AI-powered radiomics for detecting rare pituitary stalk interruption with growth hormone-insulin-like factor desynchrony

To the Editor, Pituitary stalk interruption syndrome (PSIS) is an infrequently occurring congenital condition characterized by pituitary hypoplasia and the thinning or absence of the pituitary stalk. The estimated prevalence is 0.5 per 100 000 live births, with predominance in males. It can be genetic or developmental; structural interruption leads to a deficiency of hormones; nearly all the patients present with growth hormone (GH) deficiency or growth hormone-insulin-like growth factor (GH-IGF) desynchrony, due to hypothalamic-pituitary disruption. GH release becomes absent or disrupted, leading to inappropriately low IGF levels called GH-IGF desynchrony. Common symptomatology includes short stature, poor growth velocity, delayed bone age, recurrent hypoglycemia, and hypogonadotropic hypogonadism[1]. Associations include midline craniofacial defects, ectopic posterior pituitary, and rare genetic syndromes. It is typically diagnosed based on clinical presentation, laboratory data, and imaging; treatment includes hormone replacement therapy tailored to deficient hormones and regular endocrinological follow-up. Radiomics is an advanced field in medical imaging that works by the acquisition of medical images, such as computed tomography (CT) magnetic resonance (MRI), positron emission tomography(PET) scans, segmentation – delineating the regions of interest (tumor or a lesion), extracting the quantitative features, for instance, shape, texture, size, and analyzing this data to conclude. Depending on the modality of CT, MRI, PET and the disease, it has a high 80-90% sensitivity with specificity being more variable at 70-90%[2]. Radiomics is gaining importance in all medical fields due to its potency for early detection, treatment response assessment, and detecting hormone axis abnormalities. Current applications are in oncology for tumor detection and grading, neurology for brain tumors and neurodegenerative diseases, cardiovascular imaging, liver fibrosis, and endocrinology for PSIS or GH-IGH axis abnormalities detection. Radiomics can quantitatively assess gland volume, structural abnormalities, and correlate with hormonal imbalance for detecting PSIS with GH-IGF desynchrony[3]. The GH-IGF axis is pivotal in growth regulation, whereas the desynchrony between these hormones hinders diagnostic clarity. A recent study has revealed that in hypo-insulinemic conditions (for example, type 1 diabetes, anorexia nervosa), GH amounts are raised while IGF-I is decreased. This situation can lead to a secondary association of desynchrony in as much as 100% of cases, since hepatic GH sensitivity is disturbed. In case of severe GH deficiency, 0.16-0.33 mg/kg/week of recombinant GH therapy was effective to raise IGF-I in only 50-60% of the adults, thus making their levels almost equal to those of the age-matched norm, even though normalized GH pulses were present; this was because of the persistent hepatic resistance resulting from low portal insulin. The level of bioactive IGF-I was still 30-40% lower than what was expected, thus confirming the limited efficacy of GH monotherapy in restoring full GH-IGF-I axis synchrony in hypoinsulinemic states. The combination of insulin resistance and hypersecretion of portal insulin affects the liver’s GH sensitivity and IGF-I production, which, in turn, might either mask classical deficiency presentations or change the reliability of biomarkers, especially in adults and chronic patients. Radiomics-driven identification of stalk abnormalities can provide additional support for biochemical assessments and lead to prolonged recognition of desynchrony syndromes within PSIS and better-directed hormone replacement therapy[4]. The main method of diagnosing PSIS involves subjective MRI interpretation, which is subject to differences of over 20% between various evaluators. The detection of a slender or nonexistent stalk is often not noticed on the regular 1.5 T scans, leading to a postponement in the diagnosis of 30-40% of the infants with hormone deficiency until the other hormone deficits are resolved. This consequently postpones the administration of early isolated GH therapy, and poor height outcomes are then the end result. This necessitates the extension of the radiomics model’s validation to larger multi-center patient groups and the combination of imaging biomarker assessment along with endocrine testing. Additionally, the clinical presentation of GH-IGF axis desynchrony should lead to the reconsideration of the diagnostic protocols. The future studies need to develop radiomics-based algorithms for the selection of personalized treatment management and for the prediction of therapy responses, thus improving the results in the course of this difficult endocrine disorder[5]. To conclude, PSIS, a highly complex condition, poses significant diagnostic problems due to a combination of human bias in MRI reading, overlooked stalk defects during standard 1.5-T scanning, and GH-IGF-I desynchrony that conceals typical deficiency, especially in patients with a complicated metabolism, causing a wait for isolated GH therapy to become irreversible growth impairment[1,5]. Radiomics, an advanced imaging analysis technique, provides accurate, objective, and reproducible detection of pituitary hypoplasia and stalk interruption, along with critical imaging biomarkers that support biochemical assessment. These advantages promise a new perspective in the early diagnosis of PSIS and informed decisions regarding treatment[2,3]. The validation of radiomics models on paediatric and adult cohorts in various research centers is required to make this potential a reality in clinical settings. The incorporation of radiomic signatures into diagnostic workflows should be the first step, and the development of predictive algorithms to correlate imaging characteristics with treatment response in desynchrony states should be carried out immediately[4,5]. Crucially, these measures aim to minimize the time to diagnosis, ensure optimal early hormone replacement therapy, and ultimately enhance the long-term prognosis associated with this rare congenital condition. This narrative review adheres to the Transparency in the Reporting of Artificial Intelligence in Research (TITAN) guideline[6].