Segmentation of anatomical regions and lesions

1. A computer-aided method of segmenting regions in medical images, the method comprising:

receiving input data;

analysing the input data by identifying one or more regions, wherein analysing the input data is performed using one or more Fully Convolutional Networks (FCNs), wherein the one or more FCNs do not have any dense layers, wherein each neuron of a dense layer receives input from each neuron of a previous layer;

determining one or more characteristics for the one or more regions in the input data; and

generating output segmentation data in dependence upon the characteristics for the one or more regions.

2. The method of claim 1 , wherein the or each FCN comprises one or more convolutional layers and/or one or more hidden representations.

3. The method of claim 1 , wherein the or each FCN comprises one or more hidden representations.

4. The method of claim 1 , wherein the or each FCN comprises one or more activation layers, the one or more activation layers comprising one or more rectified linear units (ReLU) and/or exponential linear units (ELU).

5. The method of claim 1 , wherein the or each FCN comprises one or more sigmoid activation layers and/or softmax functions for the or each region.

6. The method of claim 1 , wherein the input data comprises medical image data and/or one or more Digital Imaging and Communications in Medicine (DICOM) files.

7. The method of claim 6 , wherein the medical image data comprises one or more mammograms, the one or more regions comprises an anatomical region, and the anatomical region comprise at least part of a human breast area.

8. The method of claim 6 , wherein the medical image data comprises a 4D tensor.

9. The method of claim 8 , wherein: the 4D tensor is of size [1, height, width, 1].

10. The method of claim 9 , wherein pixel values of the medical image data are fit to a windowing level supplied by the DICOM file and then represented as 16-bit, and wherein the medical image data is rescaled to a width of between 750 and 900 pixels and/or a height of between 750 and 900 pixels.

11. The method of claim 1 , wherein the output segmentation data comprises an overlay.

12. The method of claim 11 , wherein the overlay comprises a segmentation outline and/or probability map showing one or more locations of the one or more regions.

13. The method of claim 12 , wherein voids within the segmentation outline are operable to be removed.

14. The method of claim 1 , further comprising one or more of:

generating one or more probability masks for the one or more regions;

converting one or more of the one or more probability masks to one or more binary masks, wherein the converting is performed by thresholding the probabilities; and

removing one or more parts of the one or more binary masks with reference to an assigned threshold.

15. The method of claim 14 , wherein the one or more binary masks are one or both of: upscaled to the original size of the input data; and/or stored in the form of a DICOM file.

16. The method of claim 14 , wherein the one or more binary masks comprise one or more identifications of masses and/or calcifications, and the one or more regions comprises an anatomical region and/or a lesion.

17. The method of claim 1 , wherein the input data comprises a plurality of patches and analysing the input data comprises:

analysing the plurality of patches of the input data through sliding windows;

calculating a prediction score for each of the plurality of patches; and

determining an overall prediction score comprising a mean score across the plurality of patches.

18. The method of claim 17 , wherein calculating the prediction score comprises:

determining a probability that a given patch is cancerous; and

applying the determined probability to each pixel within the given patch.

19. A system for segmenting regions in medical images, the system comprising:

a memory including instructions;

one or more processors to execute the instructions to receive input data;

analyse the input data by identifying one or more regions, wherein analysing the input data is performed using one or more Fully Convolutional Networks (FCNs), wherein the one or more FCNs do not have any dense layers, wherein each neuron of a dense layer receives input from each neuron of a previous layer;

determine one or more characteristics for the one or more regions in the input data; and

generate output segmentation data in dependence upon the characteristics for the one or more regions.

20. A computer program product including one or more non-transitory machine readable mediums encoded with instructions that when executed by one or more processors cause a process to be carried out for segmenting regions in medical images, the process comprising:

receiving input data;

analysing the input data by identifying one or more regions, wherein analysing the input data is performed using one or more Fully Convolutional Networks (FCNs), wherein the one or more FCNs do not have any dense layers, wherein each neuron of a dense layer receives input from each neuron of a previous layer;

determining one or more characteristics for the one or more regions in the input data; and

generating output segmentation data in dependence upon the characteristics for the one or more regions.