Editorial. Machine learning and artificial intelligence applied to the diagnosis and management of Cushing disease

View More View Less
  • Department of Neurological Surgery, Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts
Free access

If the inline PDF is not rendering correctly, you can download the PDF file here.

Cushing disease is defined as hypercortisolism resulting from a pituitary tumor that secretes abnormal and excessive amounts of adrenocorticotropic hormone (ACTH). This definition is quite straightforward; however, the actual diagnosis of Cushing disease can be elusive, confusing, and aggravating for both patients and caregivers.3 The source of these diagnostic difficulties lies in the plethora of symptoms, laboratory tests, and imaging studies that are currently utilized in attempts to establish an accurate differential diagnosis. With so many factors that must be taken into account, it would seem that these diagnostic and management efforts would be prime candidates for assessment using artificial intelligence or machine learning platforms.

Currently, the diagnosis and management of Cushing disease can be extraordinarily challenging, clouding the indications for surgery and negatively impacting the outcomes of treatment even after presumed effective therapy.6 Accurate diagnosis and long-term follow-up are critical to fully understanding the outcomes from treatment of this highly morbid disease. The entire patient assessment process hinges on the demonstration and localization of an ACTH-dependent pituitary source of pathologically elevated serum cortisol. Patients may have typical presentations, but often patient presentations are atypical and may exhibit numerous features that overlap with metabolic- and obesity-related endocrine syndromes.11 More than 20 characteristic signs and symptoms have been reported by patients with Cushing disease. These and other features combine to create sensitive but sometimes confusing diagnostic criteria. The diagnosis is further compromised by the fact that many patients suffer from neurocognitive derangements, including mood disorders, psychological instability, and impaired verbal learning and memory.1,2,10 No combination of these clinical features has yet proven specific enough for definitive diagnosis, and each case presents a unique challenge at initial and subsequent presentations.

The article by Zoli and associates describes an attempt to use machine learning in an exhaustive study of 151 consecutive patients evaluated for the diagnosis and management of Cushing disease.12 An ideal machine learning analysis for this type of investigation would include weighted estimates of as many characteristics of patients with Cushing disease as possible, including clinical signs and symptoms, a variety of separate laboratory tests,5 details of neuroimaging and nuclear imaging (PET) studies, and results of dynamic testing, such as inferior petrosal sinus sampling (IPSS) (Table 1).6 The machine learning process then correlates these elements with outcome results associated with an optimal treatment recommendation or a variety of graded treatment strategies. Once Cushing disease of pituitary origin is diagnosed, transsphenoidal surgery for resection of a presumed pituitary adenoma is usually considered the first-line therapy.4,7,9

TABLE 1.

Diagnostic and prognostic variables to be considered in the complex management of Cushing disease and biochemical outcome prediction after surgery

Clinical FindingLaboratory TestImagingOperative FactorTumor Factor
Obesity/weight gainFasting am cortisolMRI, pituitaryTumor foundPathology
Diabetes mellitusFasting am ACTHKnosp score*Tumor removedIHC/TF
Hypertension24-hr UFC ×2CT chest/abd/pelvisExtent of resectionMIB-1
Moon faciesLNSC ×2Plain radiographsCS ext/invasionCrooke’s changes
Facial plethoraSerum K, CaFDG-PETDural invasion
DysmenorrheaDHEASOctreotidePseudocapsule CSF leak
Cognitive changesThyroid (TSH, T4)DotatateDI or SIADH
Altered mood/depressionGonadal (LH, FSH, T, E)Functional MRI
PsychosisProlactin
HirsutismGH
Bruising/thin skinDST, low/high dose
Osteopenia/pathologic fracturesIPSS
Fatigue
Acne
Fungal infections, UTIs
Proximal muscle weakness
Myopathy
“Buffalo hump”
Supraclavicular fat pads
Hair loss/balding
Peripheral edema
Pigmentation changes§
Visual blurring
Poor wound healing
Decreased libido
Abdominal (and other) striae
Bleeding/clotting issues
Family history (e.g., MEN1)

Abd = abdomen; ACTH = adrenocorticotropic hormone; CS ext = cavernous sinus extension; DHEAS = dehydroepiandrosterone sulfate; DI = diabetes insipidus; DST = dexamethasone suppression test; E = estradiol; FSH = follicle-stimulating hormone; GH = growth hormone; IHC/TF = immunohistochemistry/transcription factors; IPSS = inferior petrosal sinus sampling; LH = leutenizing hormone; LNSC = late-night salivary cortisol; MEN1 = multiple endocrine neoplasia type 1; SIADH = syndrome of inappropriate antidiuretic hormone secretion; T = testosterone; TSH = thyroid-stimulating hormone; UFC = urinary free cortisol; UTI = urinary tract infection.

Estimates cavernous sinus invasion.

Cell proliferation marker.

Intraoperative or postoperative.

Knuckles and elbows.

The ambitious study by Zoli and associates is an intriguing first step. Although the number of factors analyzed was limited, the authors were able to isolate some factors that were in fact associated with their admirable postsurgical remission rate (88%). Overall, Zoli et al. selected endpoints to be examined that consisted of gross-tumor removal, postsurgical remission, and “long-term” control of disease. The reported results are quite satisfactory and are a testament to the careful analysis and follow-up provided by the authors.

As the authors noted in their discussion of the results, careful clinical interpretation of the major factors resulting in successful treatment, disease recurrence, or treatment failure is not possible without studies based on an increased patient sample size. One would also presume that more comprehensive evidence from clinical, laboratory, and imaging settings is necessary to fully assess the benefits of a machine learning approach to the thorny questions relating to diagnosis and optimal management in patients with Cushing disease. One can consider at least 29 symptoms and signs variably associated with Cushing disease that would play important roles in machine learning outcome assessment and prediction, along with 11+ potentially important laboratory tests and 11+ imaging procedures, as well as operative findings and histopathological analysis results. We hope the day will soon come when these factors can be part of a comprehensive prospective data set that will be investigated by using the principles of machine learning, thereby uncovering solutions to the confounding mysteries of Cushing disease and its optimal management.

Disclosures

The authors report no conflict of interest concerning the materials or methods used in this study or the findings specified in this paper.

References

  • 1

    Katznelson L: The cognitive, psychological, and emotional presentation of Cushing’s disease, in Laws ER Jr (ed): Cushing’s Disease: An Often Misdiagnosed and Not So Rare Disorder. Cambridge, MA: Academic Press Elsevier, 2017

    • Search Google Scholar
    • Export Citation
  • 2

    Langenecker SA, Weisenbach SL, Giordani B, Briceño EM, Breting LMG, Schallmo MP, : Impact of chronic hypercortisolemia on affective processing. Neuropharmacology 62:217225, 2012

    • Search Google Scholar
    • Export Citation
  • 3

    Laws ER Jr (ed): Cushing’s Disease: An Often Misdiagnosed and Not So Rare Disorder. Cambridge, MA: Academic Press Elsevier, 2017

  • 4

    Laws ER Jr, Jane JA Jr: Surgical treatment of Cushing’s disease, in Laws ER Jr (ed): Cushing’s Disease: An Often Misdiagnosed and Not So Rare Disorder. Cambridge, MA: Academic Press Elsevier, 2017

    • Search Google Scholar
    • Export Citation
  • 5

    Nieman L: Making the diagnosis: laboratory testing and imaging studies, in Laws ER Jr (ed): Cushing’s Disease: An Often Misdiagnosed and Not So Rare Disorder. Cambridge, MA: Academic Press Elsevier, 2017

    • Search Google Scholar
    • Export Citation
  • 6

    Oldfield EH, Doppman JL, Nieman LK, Chrousos GP, Miller DL, Katz DA, : Petrosal sinus sampling with and without corticotropin-releasing hormone for the differential diagnosis of Cushing’s syndrome. N Engl J Med 325:897905, 1991

    • Search Google Scholar
    • Export Citation
  • 7

    Pivonello R, De Leo M, Cozzolino A, Colao A: The treatment of Cushing’s disease. Endocr Rev 36:385486, 2015

  • 8

    Pouratian N, Prevedello DM, Jagannathan J, Lopes MB, Vance ML, Laws ER Jr: Outcomes and management of patients with Cushing’s disease without pathological confirmation of tumor resection after transsphenoidal surgery. J Clin Endocrin Metab 92:33833388, 2007

    • Search Google Scholar
    • Export Citation
  • 9

    Reitmeyer M, Vance ML, Laws ER Jr: The neurosurgical management of Cushing’s disease. Mol Cell Endocrinol 197:7379, 2002

  • 10

    Starkman MN, Giordani B, Berent S, Schork A, Schteingart DE: Elevated cortisol levels in Cushing’s disease are associated with cognitive decrements. Psychosom Med 63:985993, 2001

    • Search Google Scholar
    • Export Citation
  • 11

    Vance ML: Physical presentation of Cushing’s syndrome: typical and atypical presentations, in Laws ER Jr (ed): Cushing’s Disease: An Often Misdiagnosed and Not So Rare Disorder. Cambridge, MA: Academic Press Elsevier, 2017

    • Search Google Scholar
    • Export Citation
  • 12

    Zoli M, Staartjes VE, Guaraldi F, Friso F, Rustici A, Asioli S, : Machine learning–based prediction of outcomes of the endoscopic endonasal approach in Cushing disease: is the future coming? Neurosurg Focus 48(6):E5, 2020

    • Search Google Scholar
    • Export Citation

If the inline PDF is not rendering correctly, you can download the PDF file here.

Contributor Notes

Correspondence Edward R. Laws Jr.: elaws@bwh.harvard.edu.

ACCOMPANYING ARTICLE DOI: 10.3171/2020.3.FOCUS2060.

INCLUDE WHEN CITING DOI: 10.3171/2020.3.FOCUS20213.

Disclosures The authors report no conflict of interest concerning the materials or methods used in this study or the findings specified in this paper.

  • 1

    Katznelson L: The cognitive, psychological, and emotional presentation of Cushing’s disease, in Laws ER Jr (ed): Cushing’s Disease: An Often Misdiagnosed and Not So Rare Disorder. Cambridge, MA: Academic Press Elsevier, 2017

    • Search Google Scholar
    • Export Citation
  • 2

    Langenecker SA, Weisenbach SL, Giordani B, Briceño EM, Breting LMG, Schallmo MP, : Impact of chronic hypercortisolemia on affective processing. Neuropharmacology 62:217225, 2012

    • Search Google Scholar
    • Export Citation
  • 3

    Laws ER Jr (ed): Cushing’s Disease: An Often Misdiagnosed and Not So Rare Disorder. Cambridge, MA: Academic Press Elsevier, 2017

  • 4

    Laws ER Jr, Jane JA Jr: Surgical treatment of Cushing’s disease, in Laws ER Jr (ed): Cushing’s Disease: An Often Misdiagnosed and Not So Rare Disorder. Cambridge, MA: Academic Press Elsevier, 2017

    • Search Google Scholar
    • Export Citation
  • 5

    Nieman L: Making the diagnosis: laboratory testing and imaging studies, in Laws ER Jr (ed): Cushing’s Disease: An Often Misdiagnosed and Not So Rare Disorder. Cambridge, MA: Academic Press Elsevier, 2017

    • Search Google Scholar
    • Export Citation
  • 6

    Oldfield EH, Doppman JL, Nieman LK, Chrousos GP, Miller DL, Katz DA, : Petrosal sinus sampling with and without corticotropin-releasing hormone for the differential diagnosis of Cushing’s syndrome. N Engl J Med 325:897905, 1991

    • Search Google Scholar
    • Export Citation
  • 7

    Pivonello R, De Leo M, Cozzolino A, Colao A: The treatment of Cushing’s disease. Endocr Rev 36:385486, 2015

  • 8

    Pouratian N, Prevedello DM, Jagannathan J, Lopes MB, Vance ML, Laws ER Jr: Outcomes and management of patients with Cushing’s disease without pathological confirmation of tumor resection after transsphenoidal surgery. J Clin Endocrin Metab 92:33833388, 2007

    • Search Google Scholar
    • Export Citation
  • 9

    Reitmeyer M, Vance ML, Laws ER Jr: The neurosurgical management of Cushing’s disease. Mol Cell Endocrinol 197:7379, 2002

  • 10

    Starkman MN, Giordani B, Berent S, Schork A, Schteingart DE: Elevated cortisol levels in Cushing’s disease are associated with cognitive decrements. Psychosom Med 63:985993, 2001

    • Search Google Scholar
    • Export Citation
  • 11

    Vance ML: Physical presentation of Cushing’s syndrome: typical and atypical presentations, in Laws ER Jr (ed): Cushing’s Disease: An Often Misdiagnosed and Not So Rare Disorder. Cambridge, MA: Academic Press Elsevier, 2017

    • Search Google Scholar
    • Export Citation
  • 12

    Zoli M, Staartjes VE, Guaraldi F, Friso F, Rustici A, Asioli S, : Machine learning–based prediction of outcomes of the endoscopic endonasal approach in Cushing disease: is the future coming? Neurosurg Focus 48(6):E5, 2020

    • Search Google Scholar
    • Export Citation

Metrics

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 203 203 145
PDF Downloads 120 120 102
EPUB Downloads 0 0 0