Adaptable Landmark Localisation: Applying Model Transfer Learning to a Shape Model Matching System
We address the challenge of model transfer learning for a shape model matching (SMM) system. The goal is to adapt an existing SMM system to work effectively with new data without rebuilding the system from scratch.
Recently, several SMM systems have been proposed that combine the outcome of a Random Forest (RF) regression step with shape constraints. These methods have been shown to lead to accurate and robust results when applied to the localisation of landmarks annotating skeletal structures in radiographs. However, as these methods contain a supervised learning component, their performance heavily depends on the data that was used to train the system, limiting their applicability to a new dataset with different properties.
Here we show how to tune an existing SMM system by both updating the RFs with new samples and re-estimating the shape model. We demonstrate the effectiveness of tuning a cephalometric SMM system to replicate the annotation style of a new observer.
Our results demonstrate that tuning an existing system leads to significant improvements in performance on new data, up to the extent of performing a well as a system that was fully rebuilt using samples from the new dataset. The proposed approach is fast and does not require access to the original training data.
KeywordsModel transfer learning Random Forests Landmark localisation Statistical shape models Machine learning Model tuning
C. Lindner is funded by the Engineering and Physical Sciences Research Council, UK (EP/M012611/1).
- 3.Donner, R., Menze, B.H., Bischof, H., Langs, G.: Fast anatomical structure localization using top-down image patch regression. In: Menze, B.H., Langs, G., Lu, L., Montillo, A., Tu, Z., Criminisi, A. (eds.) MCV 2012. LNCS, vol. 7766, pp. 133–141. Springer, Heidelberg (2013). doi: 10.1007/978-3-642-36620-8_14 CrossRefGoogle Scholar
- 4.Gao, Y., Shao, Y., Lian, J., Wang, A., Chen, R., Shen, D.: Accurate segmentation of CT male pelvic organs via regression-based deformable models and multi-task random forests. IEEE TMI 35(6), 1532–1543 (2016)Google Scholar
- 8.Lindner, C., Thiagarajah, S., Wilkinson, M., The arcOGEN Consortium, Wallis, G., Cootes, T.: Fully automatic segmentation of the proximal femur using random forest regression voting. IEEE TMI 32(8), 1462–1472 (2013)Google Scholar
- 11.Roberts, M.G., Cootes, T.F., Adams, J.E.: Automatic location of vertebrae on DXA images using random forest regression. In: Ayache, N., Delingette, H., Golland, P., Mori, K. (eds.) MICCAI 2012. LNCS, vol. 7512, pp. 361–368. Springer, Heidelberg (2012). doi: 10.1007/978-3-642-33454-2_45 CrossRefGoogle Scholar
- 12.Schulter, S., Leistner, C., Roth, P., Gool, L., Bischof, H.: On-line Hough forests. In: BMVC (2011)Google Scholar
- 13.Segev, N., Harel, M., Mannor, S., Crammer, K., El-Yaniv, R.: Learn on source, refine on target: a model transfer learning framework with random forests. IEEE TPAMI (2016). doi: 10.1109/TPAMI.2016.2618118 (ePub)
- 14.Štern, D., Ebner, T., Urschler, M.: From local to global random regression forests: exploring anatomical landmark localization. In: Ourselin, S., Joskowicz, L., Sabuncu, M.R., Unal, G., Wells, W. (eds.) MICCAI 2016. LNCS, vol. 9901, pp. 221–229. Springer, Cham (2016). doi: 10.1007/978-3-319-46723-8_26 CrossRefGoogle Scholar