Using Deep Learning to Automate Eosinophil Counting in Pediatric Ulcerative Colitis Histopathological Images

Background

Accurate identification of inflammatory cells from mucosal histopathology images is important in diagnosing ulcerative colitis. The identification of eosinophils in the colonic mucosa has been associated with disease course. Cell counting is not only time-consuming but can also be subjective to human biases. In this study we developed an automatic eosinophilic cell counting tool from mucosal histopathology images, using deep learning.

Method

Four pediatric IBD pathologists from two North American pediatric hospitals annotated 530 crops from 143 standard-of-care hematoxylin and eosin (H & E) rectal mucosal biopsies. A 305/75 split was used for training/validation to develop and optimize a U-Net based deep learning model, and 150 crops were used as a test set. The U-Net model was then compared to SAU-Net, a state-of-the-art U-Net variant. We undertook post-processing steps, namely, (1) the pixel-level probability threshold, (2) the minimum number of clustered pixels to designate a cell, and (3) the connectivity. Experiments were run to optimize model parameters using AUROC and cross-entropy loss as the performance metrics.

Results

The F1-score was 0.86 (95%CI:0.79-0.91) (Precision: 0.77 (95%CI:0.70-0.83), Recall: 0.96 (95%CI:0.93-0.99)) to identify eosinophils as compared to an F1-score of 0.2 (95%CI:0.13-0.26) for SAU-Net (Precision: 0.38 (95%CI:0.31-0.46), Recall: 0.13 (95%CI:0.08-0.19)). The inter-rater reliability was 0.96 (95%CI:0.93-0.97). The correlation between two pathologists and the algorithm was 0.89 (95%CI:0.82-0.94) and 0.88 (95%CI:0.80-0.94) respectively.

Conclusion

Our results indicate that deep learning-based automated eosinophilic cell counting can obtain a robust level of accuracy with a high degree of concordance with manual expert annotations.

Biomarkers for Breast Adenocarcinoma Using In Silico Approaches

This work elucidates the idea of finding probable critical genes linked to breast adenocarcinoma. In this study, the GEO database gene expression profile data set (GSE70951) was retrieved to look for genes that were expressed variably across breast adenocarcinoma samples and healthy tissue samples. The genes were confirmed to be part of the PPI network for breast cancer pathogenesis and prognosis. In Cytoscape, the CytoHubba module was used to discover the hub genes. For correlation analysis, the predictive biomarker of these hub genes, as well as GEPIA, was used. A total of 155 (85 upregulated genes and 70 downregulated genes) were identified. By integrating the PPI and CytoHubba data, the major key/hub genes were selected from the results. The KM plotter is employed to find the prognosis of those major pivot genes, and the outcome shows worse prognosis in breast adenocarcinoma patients. Further experimental validation will show the predicted expression levels of those hub genes. The overall result of our study gives the consequences for the identification of a critical gene to ease the molecular targeting therapy for breast adenocarcinoma. It could be used as a prognostic biomarker and could lead to therapy options for breast adenocarcinoma.

Moderately Hypofractionated Radiotherapy and Androgen Deprivation Therapy for High-risk Localised Prostate Cancer: Predictors of Long-term Biochemical Control and Toxicity

AIMS

There is a paucity of long-term data on outcomes of high-risk prostatic adenocarcinoma after moderately hypofractionated radiotherapy with elective nodal treatment and long-term androgen deprivation therapy (ADT). We report long-term control and toxicity outcomes and analyse the predictors of failure and toxicity.

MATERIALS AND METHODS

The records of 120 consecutive high-risk prostate cancer patients treated in a single institution between February 2012 and December 2016 were retrospectively analysed. A moderately hypofractionted radiotherapy (HypoRT) regimen of 60 Gy in 20 fractions over 4 weeks with simultaneous elective pelvic irradiation to 44 Gy in 20 fractions with intensity-modulated radiotherapy was used, together with long-term ADT with either orchiectomy or medical castration for a total duration of 2-3 years. We analysed biochemical control, metastasis-free survival and late toxicities and their predictive factors using survival analysis.

RESULTS

Patients had locally advanced cancers (cT3 77.5%, median pretreatment prostate-specific antigen 30 ng/ml, Gleason score 8-10 in 45.8%). The median follow-up time was 70 months. The 3- and 5-year probability of freedom from biochemical progression was 93% and 80%, respectively. The 5-year probability of freedom from local relapse/intra-pelvic nodal relapse/distant metastases as the site of first failure was 96%/97%/86%, respectively. Gleason score 8-10 and medical ADT for 2-3 years (as opposed to orchidectomy) were independent risk factors for distant metastases. A total of 18 grade 2 and above late gastrointestinal toxicity events and a total of 23 grade 2 and above late genitourinary toxicity events were documented. Patients who underwent a transurethral resection of prostate prior to radiotherapy had worse urological toxicity.

CONCLUSIONS

HypoRT with elective nodal treatment results in excellent pelvic control. Distant metastases are the primary mode of failure. Risk of metastases is associated with Gleason score and the duration of ADT. Late urinary toxicities are more common in those with prior transurethral resection of prostate.

Helical Radiotherapy in Early Laryngeal Cancers Could Lead to Excess Local Recurrence: Lessons From a Phase II Prospective Study

AIMS

A prospective study was conducted to investigate the feasibility and efficacy of carotid-sparing intensity-modulated radiotherapy (CSIMRT) in early glottic cancers (EGC).

MATERIALS AND METHODS

Eighteen patients underwent CSIMRT using helical tomotherapy to a dose of 55 Gy/20 fractions/4 weeks. Carotid intimal thickness (CIT) at prespecified carotid levels was measured using B-mode ultrasound at 6, 18 and 36 months. Serial changes in CIT were also measured in a control prospective cohort of 18 patients with head and neck cancers receiving bilateral neck nodal radiation over the same time period (54-60 Gy/30 fraction/6 weeks). The outcomes of 18 patients undergoing CSIMRT were compared against a retrospective consecutive cohort of 41 patients with EGC to confirm comparable local control.

RESULTS

No significant CIT differences were identified between patients undergoing CSIMRT versus the control group. However, four patients in the CSIMRT group had a local recurrence between 8 and 39 months. In all patients the epicentre of the recurrence was noted at the anterior part of the larynx. The 5-year local recurrence-free survival was 75.1% (95% confidence interval 56.6-99.7%). By contrast, in the group of EGC patients treated without carotid sparing, local recurrence was noted only in a single patient (patient treated with helical tomotherapy) and the 5-year local recurrence-free survival was 97.1% (95% confidence interval 91.8-100%) (Log-rank P = 0.01).

CONCLUSION

We failed to show the safety of CSIMRT using helical tomotherapy in this population of EGC patients. Use of CSIMRT also did not translate into a substantial reduction in CIT until 36 months. Use of CSIMRT using rotational arc techniques such as helical tomotherapy may be associated with a greater risk of local recurrence due to intrafractional motion interplay effects.

Molecular docking based virtual screening of carbonic anhydrase IX with coumarin (a cinnamon compound) derived ligands

It is of interest to design carbonic anhydrase IX (CAIX) inhibitors with improved features using molecular docking based virtual high through put screening of ligands. Coumarin (a cinnamon compound with pharmacological activity) is known as a potent phytal compound blocking tumor growth. Hence, a series of 17 coumarin derivatives were designed using the CHEMSKETCH software for docking analysis with CAIX. The catalytic site analysis of CAIX for binding with ligand molecules was completed using the SCHRODINGER package (2009). Thus, 17 ligands with optimal binding features with CAIX were selected following the calculation of ADME/T properties. We report ligands #41, #42, #19 and #15 showed good docking score, glide energy and hydrogen bond interactions without vdW clash. We further show that N-(3,4,5-trimethoxy-phenylcarbamoylmethyl) designated as compound #41 have the highest binding energy (-61.58) with optimal interactions with the catalytic residues (THR 199, PRO 201, HIS 119, HIS 94) of CAIX.

A Phase I/II Study of Stereotactic Hypofractionated Once-weekly Radiation Therapy (SHORT) for Prostate Cancer

AIMS

Stereotactic radiation therapy has been investigated predominantly in patients with low-intermediate-risk disease. We conducted a clinical trial of stereotactic hypofractionated radiation therapy delivered in once-weekly fractions on patients with all-risk non-metastatic disease to test feasibility, acute toxicities and patient-reported outcomes.

MATERIALS AND METHODS

In this phase I/II study, 30 patients with prostatic adenocarcinoma, any Gleason score, T1-4N0 and prostate-specific antigen ≤60 ng/ml were treated with volumetric intensity modulated arc radiation therapy to a dose of 35 Gy in five fractions delivered once weekly. Patients with high-risk disease also received elective nodal irradiation to a dose of 25 Gy in five fractions simultaneously. Androgen deprivation was offered to intermediate- and high-risk patients. The primary outcome was acute toxicity. Secondary outcome measures included biochemical control and late toxicity. Patient-reported outcomes were measured using the International Prostate Symptom Score and European Organization for Research and Treatment of Cancer (EORTC) Quality-of-Life Questionnaire (QLQ).

RESULTS

All 30 patients completed treatment per-protocol. Most patients had T3 (60%) and Gleason 7 (50%) tumours. The median prostate-specific antigen was 17 ng/ml. High-risk disease was present in 20 patients (66.7%). There was a low incidence of acute toxicities (grade 2 + urinary 3.3%, grade 2 rectal 0%). Within the EORTC QLQ framework, only the urinary symptom score showed a clinically meaningful worsening from a mean of 20/100 at baseline to 34/100 at the end of treatment (P < 0.001), but reduced to 24/100 at 6 months (P = 0.08). With a median follow-up of 41.5 months, two patients each reported grade 2 late urinary and rectal toxicity. The 3- and 4-year biochemical control rates were 96.7 and 87.9%, respectively.

CONCLUSION

In a cohort of mainly high-risk cancers, stereotactic once-weekly radiation therapy was easy to implement and well tolerated, with a low incidence of acute and late toxicity and excellent biochemical control.

Generation of fashionable clothes using generative adversarial networks

Purpose

There are various style options available when one buys clothes on online shopping websites, however the availability the new fashion trends or choices require further user interaction in generating fashionable clothes. The paper aims to discuss this issue.

Design/methodology/approach

Based on generative adversarial networks (GANs) from the deep learning paradigm, here the authors suggest model system that will take the latest fashion trends and the clothes bought by users as input and generate new clothes. The new set of clothes will be based on trending fashion but at the same time will have attributes of clothes where were bought by the consumer earlier.

Findings

In the proposed machine learning based approach, the clothes generated by the system will personalized for different types of consumers. This will help the manufacturing companies to come up with the designs, which will directly target the customer.

Research limitations/implications

The biggest limitation of the collected data set is that the clothes in the two domains do not belong to a specific category. For instance the vintage clothes data set has coats, dresses, skirts, etc. These different types of clothes are not segregated. Also there is no restriction on the number of images of each type of cloth. There can many images of dresses and only a few for the coats. This can affect the end results. The aim of the paper was to find whether new and desirable clothes can be created from two different domains or not. Analyzing the impact of “the number of images for each class of cloth” is something which is aim to work in future.

Practical implications

The authors believe such personalized experience can increase the sales of fashion stores and here provide the feasibility of such a clothes generation system.

Originality/value

Applying GANs from the deep learning models for generating fashionable clothes.

Continuous hyperfractionated accelerated radiotherapy using modern radiotherapy techniques for nonsmall cell lung cancer patients unsuitable for chemoradiation

INTRODUCTION

The continuous hyperfractionated and accelerated radiotherapy (CHART) regimen of radiotherapy (RT) for nonsmall cell lung cancer is underused outside the UK. We present the first Indian experience of using CHART for patients, who were not suitable for chemotherapy or concurrent chemo-RT.

METHODS

We retrospectively reviewed the data of patients treated using CHART at our institution between January 2014 and December 2015.

RESULTS

Thirty-seven patients were treated using CHART. Planning methods and dosimetry parameters are described. Three-dimensional conformal RT was used for treatment planning and delivery in 23 patients and volumetric modulated arc RT was necessary for 14 patients. Patients in our series had a median age of 70 years (interquartile range 65.50-74.00) and 86.5% had Stage III disease. Median follow-up was short at 13.0 months. Actuarial rates of 1-year progression-free survival, 1-year overall survival (OS), and 2-year OS were 31.9%, 59.5%, and 28.5%, respectively. This treatment was well tolerated with manageable and some reversible acute esophageal toxicity (91.9% CONCLUSION: Our results indicate that CHART is feasible, safe, and well tolerated in Indian patients who are clinically found to be not suitable for either sequential or concurrent chemo- RT.

Actual gains in dosimetry and treatment delivery efficiency from volumetric modulated arc radiotherapy for inoperable, locally advanced lung cancer over five-field forward-planned intensity-modulated radiotherapy

AIMS

Volumetric modulated arc radiotherapy (VMAT) is used for inoperable, locally advanced nonsmall cell lung cancer, where three-dimensional conformal radiotherapy (3D-CRT) cannot yield an acceptable plan.

METHODS

The planning and treatment data were prospectively collected on the first 18 patients treated using VMAT plans. We analyzed the actual dosimetric gain and impact on treatment, compared with complex multisegment 3D-CRT (five-field forward-planned intensity-modulated radiotherapy [IMRT]) that were generated for treatment. Proportion of planning target volume (PTV) receiving 95% dose (PTV-V95%) conformity index (CI), conformity number (CN), dose homogeneity index (DHI), monitor units (MUs), and treatment time were also analyzed.

RESULTS

The PTV coverage (PTV-V95%) was improved from a median of 91.41% for 5-F forward-IMRT to 98.25% for VMAT (P < 0.001). The CI improved with a mean of 1.12 for VMAT and 1.31 for 5-F forward-IMRT (P < 0.001). The mean DHI improved from 1.15 for forward-IMRT to 1.08 for VMAT (P < 0.001). The mean CN improved from 0.62 for forward-IMRT to 0.87 for VMAT (P < 0.001). No significant increase in the low-dose bath (V5, V10 and mean lung dose) to the lung was seen. Significantly higher number of MUs (P < 0.001) and shorter treatment delivery times (P = 0.03) were seen with VMAT.

CONCLUSION

VMAT resulted in improvement in target volume coverage, demonstrated by PTV-V95%, CI, CN, and DHI, without any increase in the low-dose bath to the lung. For conventional fractionation, VMAT requires more MUs (P < 0.001) but has a shorter treatment delivery time (P = 0.03) per fraction.

Automatic polyp detection in pillcam colon 2 capsule images and videos: preliminary feasibility report

Background. The aim of this work is to present an automatic colorectal polyp detection scheme for capsule endoscopy. Methods. PillCam COLON2 capsule-based images and videos were used in our study. The database consists of full exam videos from five patients. The algorithm is based on the assumption that the polyps show up as a protrusion in the captured images and is expressed by means of a P-value, defined by geometrical features. Results. Seventeen PillCam COLON2 capsule videos are included, containing frames with polyps, flat lesions, diverticula, bubbles, and trash liquids. Polyps larger than 1 cm express a P-value higher than 2000, and 80% of the polyps show a P-value higher than 500. Diverticula, bubbles, trash liquids, and flat lesions were correctly interpreted by the algorithm as nonprotruding images. Conclusions. These preliminary results suggest that the proposed geometry-based polyp detection scheme works well, not only by allowing the detection of polyps but also by differentiating them from nonprotruding images found in the films.