Skip to content

test(intensity-histogram): vet the 21 untested IBSI IH dispersion/index features #367

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
sameeul wants to merge 1 commit into
base: main
Choose a base branch
from
Open

test(intensity-histogram): vet the 21 untested IBSI IH dispersion/index features #367

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 5 additions & 0 deletions tests/test_all.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -773,6 +773,11 @@ TEST(TEST_NYXUS, TEST_IH_REQUIRED_PREDICATE)
ASSERT_NO_THROW(test_ih_required_predicate());
}

TEST(TEST_NYXUS, TEST_IH_DISPERSION_AND_INDEX_VALUES)
{
ASSERT_NO_THROW(test_ih_dispersion_and_index_values());
}

TEST(TEST_NYXUS, TEST_PIXEL_INTENSITY_VARIANCE_BIASED)
{
ASSERT_NO_THROW(test_pixel_intensity_variance_biased());
Expand Down
122 changes: 122 additions & 0 deletions tests/test_intensity_histogram.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -45,6 +45,19 @@ static const NyxusPixel intensityHistogramTestData[] =
{0, 0, 1}, {1, 0, 1}, {2, 0, 3}, {3, 0, 5}, {4, 0, 7}
};

// Second ROI for the dispersion / robust family. 17 pixels, N=5 bins over [0,40]
// (binWidth 8, centers {4,12,20,28,36}); freq = {1,5,6,4,1}. Chosen so the robust
// window [p10Index=1, p90Index=3] strictly trims the two tail bins — unlike the 5-px
// fixture above, where the window covers all bins and robust == non-robust.
static const NyxusPixel intensityHistogramRobustData[] =
{
{0,0,0},
{1,0,10},{2,0,10},{3,0,10},{4,0,10},{5,0,10},
{6,0,20},{7,0,20},{8,0,20},{9,0,20},{10,0,20},{11,0,20},
{12,0,30},{13,0,30},{14,0,30},{15,0,30},
{16,0,40}
};

// Builds settings with the common knobs the IH family consumes.
static Fsettings ih_make_settings(int nbins, bool ibsi, double softnan = -7777.0)
{
Expand Down Expand Up @@ -98,6 +111,41 @@ static double ih_get(const std::vector<std::vector<double>>& fvals, Nyxus::Featu
return fvals[(int)fc][0];
}

// Integer-domain run on an arbitrary pixel cloud (no slide / float-domain plumbing).
static void ih_run_px(std::vector<std::vector<double>>& fvals, const Fsettings& s,
const NyxusPixel* px, size_t npx)
{
Dataset ds;
ds.dataset_props.push_back(SlideProps("", ""));

LR roidata(100);
roidata.slide_idx = -1;
load_test_roi_data(roidata, px, npx);
roidata.make_nonanisotropic_aabb();

IntensityHistogramFeatures f;
ASSERT_NO_THROW(f.calculate(roidata, s, ds));

roidata.initialize_fvals();
f.save_value(roidata.fvals);
fvals = roidata.fvals;
}

// Assert a list of (feature, ground-truth) pairs; exact zeros use an absolute tolerance
// (agrees_gt scales tolerance by the ground truth, so it cannot check gt==0).
static void ih_assert_values(const std::vector<std::vector<double>>& fv,
const std::vector<std::pair<Nyxus::Feature2D, double>>& gt)
{
for (const auto& pr : gt)
{
double v = ih_get(fv, pr.first);
if (std::abs(pr.second) < 1e-12)
ASSERT_NEAR(v, pr.second, 1e-9) << "IH feature code " << (int)pr.first;
else
ASSERT_TRUE(agrees_gt(v, pr.second, 1e4)) << "IH feature code " << (int)pr.first;
}
}

// 1) Integer-domain values vs exact hand-computed ground truth.
void test_ih_integer_domain_values()
{
Expand Down Expand Up @@ -202,3 +250,77 @@ void test_ih_required_predicate()
fs.enableFeature((int)Feature2D::IH_ENTROPY_VAL);
ASSERT_TRUE(IntensityHistogramFeatures::required(fs));
}

// 6) Dispersion + index-domain family: the 21 IH features not covered by test (1).
// Ground truth is derived independently from the IBSI first-order definitions applied
// to the discretised N-bin histogram — the 7 "...Value" dispersion features (MAD,
// robust-MAD, median-AD, IQR, CoV, QCoD, robust-mean) match the IBSI intensity-histogram
// definitions; the 14 "...Index" variants are the same statistics over the 1-based bin-index
// domain (Nyxus-specific). Reference generator + full derivation:
// .planning/test-vetting-audit/ih_oracle_ref.py.
//
// Two ROIs are checked so the robust window is exercised both degenerately and non-trivially:
// ROI-1 = {1,1,3,5,7}, N=3 (freq {2,1,2}; robust window covers all bins -> robust==full)
// ROI-2 = 17 px, N=5 (freq {1,5,6,4,1}; robust window trims both tail bins)
void test_ih_dispersion_and_index_values()
{
// ---- ROI-1 (existing 5-px fixture, N=3) ----
std::vector<std::vector<double>> fv1;
ih_run_px(fv1, ih_make_settings(/*nbins*/ 3, /*ibsi*/ true),
intensityHistogramTestData,
sizeof(intensityHistogramTestData) / sizeof(NyxusPixel));

ih_assert_values(fv1, {
// 7 "...Value" dispersion features
{ Feature2D::IH_INTERQUANTILE_RANGE_VAL, 3.5 },
{ Feature2D::IH_MEAN_ABSOLUTE_DEVIATION_VAL, 1.6 },
{ Feature2D::IH_ROBUST_MEAN_ABSOLUTE_DEVIATION_VAL, 1.6 }, // == full MAD here
{ Feature2D::IH_MEDIAN_ABSOLUTE_DEVIATION_VAL, 1.6 },
{ Feature2D::IH_COEFFICIENT_OF_VARIATION_VAL, 0.4472135955 },
{ Feature2D::IH_QUANTILE_COEFFICIENT_OF_DISPERSION_VAL, 0.4375 },
{ Feature2D::IH_ROBUST_MEAN_VAL, 4.0 }, // == full mean here
// 14 "...Index" variants
{ Feature2D::IH_VARIANCE_IDX, 0.8 },
{ Feature2D::IH_SKEWNESS_IDX, 0.0 }, // symmetric
{ Feature2D::IH_EXCESS_KURTOSIS_IDX, -1.75 },
{ Feature2D::IH_INTERQUANTILE_RANGE_IDX, 2.0 },
{ Feature2D::IH_RANGE_IDX, 2.0 },
{ Feature2D::IH_MEAN_ABSOLUTE_DEVIATION_IDX, 0.8 },
{ Feature2D::IH_ROBUST_MEAN_ABSOLUTE_DEVIATION_IDX, 0.8 },
{ Feature2D::IH_MEDIAN_ABSOLUTE_DEVIATION_IDX, 0.8 },
{ Feature2D::IH_COEFFICIENT_OF_VARIATION_IDX, 0.4472135955 },
{ Feature2D::IH_QUANTILE_COEFFICIENT_OF_DISPERSION_IDX, 0.5 },
{ Feature2D::IH_ENTROPY_IDX, 1.521928095 }, // == entropy_val
{ Feature2D::IH_UNIFORMITY_IDX, 0.36 }, // == uniformity_val
{ Feature2D::IH_ROBUST_MEAN_IDX, 1.0 },
});

// ---- ROI-2 (robust window trims tail bins, N=5) ----
std::vector<std::vector<double>> fv2;
ih_run_px(fv2, ih_make_settings(/*nbins*/ 5, /*ibsi*/ true),
intensityHistogramRobustData,
sizeof(intensityHistogramRobustData) / sizeof(NyxusPixel));

ih_assert_values(fv2, {
{ Feature2D::IH_INTERQUANTILE_RANGE_VAL, 12.3 },
{ Feature2D::IH_MEAN_ABSOLUTE_DEVIATION_VAL, 6.256055363 },
{ Feature2D::IH_ROBUST_MEAN_ABSOLUTE_DEVIATION_VAL, 4.977777778 }, // trimmed != full
{ Feature2D::IH_MEDIAN_ABSOLUTE_DEVIATION_VAL, 6.117647059 },
{ Feature2D::IH_COEFFICIENT_OF_VARIATION_VAL, 0.4089292229 },
{ Feature2D::IH_QUANTILE_COEFFICIENT_OF_DISPERSION_VAL, 0.3178294574 },
{ Feature2D::IH_ROBUST_MEAN_VAL, 19.46666667 }, // trimmed mean
{ Feature2D::IH_VARIANCE_IDX, 0.9965397924 },
{ Feature2D::IH_SKEWNESS_IDX, 0.1178511302 },
{ Feature2D::IH_EXCESS_KURTOSIS_IDX, -0.564525463 },
{ Feature2D::IH_INTERQUANTILE_RANGE_IDX, 2.0 },
{ Feature2D::IH_RANGE_IDX, 4.0 },
{ Feature2D::IH_MEAN_ABSOLUTE_DEVIATION_IDX, 0.7820069204 },
{ Feature2D::IH_ROBUST_MEAN_ABSOLUTE_DEVIATION_IDX, 0.6222222222 },
{ Feature2D::IH_MEDIAN_ABSOLUTE_DEVIATION_IDX, 0.7647058824 },
{ Feature2D::IH_COEFFICIENT_OF_VARIATION_IDX, 0.339411255 },
{ Feature2D::IH_QUANTILE_COEFFICIENT_OF_DISPERSION_IDX, 0.3333333333 },
{ Feature2D::IH_ENTROPY_IDX, 2.021614872 },
{ Feature2D::IH_UNIFORMITY_IDX, 0.2733564014 },
{ Feature2D::IH_ROBUST_MEAN_IDX, 1.933333333 },
});
}
Loading