The Hill-RBF Calculator is an advanced, self-validating method for IOL power selection employing pattern recognition by artificial intelligence. It has been optimized for use with the Haag-Streit LENSTAR LS 900 optical biometer for all axial measurements and in combination with high density autokeratometry.
Radial basis function IOL power selection performs in the same manner for short, normal and long eyes. Based in artificial intelligence, this methodology is entirely data driven and free of calculation bias.
The fundamental advantage of pattern recognition for IOL power selection is achieved through the process of adaptive learning - the ability to learn tasks based solely on data, independent of what is previously known. Current methods are prone to limit possibilities to situations that are already understood and where manual optimization methods significantly limit the number of solution options. This method is also self-organizing, meaning that it has the ability to create its own organization, or representation of data. Such an approach is well-suited to the complex, non-linear, real-world problems where ideal models are not available.
Unlike static, theoretical formulas, this approach will be an ongoing project and continuously updated as a big data exercise. The greater the number of validated surgical outcomes that are fit to the RBF artificial intelligence model, the greater the overall breadth and depth of it accuracy. In 2018, the size of the RBF calculator database for version 2 was increased to 12,419 cases. For version 3, released in December of 2020, the size of the RBF calculator database was been further increased and refined.
Hill-RBF Calculator Version 3.0 includes the following updates:
- Version 3 is now capable of in-bounds calculations for biconvex IOLs from +34.00 D to +6.00 D and for meniscus design IOLs from +5.00 D to -5.00 D.
- IOL power selection by Hill-RBF 3 is now also based on the lens thickness, WTW, central corneal thickness and gender for a total of 8 variables. For a large series of patients, it was possible to identify gender as an influencing factor for IOLs power.
- If any of the additional 4 variables are not entered, the calculation is still carried out based on those variables that were entered. In this case, however, the accuracy may be slightly less. The best accuracy is obtained when all 8 variables have been entered.
- The boundary model for Hill-RBF 3 has been greatly expanded so that very few out-of-bounds cases will be indicated. Most out-of-bounds indications are seen when an unusual post-op target SphEq is selected, like +1.50 D.
- The addition of LT, WTW, CCT and gender increases the accuracy for cases with unusual combinations of anterior segment measurements.