A Whole-Slide Image Managing Library Based on Fastai for Deep Learning in the Context of Histopathology: Two Use-Cases Explained

Background: Processing whole-slide images (WSI) to train neural networks can be intricate and labor intensive. We developed an open-source library dealing with recurrent tasks in the processing of WSI and helping with the training and evaluation of neuronal networks for classification tasks. Methods: Two histopathology use-cases were selected and only hematoxylin and eosin (H&E) stained slides were used. The first use case was a two-class classification problem. We trained a convolutional neuronal network (CNN) to distinguish between dysembryoplastic neuroepithelial tumor (DNET) and ganglioglioma (GG), two neuropathological low-grade epilepsy-associated tumor entities. Within the second use case, we included four clinicopathological disease conditions in a multilabel approach. Here we trained a CNN to predict the hormone expression profile of pituitary adenomas. In the same approach, we also predicted clinically silent corticotroph adenoma. Results: Our DNET-GG classifier achieved an AUC of 1.00 for the ROC curve. For the second use case, the best performing CNN achieved an area under the curve (AUC) of 0.97 for the receiver operating characteristic (ROC) for corticotroph adenoma, 0.86 for silent corticotroph adenoma, and 0.98 for gonadotroph adenoma. All scores were calculated with the help of our library on predictions on a case basis. Conclusions: Our comprehensive and fastai-compatible library is helpful to standardize the workflow and minimize the burden of training a CNN. Indeed, our trained CNNs extracted neuropathologically relevant information from the WSI. This approach will supplement the clinicopathological diagnosis of brain tumors, which is currently based on cost-intensive microscopic examination and variable panels of immunohistochemical stainings.

A Whole Slide Image Managing Library Based on Fastai for Deep Learning in the Context of Histopathology: Two Use-Cases Explained

Background: Processing whole-slide images (WSI) to train neural networks can be intricate and laborious. We developed an open-source library covering recurrent tasks in processing of WSI and in evaluating the performance of the trained networks for classification tasks. Methods: Two histopathology use-cases were selected. First we aimed to train a CNN to distinguish H&E-stained slides obtained from neuropathologically classified low-grade epilepsy-associated dysembryoplastic neuroepithelial tumor (DNET) and ganglioglioma (GG). The second project we trained a convolutional neural network (CNN) to predict the hormone expression of pituitary adenoms only from hematoxylin and eosin (H&E) stained slides. In the same approach, we addressed the issue to also predict clinically silent corticotroph adenoma. We included four clinico-pathological disease conditions in a multilabel approach. Results: Our best performing CNN achieved an area under the curve (AUC) of 0.97 for the receiver operating characteristic (ROC) for corticotroph adenoma, 0.86 for silent corticotroph adenoma and 0.98 for gonadotroph adenoma. Our DNET-GG classifier achieved an AUC of 1.00 for the ROC curve. All scores were calculated with the help of our library on predictions on a case basis. Conclusions: Our comprehensive library is most helpful to standardize the work-flow and minimize the work-burden in training CNN. It is also compatible with fastai. Indeed, our new CNNs reliably extracted neuropathologically relevant information from the H&E staining only. This approach will supplement the clinico-pathological diagnosis of brain tumors, which is currently based on cost-intense microscopic examination and variable panels of immunohistochemical stainings.

A Whole Slide Image Managing Library Based on Fastai for Deep Learning in the Context of Histopathology: Two Use-Cases Explained

Background: Processing whole-slide images (WSI) to train neural networks can be intricate and laborious. We developed an open-source library covering recurrent tasks in processing of WSI and in evaluating the performance of the trained networks for classification tasks. Methods: Two histopathology use-cases were selected. First we aimed to train a CNN to distinguish H&E-stained slides obtained from neuropathologically classified low-grade epilepsy-associated dysembryoplastic neuroepithelial tumor (DNET) and ganglioglioma (GG). The second project we trained a convolutional neural network (CNN) to predict the hormone expression of pituitary adenoms only from hematoxylin and eosin (H&E) stained slides. In the same approach, we addressed the issue to also predict clinically silent corticotroph adenoma. We included four clinico-pathological disease conditions in a multilabel approach. Results: Our best performing CNN achieved an area under the curve (AUC) of 0.97 for the receiver operating characteristic (ROC) for corticotroph adenoma, 0.86 for silent corticotroph adenoma and 0.98 for gonadotroph adenoma. Our DNET-GG classifier achieved an AUC of 1.00 for the ROC curve. All scores were calculated with the help of our library on predictions on a case basis. Conclusions: Our comprehensive library is most helpful to standardize the work-flow and minimize the work-burden in training CNN. It is also compatible with fastai. Indeed, our new CNNs reliably extracted neuropathologically relevant information from the H&E staining only. This approach will supplement the clinico-pathological diagnosis of brain tumors, which is currently based on cost-intense microscopic examination and variable panels of immunohistochemical stainings.

Cushing’s disease presenting with psychosis

Cushing’s disease is a rare endocrine condition in which a pituitary corticotroph adenoma drives excess adrenal cortisol production, and is one cause of endogenous Cushing’s syndrome. We present a young woman with 3 weeks of headaches and cognitive disturbance who subsequently developed florid psychosis requiring multiple admissions under neurology and psychiatry. Her clinical stigmata of hypercortisolism and biochemical abnormalities prompted an MR scan of the pituitary, which confirmed a pituitary microadenoma. Treatment with metyrapone and subsequent surgery led to complete recovery within 2 months. Cushing’s disease commonly causes neuropsychiatric symptoms and can present with psychosis. Diagnosing Cushing’s disease can be challenging, but with early diagnosis and treatment it has an excellent prognosis.

Neuromedin B Receptor Antagonist Suppresses Pomc Expression in AtT-20 Cells and Human Corticotroph Adenoma Cells

Objective: We previously reported that Neuromedin B (NMB) is expressed in murine pituitary corticotrophs under adrenal insufficiency (1). Because NMB is also expressed in several cancer cells and stimulates ACTH secretion, we hypothesized that NMB is related to corticotroph adenoma cell proliferation and hormone secretion. To examine this hypothesis, we investigated the expression of NMB and its receptor NMBR in human corticotroph adenoma and the effects of a NMBR antagonist on AtT-20 cells, a mouse corticotroph adenoma cell line, and patient-derived corticotroph adenoma cells. Methods: 1. NMB and NMBR expression in human pituitary adenoma: We performed real-time qPCR and immunostaining on human pathological specimens of corticotrophs, somatotrophs, and non-functioning pituitary adenoma to investigate NMB and NMBR expression. 2. Experiments in AtT-20 cells: We extracted mRNAs and proteins from AtT-20 cells after incubation with 100nM NMBR antagonist PD168368, and performed real-time qPCR and western blotting analyses to investigate Pomc expression. 3. Experiments in patient-derived corticotroph adenoma cells: We isolated surgically resected human corticotroph adenoma cells from patients who underwent trans-sphenoidal surgery and investigated POMC mRNA expression by real-time qPCR after incubation with PD168368. Statistical analysis: One-way ANOVA was employed to compare values among multiple groups. If the ANOVA revealed significant differences, the Tukey-Kramer post-hoc test was employed to compare values between two specific groups. Dunnett’s post-hoc test was employed to compare values with the control group. Statistical significance was defined as p < 0.05. Results: 1. NMB and NMBR expression levels were significantly higher in human corticotroph adenoma (13 and 33 times higher than non-functioning adenoma, respectively) than in somatotroph adenoma (2 and 3 times higher than non-functioning adenoma, respectively) and non-functioning adenoma in the qPCR analyses. Immunostaining confirmed higher expression of NMB and NMBR in corticotroph adenoma than in somatotroph and non-functioning adenoma. 2. Treatment with 100 nM PD168368 significantly suppressed Pomc mRNA and protein expression in AtT-20 cells by 22%±3% and 25%±10%, respectively. 3. Treatment with 1 µM PD168368 significantly suppressed POMC mRNA expression in human corticotroph adenoma cells by 18%±1%. Conclusions: NMB and NMBR were both expressed in human corticotroph adenoma, suggesting that NMB may stimulate adenoma cell proliferation and hormone secretion in autocrine or paracrine manners. Because the NMBR antagonist suppressed Pomc expression in both AtT-20 cells and human corticotroph adenoma cells, it may represent a potential treatment for Cushing disease. Reference: (1) Kameda H et al., Endocrinology 2014;155(7):2492-9.

Co Existence of a Corticotroph Adenoma and a Multifunctional Pituitary Cyst

The co-existence of a corticotroph adenoma and a pituitary cyst is very unusual. We present the case of a 50-year-old female who presented with a Cushingoid phenotype, severe hypokalaemia, hyperglycaemia and hypertension. Urinary free cortisol was markedly elevated at 50 fold. ACTH levels were elevated at 121.4 pg/ml. She failed both the low and high dose dexamethasone suppression tests. The CRF test did not show a satisfactory rise in ACTH levels but the Inferior petrosal sinus sampling revealed a central to peripheral ACTH gradient highly suggestive of Pituitary dependant Cushing’s. MRI pituitary revealed a large cystic lesion with a small solid component. Computed Tomography (CT) of thorax abdomen and pelvis was normal. A trans-sphenoidal hypophysectomy was performed, during which a fine needle aspiration of the intra-cystic fluid was obtained. This showed markedly elevated pituitary hormone levels of ACTH (1399pg/ml), prolactin (353,084mIU/L), TSH (217IU/L) FSH (205mIU/ml) and GH (519 ng/ml) consistent with a multifunctional pituitary cyst. Neuropathology of the solid component confirmed a corticotroph adenoma. Post-operative am cortisol levels were persistently suppressed to less than 50 nmol/l with marked improvement in clinical features. This case highlights the challenges in the work up of Cushing’s syndrome and the limitations of diagnostic tests. The coexistence of a corticotroph adenoma and a multifunctional pituitary cyst is very unusual and to our knowledge has not been reported before.

Cushing’s Disease: Not Always Black and White

Cushing’s Disease is a well known entity but the difficulty of diagnosis is often underappreciated. Given the wide spectrum of clinical presentation and pitfalls with each diagnostic evaluation, diagnosis of Cushing’s Disease (CD) is often difficult in clinical practice, even with pathological analysis. A 37 year old female with a history of Cushing’s Disease presented for her third transphenoidal resection to our institution. She was diagnosed with a pituitary microadenoma 16 years ago based on MRI, as she had persistent headaches and vision abnormalities. The patient also reported unexplained weight gain, “buffalo hump”, and moon facies at the time. She was diagnosed with CD due to an abnormal 1 mg overnight dexamethasone suppression test (DST) with AM cortisol of 3.03 ug/dl (normal <1.8 microgram/dl). She was on an oral contraceptive pill (OCP) at the time. The patient underwent her first pituitary transsphenoidal resection and was symptom free for about 10 years, after which she had a recurrence of her initial symptoms. She had another abnormal DST while on an OCP and pituitary MRI revealed growth of the pituitary adenoma. Patient underwent a second pituitary surgery with benign postoperative course with a recurrence about 5 years later. The workup prior to her third surgery revealed an abnormal DST while on OC pills with the 8am cortisol being 3.36 ug/dl (<1.8ug/dl), urinary free cortisol 35.4 mcg/24 hour (4-50 mcg/24 hour). 8am ACTH done on a separate day was 48 pg/ml (6-50 pg/ml) with a cortisol of 14.5 ug/dl. Midnight salivary cortisol was not performed. Interval history was still positive for weight gain and headaches, hence she was referred for her third pituitary surgery. Post surgery, the patient was on a short taper of hydrocortisone and 8am cortisol was 32 ug/dl the next day. After a discussion with the pathologist, it was determined that the pathology was suggestive of a corticotroph adenoma with moderate ACTH staining and patchy nuclear staining for TPIT, although the pathologist stated that it was difficult to be certain due to the small tissue size. From the current literature, this patient had an incomplete and equivocal biochemical work up while on OCPs but still had tissue diagnosis supporting CD. Whether this was a recurrence of Cushing’s disease, silent corticotroph adenoma, or non-functioning pituitary adenoma is unclear. This case illustrates the multiple challenges in the diagnosis of Cushing’s disease that may be encountered. 1.Braun LT et al. Recurrence after pituitary surgery in adult Cushing’s disease: a systematic review on diagnosis and treatment. Endocrine. 2020;70(2):218-231.

Canine Pituitary Organoids as 3D In Vitro Model for Cushing Disease

Cushing disease (CD) is a serious endocrine disorder that is most often caused by an ACTH-secreting pituitary adenoma. Patients can be treated medically when surgery is not an option or was unsuccessful. However, currently used pituitary-targeting drugs are effective in only 40% of patients. To efficiently identify new pituitary-targeting treatment options, we need an in vitro system that closely mimics in vivo conditions. We therefore aimed to establish organoid cultures of normal anterior pituitary and corticotroph adenomas. Organoids or tumoroids are miniature three-dimensional (3D) structures grown from stem cells, that closely resemble the organ or tumor they originate from. Because CD is a thousand times more prevalent in dogs than in humans, and hypophysectomy is the treatment of choice, we used canine tissues. Normal anterior pituitary glands were collected from three healthy dogs that were euthanized for reasons unrelated to the current study. Corticotroph adenomas were collected from six dogs that underwent transsphenoidal hypophysectomy at our University Clinic. The dogs were diagnosed with CD based on clinical signs, endocrine testing, and CT scan imaging. Normal anterior pituitary and corticotroph adenoma cells were cultured in a 3D matrix (basement membrane extract) with anterior pituitary organoid medium containing specific growth factors and ligands, which was refreshed twice a week. The organoids and tumoroids were characterized with histopathology and RT-qPCR. Structures resembling organoids or tumoroids grew from all nine samples (3 normal, 6 adenoma) that were put in culture. Both cystic and dense structures were observed. The organoids and tumoroids expanded rapidly, and could be passaged once every week. The organoids and tumoroids were successfully cultured up until passage number 10, and were then frozen down. Histopathology showed that the organoid or tumoroid cells morphologically resembled healthy anterior pituitary or corticotroph adenoma cells. All organoids cultures expressed mRNA of pituitary stem cell markers SOX2 and SOX9. This study shows that corticotroph adenomas can be cultured as tumoroids in vitro, something not previously published in any species. Based on the many opportunities in organoid culture (e.g., high-throughput drug screenings, gene editing, studying developmental processes), we expect that this in vitro model will pave the way to efficiently and reliably identify new treatment options for CD. Not only for humans, but also for our best friends: dogs.