks - game-changing paper quoted at #RCPIAC19 htt…
RT @Age_Matters: Dermatologist-level classification of skin cancer with deep neural networks - game-changing paper quoted at #RCPIAC19 htt…
Dermatologist-level classification of skin cancer with deep neural networks - game-changing paper quoted at #RCPIAC19 https://t.co/29y80jFJJD
@DrLukeOR Reminds me of the time a deep network learned that when a ruler is held next to the skin lesion on the photo, it's more likely malignant/a melanoma... https://t.co/0Y4V9qMnaw
RT @bedch: @cepcam 2017: https://t.co/tgLx6RglKF Les médecins ajoutent une règle dans l'image pour mesurer le cancer s'il y a un doute et/o…
RT @bedch: @cepcam 2017: https://t.co/tgLx6RglKF Les médecins ajoutent une règle dans l'image pour mesurer le cancer s'il y a un doute et/o…
@cepcam 2017: https://t.co/tgLx6RglKF Les médecins ajoutent une règle dans l'image pour mesurer le cancer s'il y a un doute et/ou quand c'est évident. Le DNN l'a appris. C'est subtil car il a aussi appris sur d'autres features. ~1000 citations avant que la
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @quocdngu: This is a whole new level of confounding: https://t.co/bFuX0HB0Do
RT @quocdngu: This is a whole new level of confounding: https://t.co/bFuX0HB0Do
RT @quocdngu: This is a whole new level of confounding: https://t.co/bFuX0HB0Do
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
This is a whole new level of confounding:
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
Looks like machine learning isn't the best way to rule out cancer--apparently rulers=melanomas
RT @EricTopol: @JAMADerm @UniHeidelberg @nature Other points noteworthy about this report 1. Not seen here, but previously there has been a…
RT @EricTopol: @JAMADerm @UniHeidelberg @nature Other points noteworthy about this report 1. Not seen here, but previously there has been a…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg @nature Other points noteworthy about this report 1. Not seen here, but previously there has been a…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
@JAMADerm @UniHeidelberg @nature Other points noteworthy about this report 1. Not seen here, but previously there has been a drop-off of accuracy for the prospective study cf the retrospective one. AUC at left @Nature paper, at right the new @JAMADerm stud
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…
RT @dermotor: This is lovely example of how computers can be incredibly intelligent and yet not have "common sense" https://t.co/UY0NXaVkR9
RT @EricTopol: @JAMADerm @UniHeidelberg The classic medical neural net fake out was the ruler in the @Nature 2017 skin cancer paper, which…