Everett Schwieg

[Everett Schwieg @everett_schwieg]

The Swin UNETR model was able to outperform all other models in the MSD and BTCV datasets for the segmentation task and is considered SOTA.
#SMILEJournalClub

[Everett Schwieg @everett_schwieg]

In this paper by Tang et al, a new 3D vision transformer model was created for use with medical images. The model dubbed the Swin UNET Transformers, was able to be pre-trained via CT images from a variety of human organs.
https://t.co/6rlbp8wJ73

[Everett Schwieg @everett_schwieg]

#SMILEJournalClub

[Everett Schwieg @everett_schwieg]

In this paper by Athey and Wager, the results by running causal forest models on the NSLM dataset were analyzed. It was determined that by using this method, the problems with outcome clustering and unknown treatment propensities were able to be addressed.

[Everett Schwieg @everett_schwieg]

However, there were significant differences in the outcomes depending on how much emphasis was put into capturing the clustering phenomena between students of different schools. The leads to a potential future discussion on how to address clustered data.
https://t.co/EPW6u6KMT5

[Everett Schwieg @everett_schwieg]

Their method of using video recordings of the infants' movements as the data was interesting and tests how far DL models can be pushed. While static images can provide models with a wealth of information, videos add a whole new dimension of data analysis.

[Everett Schwieg @everett_schwieg]

In this paper by Groos et al, a deep learning model was used to classify the presence of cerebral palsy in high risk infants.
https://t.co/8A1XGGAIhl

[Everett Schwieg @everett_schwieg]

#SMILEJournalClub
The creation of an ImageNet-like database but for radiological studies is an impressive feat. Using this data for pre-training instead of the typical ImageNet dataset is a large step forward for interpretability and performance in CNNs.
https://t.co/DBnefyk1Q6

[Everett Schwieg @everett_schwieg]

#SMILEJournalClub
The ability to discern a brain pattern associated with one's polygenic risk score for schizophrenia at an 83% accuracy is astounding. The association of reduced gray matter and decreased functional activity in the FTC is also a great find
https://t.co/fotDBohVZA

[Everett Schwieg @everett_schwieg]

Processing these large datasets is difficult, especially with the varying quality of fundus images. In addition to the dataset assessment, the real-time assessment with DeepDR can be beneficial for the acquisition of higher-quality and more "diagnosable" images.
[2/2]

[Everett Schwieg @everett_schwieg]

#SMILEJournalClub
While the diabetic retinopathy classification results of their developed DeepDR system are impressive, the use of their image quality assessment subnetwork to evaluate over 600k images is notable.
https://t.co/F7vd9brhYk
[1/2]

[Everett Schwieg @everett_schwieg]

Furthering to improve the image modalities used in the medical industry is a clear step towards more accurate diagnoses and less repeat scans.
https://t.co/8MX6nn7VMH

[Everett Schwieg @everett_schwieg]

With newly-developed photon-counting CT imaging systems, image quality for coronary CT angiography (CCTA) can be improved. In photon-counting CT machines, the number and total energy of photon released are tracked, enabling enhanced resolution and image contrast
@anahitaheshmat

[Everett Schwieg @everett_schwieg]

While photo-counting CT hasn't been utilized for CCTA yet, research currently indicates it has the ability to improve image quality by 57% based on a small study of 14 patients with coronary artery disease.

[Everett Schwieg @everett_schwieg]

[3/3] As RFA is a simple treatment which can be performed within 90 minutes in a doctor's office, this may be able to replace more invasive and costly surgeries.
https://t.co/0MMAxLRQCx

[Everett Schwieg @everett_schwieg]

[1/3] Radiofrequency ablation (RFA) deemed a safe and effective procedure for treating small thyroid tumors. RFA is a fluoroscopically-guided procedure which a needle is inserted into the patient and uses heat to ablate tissue, typically an overactive nerve
@anahitaheshmat

[Everett Schwieg @everett_schwieg]

[2/3] In a study, a total of 1770 patients underwent RFA to treat Papillary thyroid carcinoma, the most common type of thyroid cancer. In 79% of cases, complete disappearance of the tumors was observed with only 3 cases of major complications.

[Everett Schwieg @everett_schwieg]

[3/3] Overall, the use of F-18 SMBT-1 as an indication of future neurodegenerative disease is a possibility which can help doctors make more accurate predictions. This tracer can also help researchers better understand the pathophysiology of neurodegenerative disorders.

[Everett Schwieg @everett_schwieg]

https://t.co/sujwxE54YW
[1/3] A newly developed PET radiotracer has been shown to reveal brain activity which may indicate early signs of Alzheimer's. The tracer named F-18 SMBT-1 reveals the activity of reactive astrocytes, believed to indicate cognitive decline @anahitaheshmat

[Everett Schwieg @everett_schwieg]

[2/3] Patients who were beta amyloid-positive and had Alzheimer's showed significantly higher F-18 SMBT-1 binding on reactive astrocytes. Those who were positive and cognitively normal had higher levels than those that were negative and normal.