Segmenting Skin Biopsy Images with Coarse and Sparse Annotations using U-Net.

The number of melanoma diagnoses has increased dramatically over the past three decades, outpacing almost all other cancers. Nearly 1 in 4 skin biopsies is of melanocytic lesions, highlighting the clinical and public health importance of correct diagnosis. Deep learning image analysis methods may improve and complement current diagnostic and prognostic capabilities. The histologic evaluation of melanocytic lesions, including melanoma and its precursors, involves determining whether the melanocytic population involves the epidermis, dermis, or both. Semantic segmentation of clinically important structures in skin biopsies is a crucial step towards an accurate diagnosis. While training a segmentation model requires ground-truth labels, annotation of large images is a labor-intensive task. This issue becomes especially pronounced in a medical image dataset in which expert annotation is the gold standard. In this paper, we propose a two-stage segmentation pipeline using coarse and sparse annotations on a small region of the whole slide image as the training set. Segmentation results on whole slide images show promising performance for the proposed pipeline.

Impact of clinical photographs on the accuracy and confidence in the histopathological diagnosis of mycosis fungoides.

BACKGROUND: The histopathological diagnosis of MF is challenging, and there is significant overlap with benign inflammatory processes. Clinical features may be relevant in the assessment of skin biopsies. METHODS: We provided photomicrographs to board-certified dermatopathologists and one hematopathologist with and without accompanying clinical photographs and assessed accuracy and confidence in diagnosing MF. RESULTS: We found that access to clinical photographs improved diagnostic accuracy in both MF and non-MF (distractors); the degree of improvement was significantly higher in the non-MF/distractor category. Across all categories, diagnostic confidence level was higher when clinical images were available. CONCLUSION: These findings suggest that clinical images are useful in making an accurate diagnosis of MF, and may be particularly helpful in ruling it out when an inflammatory disorder is clinically suspected.

Factors associated with use of immunohistochemical markers in the histopathological diagnosis of cutaneous melanocytic lesions.

BACKGROUND: Melanocytic tumors are often challenging and constitute almost one in four skin biopsies. Immunohistochemical (IHC) studies may assist diagnosis; however, indications for their use are not standardized. METHODS: A test set of 240 skin biopsies of melanocytic tumors was examined by 187 pathologists from 10 US states, interpreting 48 cases in Phase I and either 36 or 48 cases in Phase II. Participant and diagnosis characteristics were compared between those who reported they would have ordered, or who would have not ordered IHC on individual cases. Intraobserver analysis examined consistency in the intent to order when pathologists interpreted the same cases on two occasions. RESULTS: Of 187 participants interpreting 48 cases each, 21 (11%) did not request IHC tests for any case, 85 (45%) requested testing for 1 to 6 cases, and 81 (43%) requested testing for ≥6 cases. Of 240 cases, 229 had at least one participant requesting testing. Only 2 out of 240 cases had more than 50% of participants requesting testing. Increased utilization of testing was associated with younger age of pathologist, board-certification in dermatopathology, low confidence in diagnosis, and lesions in intermediate MPATH-Dx classes 2 to 4. The median intraobserver concordance for requesting tests among 72 participants interpreting the same 48 cases in Phases I and II was 81% (IQR 73%-90%) and the median Kappa statistic was 0.20 (IQR 0.00, 0.39). CONCLUSION: Substantial variability exists among pathologists in utilizing IHC.

Hypercellular Encapsulated Neuroma Imitating Desmoplastic Melanoma.

Palisaded encapsulated neuromas (PENs) are benign cutaneous nerve sheath proliferations that typically occur as flesh-colored papules solitarily on the head and neck in adults, with a slight predilection for females. Histopathologically, they are partially or completely encapsulated intradermal nodules with Schwann cells and axons in fascicles separated by clefts. Although these features are often characteristic, the hypercellular variant of PEN can pose a diagnostic challenge in distinguishing between other cellular neural and melanocytic lesions. We herein report a case of hypercellular PEN, which showed striking similarity to desmoplastic melanoma.

Lichenoid Drug Eruption with Prominent Nail Changes Due to Leflunomide in a 12-Year-Old Child.

We present the case of a 12-year-old-girl who developed lichenoid dermatitis approximately 1 year after starting leflunomide for juvenile idiopathic arthritis. The eruption resolved promptly with discontinuation of the suspected culprit agent, supportive of a lichenoid drug eruption, but she subsequently developed markedly dystrophic nails with lichen planus-like features. A biopsy of her cutaneous findings at the time of initial presentation demonstrated lichenoid dermatitis, and a nail matrix biopsy was deferred given clinical correlation. Prominent nail changes in lichenoid drug eruptions, particularly in children, are rare but should be considered in children with new-onset nail dystrophy.

A giant apocrine hidrocystoma of the trunk.

Hidrocystomas are benign cysts that typically present as translucent, bluish dermal nodules on the face and are rarely > 1 cm in size. They are classically categorized as eccrine or apocrine based on histologic features. We present a rare case of a giant apocrine hidrocystoma of the trunk, demonstrating that, although a rare variant, apocrine hidrocystomas can present both off the head and neck, and can be significantly larger in size than previously reported.

Analysis of Clinical and Molecular Factors Impacting Oncologic Outcomes in Undifferentiated Pleomorphic Sarcoma.

BACKGROUND: Undifferentiated pleomorphic sarcomas (UPS) present a diagnostic and therapeutic challenge. Identification of prognostic molecular markers is required for the discovery of novel treatment approaches. The purpose of this study was to correlate clinicopathologic variables, expression of tyrosine kinase receptors, and markers of cell cycle progression and survival with oncologic outcomes. METHODS: A tissue microarray containing 208 primary UPS samples was analyzed by immunohistochemistry for protein markers and in situ hybridization for microRNA. Staining results were correlated with clinicopathologic features and oncologic outcomes. Univariate and multivariate analyses were conducted to assess associations between expression of protein markers, mi-RNA, and outcome. RESULTS: At a median follow-up of 3.9 years (9 years for survivors), 5-year disease-specific survival (DSS) was 63 %. Clinical variables associated with improved DSS included age <61 years, tumor size <10 cm, margin-negative resection, and sporadic-tumor status. At the protein level, loss of cyclin D1 (p = 0.06), pEGFR (p = 0.023), pIGF-1R (p = 0.022), and PTEN (p < 0.001) and overexpression of AXL (p = 0.015) were associated with reduced DSS on univariate analysis. Ki67, PCNA, and pEGFR were more highly expressed in sporadic UPS than radiation-associated (RA-UPS), whereas RA-UPS samples expressed higher levels of both phosphorylated and total IGF-1R. DISCUSSION: Loss of cyclin D1, overexpression of AXL, and loss of PTEN are associated with poor cancer-specific outcomes and warrant further investigation in UPS. The differences in protein expression in sporadic versus RA-UPS may indicate that the activated molecular signaling nodes may be different for each specific histology and also could explain the aggressive phenotype seen in RA-UPS compared with the sporadic lesions.

The hepatocyte growth factor receptor as a potential therapeutic target for dedifferentiated liposarcoma.

Dedifferentiated liposarcomas (DDLPS) are highly resistant to conventional chemo- and radiotherapies, with surgical resection remaining the classic treatment strategy; therefore, there is a pressing need for novel anti-DDLPS-targeted chemotherapeutics. Hepatocyte growth factor receptor (Met) expression is elevated in DDLPS, but the functional role of Met signaling in this disease is not known. We found that the in vitro stimulation of DDLPS cells with hepatocyte growth factor (HGF) elevated the degree of PI3K/AKT and MAPK pathway signaling, and that pro-tumorigenic phenotypes such as cell proliferation, invasion, and migration were significantly enhanced. Conversely, Met knockdown using shRNA-mediated interference decreased HGF-induced Met signaling, the invasive and migratory nature of DDLPS cells in vitro, and the tumorigenicity of DDLPS cells in vivo. These data strongly support the role for Met as a DDLPS therapeutic target. To that end, using EMD1214063, an ATP-competitive kinase inhibitor that targets Met more specifically than other kinases, inhibited Met-dependent signaling, reduced the oncogenicity of DDLPS cells in vitro, and significantly increased the survival of nude mice bearing subcutaneous DDLPS xenografts. These findings support further investigations of HGF-induced Met signaling inhibition in DDLPS, as a potential strategy to enhance clinical outcomes for this disease.

Radiation-Associated Undifferentiated Pleomorphic Sarcoma is Associated with Worse Clinical Outcomes than Sporadic Lesions.

BACKGROUND: Radiation therapy is used increasingly as a component of multidisciplinary treatment for many solid tumors. One complication of such treatment is the development of radiation-associated sarcoma (RAS). Undifferentiated pleomorphic sarcoma (UPS), previously termed "malignant fibrous histiocytoma" (MFH) is the most common histologic subtype of RAS. This study investigated the clinical outcomes for patients with radiation-associated UPS (RA-UPS/MFH). METHODS: The study identified 1068 patients with UPS/MFH treated at the authors' institution. Patient and tumor factors were collected and compared. Regression analysis was performed to identify independent predictors of survival. A matched-cohort survival and recurrence analysis was performed for radiation-associated and sporadic UPS/MFH. RESULTS: The findings showed that RA-UPS/MFH comprised 5.1 % of the UPS population. The median latency to the development of RA-UPS/MFH was 9.3 years. The 5-year disease-specific survival (DSS) was 52.2 % for patients identified with RA-UPS/MFH (n = 55) compared with 76.4 % for patients with unmatched sporadic UPS/MFH (n = 1,013; p < 0.001). A matched-cohort analysis also demonstrated that the 5-year DSS was significantly worse for RA-UPS/MFH (52.2 vs 73.4 %; p = 0.002). Furthermore, higher local recurrence rates were observed for patients with RA-UPS/MFH than for patients with sporadic lesions (54.5 vs 23.5 %; p < 0.001). Radiation-associated status and incomplete resection were identified as independent predictors of local recurrence. CONCLUSION: This study demonstrated worse clinical outcomes for patients with RA-UPS/MFH than for patients with sporadic UPS/MFH. Local recurrence was significantly higher for patients with RA-UPS/MFH, suggesting a unique tumor biology for this challenging disease.

AXL is a potential therapeutic target in dedifferentiated and pleomorphic liposarcomas.

BACKGROUND: AXL is a well-characterized, protumorigenic receptor tyrosine kinase that is highly expressed and activated in numerous human carcinomas and sarcomas, including aggressive subtypes of liposarcoma. However, the role of AXL in the pathogenesis of well-differentiated (WDLPS), dedifferentiated (DDLPS), and pleomorphic liposarcoma (PLS) has not yet been determined. METHODS: Immunohistochemical analysis of AXL expression was conducted on two tissue microarrays containing patient WDLPS, DDLPS, and PLS samples. A panel of DDLPS and PLS cell lines were interrogated via western blot for AXL expression and activity and by ELISA for growth arrest-specific 6 (GAS6) production. AXL knockdown was achieved by siRNA or shRNA. The effects of AXL knockdown on cell proliferation, migration, and invasion were measured in vitro. In addition, AXL shRNA-containing DDLPS cells were assessed for their tumor-forming capacity in vivo. RESULTS: In this study, we determined that AXL is expressed in a subset of WDLPS, DDLPS, and PLS patient tumor samples. In addition, AXL and its ligand GAS6 are expressed in a panel of DDLPS and PLS cell lines. We show that the in vitro activation of AXL via stimulation with exogenous GAS6 resulted in a significant increase in cell proliferation, migration, and invasion in DDLPS and PLS cell lines. Transient knockdown of AXL resulted in attenuation of these protumorigenic phenotypes in vitro. Stable AXL knockdown not only decreased migratory and invasive characteristics of DDLPS and PLS cells in vitro but also significantly diminished tumorigenicity of two dedifferentiated liposarcoma xenograft models in vivo. CONCLUSIONS: Our results suggest that AXL signaling contributes to the aggressiveness of DDLPS and PLS, and that AXL is therefore a potential therapeutic target for treatment of these rare, yet devastating tumors.

Long-term outcomes in patients with radiation-associated angiosarcomas of the breast following surgery and radiotherapy for breast cancer.

BACKGROUND: Radiation-associated angiosarcoma (RAAS) is a devastating disease occasionally observed in breast cancer patients treated with radiation. Due to its rarity, our knowledge-of disease risk factors, epidemiology, treatment, and outcome-is extremely limited. Therefore, we sought to identify clinicopathologic factors associated with local and distant recurrence and disease-specific survival (DSS). METHODS: Radiation-associated angiosarcoma was defined as pathologically confirmed breast or chest wall angiosarcoma arising within a previously irradiated field. A comprehensive search of our institutional tumor registry (1/1/93 through 2/28/11) was used to identify patients (n = 95 females). Patient, original tumor, RAAS treatment, and outcome variables were retrospectively retrieved and assembled into a database. RESULTS: The median follow-up for all RAAS patients was 10.3 (range, 2.4-31.8) years. The latency period following radiation exposure ranged from 1.4 to 26 (median, 7) years. One-year and 5-year DSS rates were 93.5 and 62.6 %, respectively. Reduced risk of local recurrence was observed in patients who received chemotherapy (P = 0.0003). In multivariable analysis, size was found to be an independent predictor of adverse outcome (P = 0.015). CONCLUSIONS: Our study demonstrates that RAAS exhibits high recurrence rates. It also highlights the need for well-designed, multicenter, clinical trials to inform the true utility of chemotherapy in this disease.

High-resolution analysis of four efficient yeast replication origins reveals new insights into the ORC and putative MCM binding elements.

In budding yeast, the eukaryotic initiator protein ORC (origin recognition complex) binds to a bipartite sequence consisting of an 11 bp ACS element and an adjacent B1 element. However, the genome contains many more matches to this consensus than actually bind ORC or function as origins in vivo. Although ORC-dependent loading of the replicative MCM helicase at origins is enhanced by a distal B2 element, less is known about this element. Here, we analyzed four highly active origins (ARS309, ARS319, ARS606 and ARS607) by linker scanning mutagenesis and found that sequences adjacent to the ACS contributed substantially to origin activity and ORC binding. Using the sequences of four additional B2 elements we generated a B2 multiple sequence alignment and identified a shared, degenerate 8 bp sequence that was enriched within 228 known origins. In addition, our high-resolution analysis revealed that not all origins exist within nucleosome free regions: a class of Sir2-regulated origins has a stably positioned nucleosome overlapping or near B2. This study illustrates the conserved yet flexible nature of yeast origin architecture to promote ORC binding and origin activity, and helps explain why a strong match to the ORC binding site is insufficient to identify origins within the genome.

VSGD-Net: Virtual Staining Guided Melanocyte Detection on Histopathological Images.

Detection of melanocytes serves as a critical prerequisite in assessing melanocytic growth patterns when diagnosing melanoma and its precursor lesions on skin biopsy specimens. However, this detection is challenging due to the visual similarity of melanocytes to other cells in routine Hematoxylin and Eosin (H&E) stained images, leading to the failure of current nuclei detection methods. Stains such as Sox10 can mark melanocytes, but they require an additional step and expense and thus are not regularly used in clinical practice. To address these limitations, we introduce VSGD-Net, a novel detection network that learns melanocyte identification through virtual staining from H&E to Sox10. The method takes only routine H&E images during inference, resulting in a promising approach to support pathologists in the diagnosis of melanoma. To the best of our knowledge, this is the first study that investigates the detection problem using image synthesis features between two distinct pathology stainings. Extensive experimental results show that our proposed model outperforms state-of-the-art nuclei detection methods for melanocyte detection. The source code and pre-trained model are available at: https://github.com/kechunl/VSGD-Net.

Improving the Diagnosis of Skin Biopsies Using Tissue Segmentation.

Invasive melanoma, a common type of skin cancer, is considered one of the deadliest. Pathologists routinely evaluate melanocytic lesions to determine the amount of atypia, and if the lesion represents an invasive melanoma, its stage. However, due to the complicated nature of these assessments, inter- and intra-observer variability among pathologists in their interpretation are very common. Machine-learning techniques have shown impressive and robust performance on various tasks including healthcare. In this work, we study the potential of including semantic segmentation of clinically important tissue structure in improving the diagnosis of skin biopsy images. Our experimental results show a 6% improvement in F-score when using whole slide images along with epidermal nests and cancerous dermal nest segmentation masks compared to using whole-slide images alone in training and testing the diagnosis pipeline.

Epithelial-mesenchymal transition and cancer stem cells: a dangerously dynamic duo in breast cancer progression.

Aberrant activation of a latent embryonic program - known as the epithelial-mesenchymal transition (EMT) - can endow cancer cells with the migratory and invasive capabilities associated with metastatic competence. The induction of EMT entails the loss of epithelial characteristics and the de novo acquisition of a mesenchymal phenotype. In breast cancer, the EMT state has been associated with cancer stem cell properties including expression of the stem cell-associated CD44+/CD24-/low antigenic profile, self-renewal capabilities and resistance to conventional therapies. Intriguingly, EMT features are also associated with stem cells isolated from the normal mouse mammary gland and human breast reduction tissues as well as the highly aggressive metaplastic and claudin-low breast tumor subtypes. This has implications for the origin of these breast tumors as it remains unclear whether they derive from cells that have undergone EMT or whether they represent an expansion of a pre-existing stem cell population that expresses EMT-associated markers to begin with. In the present review, we consider the current evidence connecting EMT and stem cell attributes and discuss the ramifications of these newly recognized links for our understanding of the emergence of distinct breast cancer subtypes and breast cancer progression.

Scale-Aware Transformers for Diagnosing Melanocytic Lesions.

Diagnosing melanocytic lesions is one of the most challenging areas of pathology with extensive intra- and inter-observer variability. The gold standard for a diagnosis of invasive melanoma is the examination of histopathological whole slide skin biopsy images by an experienced dermatopathologist. Digitized whole slide images offer novel opportunities for computer programs to improve the diagnostic performance of pathologists. In order to automatically classify such images, representations that reflect the content and context of the input images are needed. In this paper, we introduce a novel self-attention-based network to learn representations from digital whole slide images of melanocytic skin lesions at multiple scales. Our model softly weighs representations from multiple scales, allowing it to discriminate between diagnosis-relevant and -irrelevant information automatically. Our experiments show that our method outperforms five other state-of-the-art whole slide image classification methods by a significant margin. Our method also achieves comparable performance to 187 practicing U.S. pathologists who interpreted the same cases in an independent study. To facilitate relevant research, full training and inference code is made publicly available at https://github.com/meredith-wenjunwu/ScATNet.

Machine learning techniques for mitoses classification.

BACKGROUND: Pathologists analyze biopsy material at both the cellular and structural level to determine diagnosis and cancer stage. Mitotic figures are surrogate biomarkers of cellular proliferation that can provide prognostic information; thus, their precise detection is an important factor for clinical care. Convolutional Neural Networks (CNNs) have shown remarkable performance on several recognition tasks. Utilizing CNNs for mitosis classification may aid pathologists to improve the detection accuracy. METHODS: We studied two state-of-the-art CNN-based models, ESPNet and DenseNet, for mitosis classification on six whole slide images of skin biopsies and compared their quantitative performance in terms of sensitivity, specificity, and F-score. We used raw RGB images of mitosis and non-mitosis samples with their corresponding labels as training input. In order to compare with other work, we studied the performance of these classifiers and two other architectures, ResNet and ShuffleNet, on the publicly available MITOS breast biopsy dataset and compared the performance of all four in terms of precision, recall, and F-score (which are standard for this data set), architecture, training time and inference time. RESULTS: The ESPNet and DenseNet results on our primary melanoma dataset had a sensitivity of 0.976 and 0.968, and a specificity of 0.987 and 0.995, respectively, with F-scores of .968 and .976, respectively. On the MITOS dataset, ESPNet and DenseNet showed a sensitivity of 0.866 and 0.916, and a specificity of 0.973 and 0.980, respectively. The MITOS results using DenseNet had a precision of 0.939, recall of 0.916, and F-score of 0.927. The best published result on MITOS (Saha et al. 2018) reported precision of 0.92, recall of 0.88, and F-score of 0.90. In our architecture comparisons on MITOS, we found that DenseNet beats the others in terms of F-Score (DenseNet 0.927, ESPNet 0.890, ResNet 0.865, ShuffleNet 0.847) and especially Recall (DenseNet 0.916, ESPNet 0.866, ResNet 0.807, ShuffleNet 0.753), while ResNet and ESPNet have much faster inference times (ResNet 6 s, ESPNet 8 s, DenseNet 31 s). ResNet is faster than ESPNet, but ESPNet has a higher F-Score and Recall than ResNet, making it a good compromise solution. CONCLUSION: We studied several state-of-the-art CNNs for detecting mitotic figures in whole slide biopsy images. We evaluated two CNNs on a melanoma cancer dataset and then compared four CNNs on a public breast cancer data set, using the same methodology on both. Our methodology and architecture for mitosis finding in both melanoma and breast cancer whole slide images has been thoroughly tested and is likely to be useful for finding mitoses in any whole slide biopsy images.

Melanoma in the Blink of an Eye: Pathologists' Rapid Detection, Classification, and Localization of Skin Abnormalities.

Expert radiologists can quickly extract a basic "gist" understanding of a medical image following less than a second exposure, leading to above-chance diagnostic classification of images. Most of this work has focused on radiology tasks (such as screening mammography), and it is currently unclear whether this pattern of results and the nature of visual expertise underlying this ability are applicable to pathology, another medical imaging domain demanding visual diagnostic interpretation. To further characterize the detection, localization, and diagnosis of medical images, this study examined eye movements and diagnostic decision-making when pathologists were briefly exposed to digital whole slide images of melanocytic skin biopsies. Twelve resident (N = 5), fellow (N = 5), and attending pathologists (N = 2) with experience interpreting dermatopathology briefly viewed 48 cases presented for 500 ms each, and we tracked their eye movements towards histological abnormalities, their ability to classify images as containing or not containing invasive melanoma, and their ability to localize critical image regions. Results demonstrated rapid shifts of the eyes towards critical abnormalities during image viewing, high diagnostic sensitivity and specificity, and a surprisingly accurate ability to localize critical diagnostic image regions. Furthermore, when pathologists fixated critical regions with their eyes, they were subsequently much more likely to successfully localize that region on an outline of the image. Results are discussed relative to models of medical image interpretation and innovative methods for monitoring and assessing expertise development during medical education and training.