Koike et al identified 5 features on ultrasound that could help identify malignant nonfollicular thyroid nodules. This can help in planning surgical intervention. The authors are from the Noguchi Thyroid Clinic in Oita, Japan.
Nonfollicular lesions:
(1) papillary carcinoma
(2) medullary carcinoma
(3) mucoepidermoid carcinoma
(4) adenomatous goiter
(5) cyst
(6) calcified nodule
Ultrasound parameters:
(1) margin
(2) shape
(3) echo structure
(4) echogenicity
(5) calcification
Parameter |
Ultrasound Finding |
Points |
Estimate |
margin |
well-defined |
0 |
NA |
|
ill-defined |
1 |
1.034 |
shape |
regular |
0 |
NA |
|
irregular |
1 |
3.079 |
echo structure |
cystic |
0 |
NA |
|
solid |
1 |
0.457 |
|
mixed |
2 |
-0.871 |
echogenicity |
isoechoic |
0 |
NA |
|
hyperechoic |
0 |
NA |
|
hypoechoic |
1 |
1.717 |
|
hypoisoechoic |
2 |
0.491 |
calcification |
absent |
0 |
NA |
|
fine |
1 |
1.382 |
|
other |
2 |
0.452 |
X =
= ((points for margin) * (estimate for margin)) + ((points for shape) * (estimate for shape)) + ((points for echo structure) * (estimate for echo structure)) + ((points for echogenicity) * (estimate for echogenicity)) + ((points for calcification) * (estimate for calcification)) + 0.355
probability of malignancy =
= 1 / (1 + EXP((-1) * X))
NOTE : When the data in Figure 3, page 337 is entered into the equation, I get probabilities that are about half that given in the legend to the figure.
Limitations:
• Multivariate analysis was unreliable for predicting malignancy of follicular neoplasms.
• The lack of coverage for follicular lesions would seem to limit the applicability of the method in practice.
Specialty: Hematology Oncology, Surgery, general, Endocrinology