We need new tools to help us compare two treatment methods, consequently improve our decision making in orthodontic treatment of class III malocclusion, which are the most difficult to treat non surgically.
Amine Fattal
If i undersand the question right: you have preoperative 3D CBCT datasets and the 2 cohorts after treatment with (obviously) known postoperative result (with a scale like good -- > poor) and you would like to evaluate what are the image features that are associated with good outcome for each treatment, in other words wich treatment to choose by wich patient based on the image?
That is exactly what i want to accomplish, what are the tools available for this?
thanks for your reply!
Amine Fattal
As far as i can contribute, in general:
you should extract the image features (for example with pyradiomics), and select the stabile features (through classification), perform a clustering to find distinct clusters, then compare these with the therapy status (good or bad) to find associations with the extracted features.
Here is an article about the pyradiomics test:
Article Computational Radiomics System to Decode the Radiographic Phenotype
and Machine Learning methods for Quantitative Radiomic Biomarker
https://www.nature.com/articles/srep13087
Butprobably the most important thing to do is to preprocess your CT data. I would try resampling to small volumes or better perform segmentation of the mandible and do the above steps on that segmented data. (manual segmentation for example here: https://www.youtube.com/watch?v=P44m3MZuv5A).
You may take another route and follow this guide:
https://towardsdatascience.com/deep-learning-with-magnetic-resonance-and-computed-tomography-images
The answer to that question is not as simple. The difference between the use of deep learning in medical field and orthodontics is that the abnormalities are not as different. For example, even in the paper mentioned in the comment above, the differences between lungs with cancer and without cancer were identified which can be done with volumetric analysis and so on. However, in orthodontic treatment of class III, you would have to determine whether the mandibular angle was different in the groups, whether incisor proclined more, or was the overall volume of mandible was different and you may not get a clear difference in the two groups due to the variation in treatment protocols and results. Unless, you wish to compare cohort of different skeletal malocclusion pre-treatment where you can identify the patterns more easily.
- Badges
- Science topic
- Similar topics
- Biological Science
- Histology