Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

The article itself did not identify what the five patterns were, so that leads me to believe that they are vector-space patterns that aren’t related to any type of existing set of discrete inputs, despite them hinting at alcohol and smoking as being factors.

Does anybody have a sense of what these five models are?

The research itself is not open access so I wasn’t able to actually read it.



I got curious and accessed the full paper. The five patterns (so called "R-indices") are:

R1. Subcortical atrophy • Stress-related gene set • Pregnancy

R2: MTL atrophy • Dementia • CN-MCI-dementia progression • Amyloid and tau • Cognitive dysfunction, mainly memory impairment • Birth weight

R3: Parieto-temporal atrophy • Dementia; schizophrenia;Parkinson’s; multiple sclerosis • MCI-dementia progression • Amyloid and tau • Cognitive dysfunction, mainlyin executive function • Pregnancy • Social/recreational activity

R4: Diffuse cortical atrophy • Multiple sclerosis • Smoking and alcohol consumption • Diet

R5: Perisylvian atrophy • Multi-organ chronic conditions • Psychological factors • Psychiatric diseases • Cardiovascular factors • WMH• Mortality risk • Smoking and alcohol consumption

These patterns were identified using a type of Deep Learning model, the paper then goes on studying the association between different factors (such as chronic diseases) and brain atrophy.

So in conclusion, this study have demonstrated how brain atrophy manifests itself in MRI brain scans and how we can "classify" them into one of five categories (or "patterns", if you like). The gain here is in diagnosis, imagine if a doctor could easily determine that a specific person exhibits " Type 2" atrophy? Much time, money and human suffering could be saved providing the patient with tailored treatment at the get-go.


Thank you! Exceptionally clarifying




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: