Theo Dapamede

🚀 @TheoDapamede and I are thrilled to announce #RadPrompter - a Python package for simplified, reproducible, and composable #LLM prompting, providing a flexible framework for crafting complex prompts from reusable components. 1/7

@khosravi_bardia @YaleRadiology @YaleIMed @ArdestaniBahar @YaleRadRes @MayoMrfa @StoriesImg @Inside_TheMatch Congratulations, Bardia!

Dreams coming true! 🎉 Radiology @YaleRadiology 🫱🏻‍🫲🏼 Internal Medicine @YaleIMed Thank you for being my 🪨, @ArdestaniBahar. (1/3) #MatchDay2024 #MedTwitter #RadRes #CouplesMatch #IMRes #Match2024

@ShahriarFaghani @PennRadiology @PennRadRes Congratulations, Shahriar!

I am thrilled to join @PennRadiology family! My sincere gratitude to @PennRadRes for their trust in me! I can’t wait to learn from and work with the very best. #Match2024 #RadRes

Similar to the reporting using the TRIPOD statement, journals could ensure authors follow a similar standard reporting framework, for example: bmjopen.bmj.com/content/11/7/e… #RadAIchat

Training set is used to train the model's weights and biases. Validation is used to perform hyperparameter tuning or to select the best model and the test sets are used to evaluate the model's performance. The test sets may be from internal or external datasets. #RadAIchat

@ShahriarFaghani Could the validation set also be used for model calibration? #RadAIchat

The validation set is used to fine-tune the model's hyperparameters (which is why "tuning set" might be a more appropriate term). Finally, after the training and tuning process, the test set is used to assess the final model's performance on unseen data. #RadAIchat

The training set is used to train the model, meaning that the data from the training set is fed into the model to adjust its parameters (e.g., weights in neural networks) based on the input features and the target output. #RadAIchat

Ground truth labels can be used as benchmark for the performance of the model. This paper by @ImonBanerjee6 shows examples of self-supervised learning tasks evaluated on different ground truths based on their specific downstream task #RadAIchat medrxiv.org/content/10.110…

@Radiology_AI For tasks such as predicting labels or segmentation radiologists or other domain experts would be annotating the datasets. For ground truths such as denoising tasks, a higher dose image are sometimes used as the ground truths. #RadAIchat

T2. How are ground truth datasets created in #radiology #AI? #RadAIchat

@khosravi_bardia @Radiology_AI Ground truths are sometimes subjective or relative, depending on the objective or the AI task. #RadAIChat

@Radiology_AI “Ground truth” refers to the real, factual data used to train and evaluate AI models, in other words we use that as the gold standard. As the saying goes, 'garbage in, garbage out' - poor ground truth leads to poor AI model performance. #RadAIchat

Let's get started! T1. What does ground truth mean? Why is it essential in #AI? #RadAIchat

@Radiology_AI @khosravi_bardia @heacockmd @ShahriarFaghani @KMagudia @RSNA @cekahn @HElhalawaniMD Good evening all! I'm Theo, a postdoc fellow at @EmoryRadiology in @judywawira and @HariTrivediMD's awesome #HITILab. I have experience in developing AI models for spectral photon counting CTs and currently working on AI for opportunistic screening studies. #RadAIchat

Welcome to @Radiology_AI chat on "Defining the ground truth for #AI" moderated by @khosravi_bardia and panelists @heacockmd @ShahriarFaghani @KMagudia @theodapamede #AI #ML #DL #radres @RSNA @cekahn @HElhalawaniMD Please introduce yourself and always remember to add #RadAIchat

Don't miss the March @Radiology_AI #RadAIchat on Mar. 6 at 8 PM ET on "Defining the ground truth for #AI" moderated by @khosravi_bardia and panelists @heacockmd @ShahriarFaghani @KMagudia @theodapamede #radiomics #AI #ML #DL #radres @RSNA @cekahn @HElhalawaniMD

Just received my Science Teaching badge from Emory Continuing Education. credly.com/badges/1159dab…

Thrilled to share our preprint on use of #SyntheticData in medical imaging research! 📝 arxiv.org/abs/2311.09402 💬 Here is a custom GPT that will answer all your questions about the paper: chat.openai.com/g/g-iCGh2R2QA-… Some key takeaways in a 🧵 /1

<=> IFF If and only if #math #life

What noise annoys an oyster? An annoying noise annoys an oyster. #tonguetwister

There's always a first time for everything.

Peek a boo

Level 4 lock down across the country. Back to last year...

@heyyyairsss All the best!

Wish me luck for tomorrow's CBT yeorobun 😷

PI and multiple authors writing different sections of the same paper. #AcademicTwitter @OpenAcademics @PhDVoice @PhD_Genie