Analysis of Artificial Intelligence-Based Automated Phenotyping of Coronary Artery Disease Diagnoses for Patients who Received Radiotherapy for Thoracic Malignancies

PURPOSE/OBJECTIVE(S)

Coronary artery disease (CAD) is a critical risk factor and toxicity endpoint in thoracic radiation oncology. The recent advent of artificial intelligence (AI) algorithms - specifically natural language processing (NLP) - allows for automated extraction of this data from electronic health records (EHR) more efficiently and at scale. We report on the application and performance characteristics of an EHR-based automated phenotyping tool to a cohort of lung cancer patients receiving radiation therapy (RT).

MATERIALS/METHODS

We applied to this radiation oncology use-case an AI-based disease phenotyping tool developed by other researchers for phenotyping CAD from the EHR. The CAD model was trained on 191,187 patients with at least one ICD code and validated with 158 gold-standard patients (Sensitivity: 73%, Specificity: 93%, PPV: 90%). NLP disease definitions were determined by UMLS and were employed with ICD codes through PheNorm classification method to create the phenotype algorithm (Yue et al., J Am Med Inform Assoc. 2018 Jan 1; 25(1):54-60). We applied the CAD phenotyping tool to two independent cohorts of RT treated lung cancer patients: 1) an annotated set of 677 patients who received RT from 1998 to 2014 and were studied in prior cardiac toxicity research; 2) an independent set of 209 patients who received RT from 1998 to 2018 and were annotated separately.

RESULTS

Only 7.7% of the AI training population had a CAD diagnosis code, compared to 32.5% and 65.0% in the 677 patient and 209 patient cohorts, respectively. Validation of the AI predicted CAD is shown in the table below. Performance on the 209 patient cohort best matched the AI training set. The 677-patient cohort differed from the 209-patient cohort in that its CAD definition included patients with radiographic coronary artery calcifications, which accounted for the difference in cohort sensitivities, F-scores, and NPVs.

CONCLUSION

The AI calculated CAD was somewhat accurate in detecting whether a patient had CAD and was very accurate in confirming no CAD. The differences in sensitivity and NPV reflect differences in defining CAD, but this study demonstrates the potential for automated AI-phenotyping to detect pre-treatment risk factors and cardiac toxicity events from the EHR.

Automated Sarcopenia Assessment and Outcomes in Head and Neck Cancer with Deep Learning Analysis of Cervical Neck Skeletal Muscle

PURPOSE/OBJECTIVE(S)

Sarcopenia is an established prognostic factor in patients diagnosed with head and neck cancers (HNC), typically measured by the skeletal muscle index (SMI) from abdominal muscle mass at L3. While sarcopenia assessment could inform HNC management, it remains impractical, time- and labor-intensive, and operator-dependent. To overcome these challenges, we developed an automated deep learning (DL) platform to calculate SMI at L3 by quantifying cross-sectional cervical skeletal muscle area (SMA) at C3 through auto-segmentation, externally validated it, and evaluated associations with clinical outcomes.

MATERIALS/METHODS

Eight hundred twenty-one patients diagnosed with HNC from multiple institutes from 1999-2013, treated with definitive chemoradiation with baseline pre-treatment CT scans, were included for model development (335 training, 96 tuning) and for independent testing (48 internal, and 342 external). Ground truth single-slice segmentations of SM at the mid-C3 vertebral level were manually annotated by radiation oncologists using an established protocol. A multi-stage DL pipeline was developed, with a 2D DenseNet to select the middle slice of C3 section and a 2D UNet to segment the SM, from which SMA was calculated. The model was evaluated using the Dice Similarity Coefficient (DC) for the internal test set, and human acceptability testing on the external test set was performed by two radiation oncologists not involved in annotations. SMI was calculated from C3 SMA based on prior literature, and sarcopenia was defined by an established, sex-specific SMI cutoff. Sarcopenia associations with overall survival (OS) and toxicities were assessed on the external dataset with Cox and logistic multivariable regressions, as indicated.

RESULTS

Model DC on the internal test set as 0.90 [95% CI: 0.90-0.91], with an intra-class coefficient of 0.96 for SMA. Human acceptability testing showed a pass rate of 94.4%. Of the 342 patients in the clinical analysis, 261 (76.3%) patients had sarcopenia. Five-year survival was 84.4% in patients without sarcopenia vs 73.1% in patients with sarcopenia (HR 2.21, p = 0.028) (median f/u: 44 mo (IQR: 25 - 66 mo)). On multivariable regression, sarcopenia (HR 2.06, p = 0.037), ACE-27 score 2+ (HR 2.25, p = 0.001), non-oropharynx diagnosis (HR 3.96, p<0.001), and T3-4 stage (HR 2.37, p<0.001) were associated with worse OS. Sarcopenia was associated with longer PEG tube duration on multivariable analysis (HR 1.59, p = 0.003), along with ACE-27 score (HR 1.20, p = 0.012) and non-oropharynx primary site (HR 1.46, p = 0.034). Sarcopenia was associated with higher risk of having PEG tube at last follow up (OR 2.25, p = 0.046). An observed increase in risk of hospitalization <3 months after RT was non-significant (OR 2.18, p = 0.117).

CONCLUSION

We developed and externally validated a fully-automated platform for sarcopenia assessment that can be used on routine HNC imaging. This algorithm is positioned for prospective testing to determine if use will inform HNC management.