Cable cell morphology

Cell morphologies are required to describe a Cable cells. Morphologies can be constructed from segment_trees, or read from a number of file formats; see The stitch-builder interface for details.

Segment tree

A segment_tree is – as the name implies – a set of segments arranged in a tree structure, ie each segment has exactly one parent and no child is the parent of any of its ancestors. The tree starts at a root segment which has no parent. Each segment comprises two points in 3d space together with the radii at these points. The segment’s endpoints are called proximal (at the parent’s distal end) and distal (farther from the root).

Segment trees are used to form morphologies which ignore the 3d information encoded in the segments and just utilise the radii, length, and tree-structure. Branches in the tree occur where a segment has more than one child. The tree is constructed by appending segments to the tree. Segments are numbered starting at 0 in the order that they are added, with the first segment getting id 0, the second segment id 1, and so forth. A segment can not be added before its parent, hence the first segment is always at the root. In this manner, a segment tree is always guaranteed to be in a correct state, with consistent parent-child indexing, and with n segments numbered from 0 to n-1. The first parent must be mnpos, indicating ‘no parent’.

class segment_tree
segment_tree()

Construct an empty segment tree.

msize_t append(msize_t parent, const mpoint &prox, const mpoint &dist, int tag)

Append a segment to the tree. Returns the new parent’s id.

msize_t append(msize_t parent, const mpoint &dist, int tag)

Append a segment to the tree whose proximal end has the location and radius of the distal end of the parent segment. Returns the new parent’s id.

This version of append can’t be used for a segment at the root of the tree, that is, when parent is mnpos, in which case both proximal and distal ends of the segment must be specified.

bool empty()

If the tree is empty (i.e. whether it has size 0)

msize_t size()

The number of segments.

std::vector<msize_t> parents()

A list of parent indices of the segments.

std::vector<msegment> segments()

A list of the segments.

std::pair<segment_tree, segment_tree> split_at(const segment_tree &t, msize_t id)

Split a segment_tree into a pair of subtrees at the given id, such that one tree is the subtree rooted at id and the other is the original tree without said subtree.

segment_tree join_at(const segment_tree &t, msize_t id, const segment_tree &o)

Join two subtrees at a given id, such that said id becomes the parent of the inserted sub-tree.

std::vector<msize_t> tag_roots(const segment_tree &t, int tag)

Get IDs of roots of a region with specific tag in the segment tree, i.e. segments whose parent is either mnpos or a segment with a different tag.

bool equivalent(const segment_tree &l, const segment_tree &r)

Two trees are equivalent if 1. the root segments’ prox and dist points and their tags are identical. 2. recursively: all sub-trees starting at the current segment are pairwise equivalent.

Note that under 1 we do not consider the id field.

segment_tree apply(const segment_tree &t, const isometry &i)

Apply an isometry to the segment tree, returns the transformed tree as a copy. Isometries are rotations around an arbritary axis and/or translations; they can be instantiated using isometry::translate and isometry::rotate and combined using the * operator.

Morphology API

Todo

Describe morphology methods.

The stitch-builder interface

Like the segment tree, the stich_builder class constructs morphologies through attaching simple components described by a pair of mpoint values, proximal and distal. These components are mstitch objects, and they differ from segments in two regards:

  1. Stitches are identified by a unique string identifier, in addition to an optional tag value.

  2. Stitches can be attached to a parent stitch at either end, or anywhere in the middle.

The ability to attach a stitch some way along another stitch dictates that one stitch may correspond to more than one morphological segment once the morphology is fully specified. When these attachment points are internal to a stitch, the corresponding geometrical point is determined by linearly interpolating between the proximal and distal points.

The required header file is arbor/morph/stitch.hpp.

mstitch has two constructors:

mstitch::mstitch(std::string id, mpoint prox, mpoint dist, int tag = 0)
mstitch::mstitch(std::string id, mpoint dist, int tag = 0)

If the proximal point is omitted, it will be inferred from the point at which the stitch is attached to its parent.

The stitch_builder class collects the stitches with the add method:

stitch_builder::add(mstitch, const std::string& parent_id, double along = 1.)
stitch_builder::add(mstitch, double along = 1.)

The first stitch will have no parent. If no parent id is specified for a subsequent stitch, the last stitch added will be used as parent. The along parameter must lie between zero and one inclusive, and determines the point of attachment as a relative position between the parent’s proximal and distal points.

A stitched_morphology is constructed from a stitch_builder, and provides both the morphology built from the stitches, and methods for querying the extent of individual stitches.

class stitched_morphology
stitched_morphology(const stitch_builder&)
stitched_morphology(stitch_builder&&)

Construct from a stitch_builder. Note that constructing from an rvalue is more efficient, as it avoids making a copy of the underlying tree structure.

arb::morphology morphology() const

Return the constructed morphology object.

region stitch(const std::string &id) const

Return the region expression corresponding to the specified stitch.

std::vector<msize_t> segments(const std::string &id) const

Return the collection of segments by index comprising the specified stitch.

label_dict labels(const std::string &prefix = "") const

Provide a label_dict with a region entry for each stitch; if a prefix is provided, this prefix is applied to each segment id to determine the region labels.

Example code, constructing a cable cell from a T-shaped morphology specified by two stitches:

using namespace arb;

mpoint soma0{0, 0, 0, 10};
mpoint soma1{20, 0, 0, 10};
mpoint dend_end{10, 100, 0, 1};

stitch_builder builder;
builder.add({"soma", soma0, soma1, 1});
builder.add({"dend", dend_end, 4}, "soma", 0.5);

stitched_morphology stitched(std::move(builder));

auto dec = decor{}.paint("\"soma\"", density("hh"));

cable_cell cell(stitched.morphology(), dec, stitched.labels());

Identifying sites and subsets of the morphology

Todo

TODO: Region and locset documentation.

Translating regions and locsets to cables and locations

Todo

TODO: mprovider, mextent and thingify.

From morphologies to points and segments

The morphology class has the branch_segments method for returning a vector of msegment objects that describe the geometry of that branch. However, determining the position in space of an mlocation, for example, requires some assumptions about how to position points which fall inside a morphological segment.

The place_pwlin class takes a morphology (and optionally an isometry) and interprets it as describing a piecewise-linear object in space. It can then be queried to find the 3-d positions in space of points on the morphology and the extents in space of morphological sub-regions.

Because the morphology need not be contiguous in space, a position query can potentially give more than one possible answer. Similarly, a description of a cable in terms of segments or partial segments in space may include multiple zero-length components as a result of such discontinuities.

class place_pwlin
place_pwlin(const morphology&, const isometry& = isometry())

Construct a piecewise linear placement of the morphology in space, optionally applying the given isometry.

mpoint at(mlocation) const

Return any single point corresponding to the given mlocation in the placement.

std::vector<mpoint> all_at(mlocation) const

Return all points corresponding to the given mlocation in the placement.

std::vector<msegment> segments(const mextent&) const

Return any minimal collection of segments and partial segments whose union is coterminous with the given mextent in the placement.

std::vector<msegment> all_segments(const mextent&) const

Return the maximal set of segments and partial segments whose union is coterminous with the given mextent in the placement.

closest(double x, double y, double z) -> std::pair<mpoint, double>

Find the closest location to p. Returns the location and its distance from the input coordinates.

Isometries

The one cellular morphology may be used to represent multiple cable cells which are in principle sited in different locations and orientations. An explicit isometry allows the one morphology to be repositioned so as to answer location queries on such cells.

An isometry consists of a rotation and a translation. Isometries can be composed; as interpreted by Arbor, translations are always regarded as being relative to the absolute, extrinsic co-ordinate system, while rotations are interpreted as intrinsic rotations: rotations are always applied with respect to the coordinate system carried with the object, not the absolute co-ordinate axes.

class isometry
isometry()

Construct an identity isometry.

static isometry translate(double x, double y, double z)

Construct a translation (x, y, z) with respect to the extrinsic coordinate system.

template<typename Point>
static isometry translate(const Point &p)

Construct a translation (p.x, p.y, p.z) from an arbitrary object with the corresponding public member variables.

static isometry rotate(double theta, double x, double y, double z)

Construct a rotation of theta radians about the axis (x, y, z) with respect to the intrinsic coordinate system.

template<typename Point>
static isometry translate(double theta, const Point &p)

Construct a rotation of theta radians about the (p.x, p.y, p.z) from an arbitrary object with the corresponding public member variables.

template<typename Point>
Point apply(Point p) const

The Point object is interpreted as a point in space given by public member variables x, y, and z. The isometry is applied to the point (x, y, z), and a copy of p is returned with the new coordinate values.

isometry operator*(const isometry &a, const isometry &b)

Compose two isometries to form a new isometry which applies the intrinsic rotation of b, and then the intrinsic rotation of a, together with the translations of both a and b.

Discretisation and CV policies

The set of boundary points used by the simulator is determined by a CV policy. These are objects of type cv_policy, which has the following public methods:

class cv_policy
locset cv_boundary_points(const cable_cell&) const

Return a locset describing the boundary points for CVs on the given cell.

region domain() const

Give the subset of a cell morphology on which this policy has been declared, as a morphological region expression.

Specific CV policy objects are created by functions described below (strictly speaking, these are class constructors for classes are implicit converted to cv_policy objects). These all take a region parameter that restrict the domain of applicability of that policy; this facility is useful for specifying differing discretisations on different parts of a cell morphology. When a CV policy is constrained in this manner, the boundary of the domain will always constitute part of the CV boundary point set.

CV policies can be composed with + and | operators. For two policies A and B, A + B is a policy which gives boundary points from both A and B, while A | B is a policy which gives all the boundary points from B together with those from A which do not within the domain of B. The domain of A + B and A | B is the union of the domains of A and B.

cv_policy_single

cv_policy_single(region domain = reg::all())

Use one CV for the whole cell, or one for each connected component of the supplied domain.

cv_policy_explicit

cv_policy_explicit(locset locs, region domain = reg::all())

Use the points given by locs for CV boundaries, optionally restricted to the supplied domain.

cv_policy_every_segment

cv_policy_every_segment(region domain = reg::all())

Use every segment in the morphology as a CV, optionally restricted to the supplied domain. Each fork point in the domain is represented by a trivial CV.

cv_policy_fixed_per_branch

cv_policy_fixed_per_branch(unsigned cv_per_branch, region domain, cv_policy_flag::value flags = cv_policy_flag::none);

cv_policy_fixed_per_branch(unsigned cv_per_branch, cv_policy_flag::value flags = cv_policy_flag::none):

For each branch in each connected component of the domain (or the whole cell, if no domain is given), evenly distribute boundary points along the branch so as to produce exactly cv_per_branch CVs.

By default, CVs will terminate at branch ends. If the flag cv_policy_flag::interior_forks is given, fork points will be included in non-trivial, branched CVs and CVs covering terminal points in the morphology will be half-sized.

cv_policy_max_extent

cv_policy_max_extent(double max_extent, region domain, cv_policy_flag::value flags = cv_policy_flag::none);

cv_policy_max_extent(double max_extent, cv_policy_flag::value flags = cv_policy_flag::none):

As for cv_policy_fixed_per_branch, save that the number of CVs on any given branch will be chosen to be the smallest number that ensures no CV will have an extent on the branch longer than max_extent micrometres.

CV discretization as mcables

It is often useful for the user to have a detailed view of the CVs generated for a given morphology and cv-policy. For example, while debugging and fine-tuning model implementations, it can be helpful to know how many CVs a cable-cell is comprised of, or how many CVs lie on a certain region of the cell.

The following classes and functions allow the user to inspect the CVs of a cell or region.

class cell_cv_data

Stores the discretisation data of a cable-cell in terms of CVs and the mcables comprising each of these CVs.

mcable_list cables(unsigned idx) const

Returns an vector of mcable representing the CV at a given index.

std::vector<unsigned> children(unsigned idx) const

Returns a vector of the indices of the CVs representing the children of the CV at index idx.

unsigned parent(unsigned idx) const

Returns the index of the CV representing the parent of the CV at index idx.

unsigned size() const

Returns the total number of CVs on the cell.

std::optional<cell_cv_data> cv_data(const cable_cell &cell)

Constructs a cell_cv_data object representing the CVs comprising the cable-cell according to the cv_policy provided in the decor of the cell. Returns std::nullopt_t if no cv_policy was provided in the decor.

class cv_proportion
unsigned idx

Index of the CV.

double proportion

Proportion of the CV by area.

std::vector<cv_proportion> intersect_region(const region &reg, const cell_cv_data &cvs, bool integrate_by_length = false)

Returns a vector of cv_proportion identifying the indices of the CVs from the cell_cv_data argument that lie in the provided region, and how much of each CV belongs to that region. The proportion of CV lying in the region is the area proportion if integrate_by_length is false, otherwise, it is the length proportion.

Supported morphology formats

Arbor supports morphologies described using the SWC file format and the NeuroML file format.

SWC

Arbor supports reading morphologies described using the SWC file format. And has three different interpretation of that format.

A parse_swc() function is used to parse the SWC file and generate a swc_data object. This object contains a vector of swc_record objects that represent the SWC samples, with a number of basic checks performed on them. The swc_data object can then be used to generate a morphology object using one of the following functions: (See the morphology concepts page for more details).

class swc_record
int id

ID of the record

int tag

Structure identifier (tag).

double x

x coordinate in space.

double y

y coordinate in space.

double z

z coordinate in space.

double r

Sample radius.

int parent_id

Record parent’s sample ID.

class swc_data
std::string metadata

Contains the comments of an SWC file.

std::vector<swc_record> records

Stored the list of samples from an SWC file, after performing some checks.

swc_data parse_swc(std::istream&)

Returns an swc_data object given an std::istream object.

morphology load_swc_arbor(const swc_data &data)

Returns a morphology constructed according to Arbor’s SWC specifications.

morphology load_swc_neuron(const swc_data &data)

Returns a morphology constructed according to NEURON’s SWC specifications.

Neurolucida ASCII

Arbor supports reading morphologies described using the Neurolucida ASCII file format.

The parse_asc() function is used to parse the SWC file and generate a asc_morphology object: a simple struct with two members representing the morphology and a label dictionary with labeled regions and locations.

class asc_morphology
arb::morphology morphology
arb::label_dict labels
asc_morphology load_asc(const std::string &filename)

Parse a Neurolucida ASCII file. Throws an exception if there is an error parsing the file.

NeuroML

Arbor offers limited support for models described in NeuroML version 2. All classes and functions provided by the arborio library are provided in the arborio namespace.

NeuroML2 morphology support

NeuroML documents are represented by the arborio::neuroml class, which in turn provides methods for the identification and translation of morphology data. neuroml objects are moveable and move-assignable, but not copyable.

An implementation limitation restricts valid segment id values to those which can be represented by an unsigned long long value.

arborio::neuroml methods can throw an arborio::xml_error in the instance that the underlying XML library reports a problem that cannot be handled by the arborio library. Otherwise, exceptions derived from aborio::neuroml_exception can be thrown when encountering problems interpreting the NeuroML document (see Exceptions below).

Special parsing behaviour can be invoked through the use of an enum value in the neuroml_options namespace.

class neuroml
neuroml(std::string)

Build a NeuroML document representation from the supplied string.

std::vector<std::string> cell_ids() const

Return the id of each <cell> element defined in the NeuroML document.

std::vector<std::string> morphology_ids() const

Return the id of each top-level <morphology> element defined in the NeuroML document.

std::optional<nml_morphology_data> morphology(const std::string&, enum neuroml_options::value = neuroml_options::none) const

Return a representation of the top-level morphology with the supplied identifier, or std::nullopt if no such morphology could be found.

std::optional<nml_morphology_data> cell_morphology(const std::string&, enum neuroml_options::value = neuroml_options::none) const

Return a representation of the morphology associated with the cell with the supplied identifier, or std::nullopt if the cell or its morphology could not be found.

enum neuroml_options::value
enumerator none

Perform no special parsing.

enumerator allow_spherical_root

Replace a zero-length root segment of constant radius with a Y-axis aligned cylindrical segment of the same radius and with length twice the radius. This cylinder will have the equivalent surface area to a sphere of the given radius.

All child segments will connect to the centre of this cylinder, no matter the value of any fractionAlong attribute.

The morphology representation contains the corresponding Arbor arb::morphology object, label dictionaries for regions corresponding to its segments and segment groups by name and id, and a map providing the explicit list of segments contained within each defined segment group.

class nml_morphology_data
std::optional<std::string> cell_id

The id attribute of the cell that was used to find the morphology in the NeuroML document, if any.

std::string id

The id attribute of the morphology.

arb::morphology morphology

The corresponding Arbor morphology.

arb::label_dict segments

A label dictionary with a region entry for each segment, keyed by the segment id (as a string).

arb::label_dict named_segments

A label dictionary with a region entry for each name attribute given to one or more segments. The region corresponds to the union of all segments sharing the same name attribute.

arb::label_dict groups

A label dictionary with a region entry for each defined segment group

std::unordered_map<std::string, std::vector<unsigned long long>> group_segments

A map from each segment group id to its corresponding collection of segments.

Exceptions

All NeuroML-specific exceptions are defined in arborio/neuroml.hpp, and are derived from arborio::neuroml_exception which in turn is derived from std::runtime_error. With the exception of the nml_no_document exception, all contain an unsigned member line which is intended to identify the problematic construct within the document.

class nml_no_document : neuroml_exception

A request was made to parse text which could not be interpreted as an XML document.

class nml_parse_error : neuroml_exception

Failure parsing an element or attribute in the NeuroML document. These can be generated if the document does not confirm to the NeuroML2 schema, for example.

class nml_bad_segment : neuroml_exception

A <segment> element has an improper id attribue, refers to a non-existent parent, is missing a required parent or proximal element, or otherwise is missing a mandatory child element or has a malformed child element.

class nml_bad_segment_group : neuroml_exception

A <segmentGroup> element has a malformed child element or references a non-existent segment group or segment.

class nml_cyclic_dependency : neuroml_exception

A segment or segment group ultimately refers to itself via parent