Fig 1.
Typical magnetic resonance features of image analysis.
(A) diffuse stalk thickening; (B) cystic change; (C) high T1 signal; (D) gland involvement; (E) extrasellar involvement.
Table 1.
Final diagnoses of pituitary stalk lesions.
Table 2.
Comparison of clinical and imaging features between non-neoplastic and neoplastic pituitary stalk lesions.
Table 3.
Clinical and imaging predictors for pituitary metastases among neoplastic pituitary stalk lesions.
Table 4.
Clinical and imaging predictors of non-neoplastic and neoplastic stalk lesions.
Fig 2.
A diagnostic model for a thickened pituitary stalk lesion in distinguishing non-neoplastic pituitary stalk lesions from neoplastic pituitary stalk lesions in the training set, based on recursive partitioning analysis.
PSLs, pituitary stalk lesions.
Fig 3.
A 25-year-old woman with pan-hypopituitarism and pituitary stalk thickening (maximal anterior–posterior diameter: 2 mm).
(A) The patient presented with clinical features of diabetes insipidus and a diffuse pattern of stalk thickening was observed. (B) Similar finding on coronal image, and this lesion was classified as a non-neoplastic based on the diagnostic model, and was diagnosed as hypophysitis via clinico-radiological follow-up after steroid therapy. (C) After 1 year of follow-up, the infundibular thickening had improved.
Fig 4.
A 21-year-old man with pituitary stalk thickening.
(A) The patient presented with symptoms and laboratory findings consistent with diabetes insipidus, showing thickened pituitary stalk with a maximal anterior-posteriro diameter of 5mm. (B) The stalk showed fusiform and non-diffuse thickening. (C) There was no remarkable abnormality in the pineal gland. The lesion was classified as a neoplastic pituitary stalk lesion based on the diagnostic model, and finally diagnosed as a germinoma via biopsy.
Fig 5.
Receiver operating characteristic curve comparison of our diagnostic model with the original radiological report in terms of differentiating non-neoplastic from neoplastic pituitary stalk lesions in the validation set.