Automatic Ploidy Prediction and Quality Assessment of Human Blastocyst Using Time-Lapse Imaging.

Assessing fertilized human embryos is crucial for in vitro-fertilization (IVF), a task being revolutionized by artificial intelligence and deep learning. Existing models used for embryo quality assessment and chromosomal abnormality (ploidy) detection could be significantly improved by effectively utilizing time-lapse imaging to identify critical developmental time points for maximizing prediction accuracy. Addressing this, we developed and compared various embryo ploidy status prediction models across distinct embryo development stages. We present BELA (Blastocyst Evaluation Learning Algorithm), a state-of-the-art ploidy prediction model surpassing previous image- and video-based models, without necessitating subjective input from embryologists. BELA uses multitask learning to predict quality scores that are used downstream to predict ploidy status. By achieving an AUC of 0.76 for discriminating between euploidy and aneuploidy embryos on the Weill Cornell dataset, BELA matches the performance of models trained on embryologists' manual scores. While not a replacement for preimplantation genetic testing for aneuploidy (PGT-A), BELA exemplifies how such models can streamline the embryo evaluation process, reducing time and effort required by embryologists.

Wearable sensor-based performance status assessment in cancer: A pilot multicenter study from the Alliance for Clinical Trials in Oncology (A19_Pilot2).

Clinical performance status is designed to be a measure of overall health, reflecting a patient's physiological reserve and ability to tolerate various forms of therapy. Currently, it is measured by a combination of subjective clinician assessment and patient-reported exercise tolerance in the context of daily living activities. In this study, we assess the feasibility of combining objective data sources and patient-generated health data (PGHD) to improve the accuracy of performance status assessment during routine cancer care. Patients undergoing routine chemotherapy for solid tumors, routine chemotherapy for hematologic malignancies, or hematopoietic stem cell transplant (HCT) at one of four sites in a cancer clinical trials cooperative group were consented to a six-week prospective observational clinical trial (NCT02786628). Baseline data acquisition included cardiopulmonary exercise testing (CPET) and a six-minute walk test (6MWT). Weekly PGHD included patient-reported physical function and symptom burden. Continuous data capture included use of a Fitbit Charge HR (sensor). Baseline CPET and 6MWT could only be obtained in 68% of study patients, suggesting low feasibility during routine cancer treatment. In contrast, 84% of patients had usable fitness tracker data, 93% completed baseline patient-reported surveys, and overall, 73% of patients had overlapping sensor and survey data that could be used for modeling. A linear model with repeated measures was constructed to predict the patient-reported physical function. Sensor-derived daily activity, sensor-derived median heart rate, and patient-reported symptom burden emerged as strong predictors of physical function (marginal R2 0.429-0.433, conditional R2 0.816-0.822). Trial Registration: Clinicaltrials.gov Id NCT02786628.

Towards artificial intelligence-driven pathology assessment for hematological malignancies.

In this issue of Blood Cancer Discovery, Brück et al applied unsupervised and supervised machine learning to bone marrow histopathology images from Myelodysplastic Syndrome (MDS) patients. Their study provides new insights into the pathobiology of MDS and paves the way for increased use of artificial intelligence for the assessment and diagnosis of hematological malignancies.

Social and Clinical Determinants of COVID-19 Outcomes: Modeling Real-World Data from a Pandemic Epicenter.

IMPORTANCE: As the United States continues to accumulate COVID-19 cases and deaths, and disparities persist, defining the impact of risk factors for poor outcomes across patient groups is imperative. OBJECTIVE: Our objective is to use real-world healthcare data to quantify the impact of demographic, clinical, and social determinants associated with adverse COVID-19 outcomes, to identify high-risk scenarios and dynamics of risk among racial and ethnic groups. DESIGN: A retrospective cohort of COVID-19 patients diagnosed between March 1 and August 20, 2020. Fully adjusted logistical regression models for hospitalization, severe disease and mortality outcomes across 1-the entire cohort and 2- within self-reported race/ethnicity groups. SETTING: Three sites of the NewYork-Presbyterian health care system serving all boroughs of New York City. Data was obtained through automated data abstraction from electronic medical records. PARTICIPANTS: During the study timeframe, 110,498 individuals were tested for SARS-CoV-2 in the NewYork-Presbyterian health care system; 11,930 patients were confirmed for COVID-19 by RT-PCR or covid-19 clinical diagnosis. MAIN OUTCOMES AND MEASURES: The predictors of interest were patient race/ethnicity, and covariates included demographics, comorbidities, and census tract neighborhood socio-economic status. The outcomes of interest were COVID-19 hospitalization, severe disease, and death. RESULTS: Of confirmed COVID-19 patients, 4,895 were hospitalized, 1,070 developed severe disease and 1,654 suffered COVID-19 related death. Clinical factors had stronger impacts than social determinants and several showed race-group specificities, which varied among outcomes. The most significant factors in our all-patients models included: age over 80 (OR=5.78, p= 2.29x10-24) and hypertension (OR=1.89, p=1.26x10-10) having the highest impact on hospitalization, while Type 2 Diabetes was associated with all three outcomes (hospitalization: OR=1.48, p=1.39x10-04; severe disease: OR=1.46, p=4.47x10-09; mortality: OR=1.27, p=0.001). In race-specific models, COPD increased risk of hospitalization only in Non-Hispanics (NH)-Whites (OR=2.70, p=0.009). Obesity (BMI 30+) showed race-specific risk with severe disease NH-Whites (OR=1.48, p=0.038) and NH-Blacks (OR=1.77, p=0.025). For mortality, Cancer was the only risk factor in Hispanics (OR=1.97, p=0.043), and heart failure was only a risk in NH-Asians (OR=2.62, p=0.001). CONCLUSIONS AND RELEVANCE: Comorbidities were more influential on COVID-19 outcomes than social determinants, suggesting clinical factors are more predictive of adverse trajectory than social factors.

Incorporating cytologic adequacy assessment into precision oncology workflow using telepathology: An institutional experience.

BACKGROUND: Tumor sample quality and quantity determine the success of somatic mutation analysis. Thus, a rapid on-site evaluation (ROSE) tumor cytology adequacy assessment was incorporated into the workflow of precision oncology at Weill Cornell Medicine in New York City. Optimal samples were obtained from 68 patients with metastatic cancer. METHODS: Cytopathologists performed ROSE on fine-needle aspirate samples via telepathology, and subsequently core-needle biopsies were obtained. In a retrospective manner, the concordance between adequacy assessment and the success rate of the procedure was evaluated to obtain sufficient tumor tissue for next-generation sequencing (NGS). RESULTS: Out of the 68 procedures, 43 were documented as adequate and 25 were documented as inadequate. The diagnostic yield of adequate procedures was 100%. Adequacy evaluation predicted the success rate of molecular profiling in 40 of 43 procedures (93%; 95% CI, 80.9-98.5 procedures). The success rate of molecular testing was significantly higher in the adequate group: 93% compared with 32% in the inadequate group (P < .0005). Seven procedures that failed to provide quality material for mutational analysis and pathological diagnosis were evaluated as inadequate. Cell block provided sufficient DNA for NGS in 6 cases. In 2 cases, a core biopsy could not be performed; hence, the fine-needle aspirate material confirmed the diagnosis and was used for NGS testing. CONCLUSION: These results support the incorporation of ROSE into the workflow of precision oncology to obtain high-quality tissue samples from metastatic lesions. In addition, NGS testing of concurrent cytology specimens with adequate cellularity can be a surrogate for NGS testing of biopsy specimens.

Deep learning enables robust assessment and selection of human blastocysts after in vitro fertilization.

Visual morphology assessment is routinely used for evaluating of embryo quality and selecting human blastocysts for transfer after in vitro fertilization (IVF). However, the assessment produces different results between embryologists and as a result, the success rate of IVF remains low. To overcome uncertainties in embryo quality, multiple embryos are often implanted resulting in undesired multiple pregnancies and complications. Unlike in other imaging fields, human embryology and IVF have not yet leveraged artificial intelligence (AI) for unbiased, automated embryo assessment. We postulated that an AI approach trained on thousands of embryos can reliably predict embryo quality without human intervention. We implemented an AI approach based on deep neural networks (DNNs) to select highest quality embryos using a large collection of human embryo time-lapse images (about 50,000 images) from a high-volume fertility center in the United States. We developed a framework (STORK) based on Google's Inception model. STORK predicts blastocyst quality with an AUC of >0.98 and generalizes well to images from other clinics outside the US and outperforms individual embryologists. Using clinical data for 2182 embryos, we created a decision tree to integrate embryo quality and patient age to identify scenarios associated with pregnancy likelihood. Our analysis shows that the chance of pregnancy based on individual embryos varies from 13.8% (age ≥41 and poor-quality) to 66.3% (age <37 and good-quality) depending on automated blastocyst quality assessment and patient age. In conclusion, our AI-driven approach provides a reproducible way to assess embryo quality and uncovers new, potentially personalized strategies to select embryos.

Molecular diagnosis of autosomal dominant polycystic kidney disease using next-generation sequencing.

Autosomal dominant polycystic kidney disease (ADPKD) is caused by mutations in PKD1 and PKD2. However, genetic analysis is complicated by six PKD1 pseudogenes, large gene sizes, and allelic heterogeneity. We developed a new clinical assay for PKD gene analysis using paired-end next-generation sequencing (NGS) by multiplexing individually bar-coded long-range PCR libraries and analyzing them in one Illumina MiSeq flow cell. The data analysis pipeline has been optimized and automated with Unix shell scripts to accommodate variant calls. This approach was validated using a cohort of 25 patients with ADPKD previously analyzed by Sanger sequencing. A total of 250 genetic variants were identified by NGS, spanning the entire exonic and adjacent intronic regions of PKD1 and PKD2, including all 16 pathogenic mutations. In addition, we identified three novel mutations in a mutation-negative cohort of 24 patients with ADPKD previously analyzed by Sanger sequencing. This NGS method achieved sensitivity of 99.2% (95% CI, 96.8%-99.9%) and specificity of 99.9% (95% CI, 99.7%-100.0%), with cost and turnaround time reduced by as much as 70%. Prospective NGS analysis of 25 patients with ADPKD demonstrated a detection rate comparable with Sanger standards. In conclusion, the NGS method was superior to Sanger sequencing for detecting PKD gene mutations, achieving high sensitivity and improved gene coverage. These characteristics suggest that NGS would be an appropriate new standard for clinical genetic testing of ADPKD.

Faster sequencers, larger datasets, new challenges.

A report on the Advances in Genome Biology and Technology (AGBT) meeting, Marco Island, Florida, USA, 15-18 February 2012.