MedNF

MedNF Dataset from the Paper: "MedFuncta: A Unified Framework for Learning Efficient Medical Neural Fields" This is the official dataset, realeased with "MedFuncta: A Unified Framework for Learning Efficient Medical Neural Fields" by Paul Friedrich, Florentin Bieder, Julian McGinnis, Julia Wolleb, Daniel Rueckert and Philippe C. Cattin. Abstract Research in medical imaging primarily focuses on discrete data representations that poorly scale with grid resolution and fail to capture the often continuous nature of the underlying signal. Neural Fields (NFs) offer a powerful alternative by modeling data as continuous functions. While single-instance NFs have successfully been applied in medical contexts, extending them to large-scale medical datasets remains an open challenge. We therefore introduce MedFuncta, a unified framework for large-scale NF training on diverse medical signals. Building on Functa, our approach encodes data into a unified representation, namely a 1D latent vector, that modulates a shared, meta-learned NF, enabling generalization across a dataset. We revisit common design choices, introducing a non-constant frequency parameter $\omega$ in widely used SIREN activations, and establish a connection between this $\omega$-schedule and layer-wise learning rates, relating our findings to recent work in theoretical learning dynamics. We additionally introduce a scalable meta-learning strategy for shared network learning that employs sparse supervision during training, thereby reducing memory consumption and computational overhead while maintaining competitive performance. Finally, we evaluate MedFuncta across a diverse range of medical datasets and show how to solve relevant downstream tasks on our neural data representation. To promote further research in this direction, we release our code, model weights and the first large-scale dataset - MedNF - containing > 500 k latent vectors for multi-instance medical NFs. Project page | Link to arXiv | Code Dataset Descriptor We release a total of 7 MedFuncta sets, with > 550 k annotated neural fields. The released representations are derived from the MedMNIST dataset. In the first version of this dataset, we publish 7 MedNF sets trained with images of 64 x 64 pixels as a supervision signal. We plan to add other supervision resolutions (28 x 28, 128 x 128, 224 x 224) in future releases. Information on the datasets and the according labels can be found in our paper (Appendix D). Data structure We release all datasets in the following form: └───PneumoniaNF └───test └───datapoint_0.pt └───datapoint_1.pt └───datapoint_2.pt ... └───train └───datapoint_0.pt └───datapoint_1.pt └───datapoint_2.pt ... └───val └───datapoint_0.pt └───datapoint_1.pt └───datapoint_2.pt ... └───model.pt└───RetinaNF └───test ... └───train ... └───val ...... Each datapoint.pt contains the representation and according labels. They can be accessed as follows: data = torch.load(self.files[idx], weights_only=False)modulations, label = data['modulations'].float(), data['label'] License ChestNF, PathNF, OctNF, PneumoniaNF, RetinaNF and TissueNF are relesed under a CC BY 4.0 License. DermaNF is released under a CC BY-NC 4.0 License.