uss: Ultrasound signatures (uss) detection

In Morphonodepredictivemodel/morphonode: The Morphonode Predictive Model for ultrasound signatures detection and malignancy prediction in vulvar cancer

Ultrasound signatures (uss) detection

Description

This function implements the decision tree described in Fragomeni et al. 2022 (Morphonode-DT module) to detect metastatic risk signatures (MRSs). The first split detects subjects with a "MET" (metastatic) signature. These individuals show at least one among the three metastatic markers (see details) and a high risk of malignancy (86-100 usually coming with multiple metastatic lymph nodes. The other decision tree branches are defined on the base of five key ultrasound features (see details), with a malignancy risk (MR) signature ranging from low ("LMR", 0-10 to high ("HMR", 52-90 in most malignancies. The main goal of MRSs is to predict single-metastatic event malignancies (HMR) and multiple-metastatic event malignancies (MET).

Usage

uss(
  x,
  dichotomous = FALSE,
  dct = 2,
  short = 8,
  cortical = 2,
  ist = 1,
  ecs = 1,
  hab = 0,
  eco = 1,
  vp = c(1, 2, 3),
  vfl = c(2, 3, 4),
  ct = c(2),
  fid = c(1, 2, 3),
  cmid = c(2, 3, 4),
  shape = c(3),
  cs = 3,
  grouping = c(2, 3),
  ...
)

Arguments

x	An (n, 14) ultrasound features data.frame, where n is the number of subjects.
dichotomous	Logical value. The first step of this function is to dichotomize the input data.frame. Set dichotomous = TRUE if the input data.frame is already dichotomous (default = FALSE).
dct	Numeric value. Moderate risk signature predicts a baseline risk of malignancy equal to 0.16 (CI95 Fragomeni et al. (2022), if at least 2 other ultrasound features than those used in the decision tree (referred to as diagnostic covariates) are above their optimal threshold, the malignancy risk increases to 0.55 (CI95 as MMR1, in contrast to the basal MMR signature (MMR0). The argument dct controls the number of diagnostic covariates needed to switch from MMR0 to MMR1 (default = 2).
short	Numeric value corresponding to the short axis cutoff in millimeters (default = 8).
cortical	Numeric value corresponding to the cortical thickness cutoff in millimeters (default = 2).
ist	Dichotomous value 0, 1 for the presence of inflammatory stroma (perinodal hyperechogenic ring; default = 1).
ecs	Dichotomous value 0, 1 for the presence of extracapsular spread (cortical interruption; default = 1).
hab	Dichotomous value 0, 1 for the absence of the hilum (nodal core sign; default = 0).
eco	Dichotomous value 0, 1 for heterogeneous echogenicity (echostructure; default = 1).
vp	Categorical value (integers from 0 to 4) associated to a high-risk vascular flow architecture pattern (default = c(1, 2, 3)).
vfl	Categorical value (integers from 0 to 4) associated to a high-risk vascular flow localization (default = c(2, 3, 4)).
ct	Categorical value (integers from 0 to 4) associated to a high-risk cortical thickening (default = 2).
fid	Categorical value (integers from 0 to 3) associated to a high-risk focal intranodal deposit (default = c(1, 2, 3)).
cmid	Categorical value (integers from 0 to 4) associated to a high-risk cortical-medullar interface distortion (default = c(2, 3, 4)).
shape	Categorical value (integers from 1 to 3) associated to a high-risk shape (default = 3).
cs	Ordinal value (integers from 1 to 5) associated to a high-risk color score (default = 3).
grouping	Categorical value (integers from 1 to 3) associated to a high-risk grouping (default = c(2, 3)).
...	Currently ignored.

Details

The core method of the Morphonode-DT model is implemented in this function. A series of binary branching points define the metastatic risk signature (MRS) of the subject. The first branch point is based on the evaluation of three metastatic markers: the absence of the nodal core sign (hilum), the presence of the perinodal hyperechogenic ring, and the presence of cortical interruption. If at least one of these conditions are true, the ultrasound profile has a high malignancy risk (86-100 multiple metastatic lymph nodes (Fragomeni et al. 2022). This is referred to as the metastatic (MET) signature. A cortical thickness below 2 mm defines a low metastatic risk (LMR) signature (0.04, CI95 malignant. Based on the values of short axis, vascular flow architecture pattern, cortical thickening, and vascular flow localization, two more signatures are defined: (i) moderate metastatic risk (MMR; 0.16, CI95 CI95 mostly malignant, but chraracterized by a single metastatic event. MRSs should be always compared to the output of the random forest classifier (Morphonode-RFC module) and robust binomial model (Morphonode-RBM module). The main advantage of MRSs is the prediction of either multiple (MET signature) or single (HMR signature) metastasis events.

Value

A list of 4 objects:

1. "signature", metastatic risk signature (MRS);

2. "p", MRS-associated malignancy risk (evaluated as positive predictive value, according to Fragomeni et al. 2022);

3. "ci95", 95

4. "y.uss", naive guess of the outcome (0: non-malignant, 1: malignant) based on the MRS (this will be less accurate than the RFC-based prediction).

Author(s)

Fernando Palluzzi [e-mail]

References

Fragomeni SM, Moro F, Palluzzi F, Mascilini F, Rufini V, Collarino A, Inzani F, Giacò L, Scambia G, Testa AC, Garganese G (2022). Evaluating the risk of inguinal lymph node metastases before surgery using the Morphonode Predictive Model: a prospective diagnostic study. Ultrasound xx Xxxxxxxxxx xxx Xxxxxxxxxx. 00(0):000-000. <https://doi.org/00.0000/00000000000000000000>

See Also

See us.predict to launch all morphonode modules at once.

Examples

# Extract 5 random subjects from the default simulated dataset
x <- mosaic::sample(mpm.us[, 2:15], 5, replace = FALSE, prob = NULL)
print(x)

# Assign a metastatic risk signature to each subject in the dataset
mrs <- uss(x)
x$signature <- mrs$signature
print(x)

Morphonodepredictivemodel/morphonode documentation built on Feb. 15, 2023, 4:51 a.m.