Johannes Lotz
@johannes.lotz.dev
Scientist in Computational Pathology @FraunhoferMEVIS / @uniluebeck.bsky.social, Vorstand im KIKS e.V. in Lübeck.
fedi: @johanneslotz@scholar.social, joha1o@chaos.social
#pathology #computationalpathology
fedi: @johanneslotz@scholar.social, joha1o@chaos.social
#pathology #computationalpathology
Had fun presenting with Carol Geppert at #ecdp2025.
Our poster “Image registration accelerates (pathology) diagnostics” shows that #imageregistration saves one minute (33%) per case in our pilot study!
Could not make it? Have a look the poster s.x91.de/ecdpposter and video s.x91.de/tnvideo
Our poster “Image registration accelerates (pathology) diagnostics” shows that #imageregistration saves one minute (33%) per case in our pilot study!
Could not make it? Have a look the poster s.x91.de/ecdpposter and video s.x91.de/tnvideo
June 27, 2025 at 9:41 AM
Had fun presenting with Carol Geppert at #ecdp2025.
Our poster “Image registration accelerates (pathology) diagnostics” shows that #imageregistration saves one minute (33%) per case in our pilot study!
Could not make it? Have a look the poster s.x91.de/ecdpposter and video s.x91.de/tnvideo
Our poster “Image registration accelerates (pathology) diagnostics” shows that #imageregistration saves one minute (33%) per case in our pilot study!
Could not make it? Have a look the poster s.x91.de/ecdpposter and video s.x91.de/tnvideo
“Tissue Concepts” is a resource-efficient foundation model for pathology. We trained with only 6% of the data compared to other (self-supervised) models, and it maintained great performance.
It outperforms self-supervised learning when the model is of similar size (we compared to ...
🧵2/5
It outperforms self-supervised learning when the model is of similar size (we compared to ...
🧵2/5
January 10, 2025 at 11:43 AM
“Tissue Concepts” is a resource-efficient foundation model for pathology. We trained with only 6% of the data compared to other (self-supervised) models, and it maintained great performance.
It outperforms self-supervised learning when the model is of similar size (we compared to ...
🧵2/5
It outperforms self-supervised learning when the model is of similar size (we compared to ...
🧵2/5
Our paper on Tissue Concepts is finally out in JCompBiomed 🥳: Supervised pre-training of foundation models needs fewer data and less compute/energy to analyze pathology whole-slide images.
How? Read more here: doi.org/10.1016/j.co...
🧵1/5
How? Read more here: doi.org/10.1016/j.co...
🧵1/5
January 10, 2025 at 11:43 AM
Our paper on Tissue Concepts is finally out in JCompBiomed 🥳: Supervised pre-training of foundation models needs fewer data and less compute/energy to analyze pathology whole-slide images.
How? Read more here: doi.org/10.1016/j.co...
🧵1/5
How? Read more here: doi.org/10.1016/j.co...
🧵1/5
Meanwhile in Munich: public transport (S-Bahn) is ready for todays start of the winter season.
November 20, 2024 at 10:47 PM
Meanwhile in Munich: public transport (S-Bahn) is ready for todays start of the winter season.
on diverse medical imaging tasks from #radiology and #pathology.
Key findings:
- UMedPT outperforms ImageNet pretraining & state-of-the-art models 💪
- Maintains performance with only 1% of training data for in-domain tasks 📈
- Enables strong results with 50% less data for out-of-domain tasks 🔍
…
Key findings:
- UMedPT outperforms ImageNet pretraining & state-of-the-art models 💪
- Maintains performance with only 1% of training data for in-domain tasks 📈
- Enables strong results with 50% less data for out-of-domain tasks 🔍
…
November 14, 2024 at 9:33 PM
on diverse medical imaging tasks from #radiology and #pathology.
Key findings:
- UMedPT outperforms ImageNet pretraining & state-of-the-art models 💪
- Maintains performance with only 1% of training data for in-domain tasks 📈
- Enables strong results with 50% less data for out-of-domain tasks 🔍
…
Key findings:
- UMedPT outperforms ImageNet pretraining & state-of-the-art models 💪
- Maintains performance with only 1% of training data for in-domain tasks 📈
- Enables strong results with 50% less data for out-of-domain tasks 🔍
…
Our medical image foundation model is recently published in Nature Computational Science: "Overcoming data scarcity in biomedical imaging with a foundational multi-task model" 🎉.
We developed a *supervised* foundational training strategy and pretrained a model (UMedPT) …
We developed a *supervised* foundational training strategy and pretrained a model (UMedPT) …
November 14, 2024 at 9:33 PM
Our medical image foundation model is recently published in Nature Computational Science: "Overcoming data scarcity in biomedical imaging with a foundational multi-task model" 🎉.
We developed a *supervised* foundational training strategy and pretrained a model (UMedPT) …
We developed a *supervised* foundational training strategy and pretrained a model (UMedPT) …
Ready to hit the road with an #EV in #winter? But where does all the energy go?
Instead of a poll, here is a plot of the consumption of lights (1), rear window (2), heated seats (3), steering wheel (4), each at <0.5 kW.
Then the air heater (5) comes in with 9 kW peak (will be lower over time) (1/2)
Instead of a poll, here is a plot of the consumption of lights (1), rear window (2), heated seats (3), steering wheel (4), each at <0.5 kW.
Then the air heater (5) comes in with 9 kW peak (will be lower over time) (1/2)
December 26, 2023 at 2:06 PM
One of the unique features of the Castro podcast app was/is the extending sleep timer (if you press play, it starts the timer again). With Castro being unavailable for now, here is a workaround based on an iOS Shortcut and Overcast (by @marcoarment.bsky.social)
outline.lotz.dev/s/64ebe0b5-d...
outline.lotz.dev/s/64ebe0b5-d...
November 29, 2023 at 1:38 PM
One of the unique features of the Castro podcast app was/is the extending sleep timer (if you press play, it starts the timer again). With Castro being unavailable for now, here is a workaround based on an iOS Shortcut and Overcast (by @marcoarment.bsky.social)
outline.lotz.dev/s/64ebe0b5-d...
outline.lotz.dev/s/64ebe0b5-d...