#mdpidiagnostics Outcome Prediction at Patient Level Derived from Pre-Treatment 18F-#FDG #PET Due to Machine Learning in Metastatic #Melanoma Treated with Anti-#PD1 Treatment. View Full Text: mdpi.com/2075-4418/12/2…. #radiomics #MachineLearning #immunotherapy #prediction

Autopsy Validation of the Diagnostic Accuracy of 123I-MIBG Myocardial Scintigraphy for Lewy Body Disease n.neurology.org/content/early/… Neurology Journal Isotopiste Sophie Dautricourt

#FDG #PET is great for #epilepsy surgery already-but resolution and contrast could be better still to find more lesions e.g. #FocalCorticalDysplasia. In this new #Frontiers paper by @enthym, we took a step towards #superresolution PET… frontiersin.org/articles/10.33… 1/8

We created a large database of simulated #PET data using a high-quality Monte-Carlo PET simulator. With digital brain phantoms from T1w #MRI as “ground-truth phantom” inputs, we created 2100 simulated standard quality PETs (yes, lots of number crunching!) 2/8

Those 2100 standard-quality PETs were used to learn the mapping back to the ground-truth phantoms using #AI - a residual neuronal network, #ResNet. Once you have this mapping, you can create enhanced “ #superresolution ” PET from standard PET… 3/8

… and this is what those “predicted high-quality PETs” look like. Stats for various quality measures improved. 4/8

So far, so good, but what does this do for lesion detectability? To find out, we created another (!) 120 phantoms with lesions designed to look like #FCD (focal cortical dysplasia), see the example … 5/8

Those simulated #FCD lesions then had more favourable stats (better contrast) and were also twice as likely to be detected by human raters (75% vs 38%). Great! But these were all simulated data -so what about clinically acquired data? 6/8

So finally, as a proof of concept Isotopiste showed that a model trained on simulated can be applied to clinical data. Again, visual image quality improved (see images) and reader confidence increased, too. However, a limitation is that we only had 10 patients’ data. 7/8

We think this is quite promising. If you’re curious to read the whole story, here’s the link, open access of course: frontiersin.org/articles/10.33…) 8/8

Great collaboration with #Lyon’s #CERMEP and #CREATIS as usual, this time also with #CAHMS and Grenoble! KCL/GSTT PET centre Biomedical Engineering & Imaging Sciences Université Lyon 1 @enthym Marc Janier @colm_j_m

A new treatment for Alzheimer’s disease, lecanemab, was approved today by the FDA. It has been shown to slow disease progression for early-stage patients. The presence of beta amyloid must be confirmed before treatment. ow.ly/MTiy50Mkriw #NuclearMedicine #PETscan #Alzheimer

Depistage : ca travaille dur au HACKING HEALTH LYON !!!

EN DIRECT sur #Periscope : #Monpointdepistage.com entre en scène pscp.tv/w/awBucTFlZGpu…

#Monpointdepistage.com : pour toutes les personnes de 18 ans de faire le point sur leur #dépistage selon âge, sexe et comportement #HHLyon

Hello London. Great seminar about #CRPC #xofigo #radium223 and all therapeutic options Guy's Hospital RFC Bayer | Pharmaceuticals Ra-223 dichloride

Expert @morr316 of @sloan_kettering discusses radium-223 combination options for mCRPC. #pcsm ht.ly/hZla306N61m