Detailed Image Data Quality and Cleaning Practices for Artificial Intelligence Tools for Breast Cancer.

Standardizing image-data preparation practices to improve accuracy/consistency of AI diagnostic tools.

Quality of Anatomic Staging of Breast Carcinoma in Hospitals in the United States, With Focus on Measurement of Tumor Dimension.

OBJECTIVES: We investigated the accuracy of clinical breast carcinoma anatomic staging and the greatest tumor dimension measurements. METHODS: We compared clinical stage and greatest dimension values with the pathologic reference standard values using 57,747 cases from the 2016 US National Cancer Institute Surveillance, Epidemiology, and End Results program who were treated by surgical resection without prior neoadjuvant therapy. RESULTS: Agreement for clinical vs pathologic anatomic TNM group stage, overall, is 74.3% ± 0.4%. Lymph node N staging overall agrees very well (85.1% ± 0.4%). Based on tumor dimension and location, T staging has an agreement of only 64.2% ± 0.4%, worsening to 55% without carcinoma in situ (Tis) cases. In approximately 25% of cases, pathologic T stage is higher than clinical T stage. The mean difference in the greatest dimension is 1.36 ± 9.59 mm with pathologic values being generally larger than clinical values; pathologic and clinical measurements correlate well. T-stage disagreement is associated with histology, tumor grade, tumor size, N stage, patient age, periodic biases in tumor size measurements, and overuse of family T-stage categories. Pathologic measurement biases include rounding and specimen-slicing intervals. CONCLUSIONS: Clinical and pathologic T-staging values agree only moderately. Pathologists face challenges in increasing the precision of gross tumor measurements, with the goal of improving the accuracy of clinical T staging and measurement.

Clinical Non-Small Cell Lung Cancer Staging and Tumor Length Measurement Results From U.S. Cancer Hospitals.

RATIONALE AND OBJECTIVES: Examine the accuracy of clinical non-small cell lung cancer staging and tumor length measurements, which are critical to prognosis and treatment planning. MATERIALS AND METHODS: Compare clinical and pathological staging and lengths using 10,320 2016 National Cancer Institute Surveillance, Epidemiology, and End Results (SEER) and 559 2010-2018 non-SEER single-institute surgically-treated cases, and analyze modifiable causes of disagreement. RESULTS: The SEER clinical and pathological group-stages agree only 62.3% ± 0.9% over all stage categories. The lymph node N-stage agrees much better at 83.0% ± 1.0%, but the tumor length-location T-stage agrees only 57.7% ± 0.8% with approximately 29% of the cases having a greater pathology than clinical T-stage. Individual T-stage category agreements with respect to the number of pathology cases are Tis, T1a, T1b, T2a, T2b, T3, T4: 89.9% ± 10.0%; 78.7% ± 1.7%; 51.8% ± 1.9%; 46.1% ± 1.3%; 40.5% ± 3.1%; 44.1% ± 2.2%; 56.4% ± 4.7%, respectively. Most of the single-institute results statistically agree with SEER's. Excluding Tis cases, the mean difference in SEER tumor length is ∼1.18 ± 9.26 mm (confidence interval: 0.97-1.39 mm) with pathological lengths being longer than clinical lengths except for small tumors; the two measurements correlate well (Pearson-r >0.87, confidence interval: 0.86-0.87). Reasons for disagreement include the use of family-category descriptors (e.g., T1) instead of their subcategories (e.g., T1a and T1b), which worsens the T-stage agreement by over 15%. Disagreement is also associated with higher tumor grade, larger resected specimens, higher N-stage, patient age, and periodic biases in clinical and pathological tumor size measurements. CONCLUSIONS: By including preliminary non-small cell lung cancer clinical stage values in their evaluation, diagnostic radiologists can improve the accuracy of staging and standardize tumor-size measurements, which improves patient care.

Simplified, standardized methods to assess the accuracy of clinical cancer staging.

BACKGROUND: Hospitals lack intuitive methods to monitor their accuracy of clinical cancer staging, which is critical to treatment planning, prognosis, refinements, and registering quality data. METHODS: We introduce a tabulation framework to compare clinical staging with the reference-standard pathological staging, and quantify systematic errors. As an example, we analyzed 9,644 2016 U.S. National Cancer Institute SEER surgically-treated non-small cell lung cancer (NSCLC) cases, and computed concordance with different denominators to compare with incompatible past results. RESULTS: The concordance for clinical versus pathological lymph node N-stage is very good, 83.4 ± 1.0%, but the tumor length-location T-stage is only 58.1 ± 0.9%. There are intuitive insights to the causes of discordance. Approximately 29% of the cases are pathological T-stage greater than clinical T-stage, and 12% lower than the clinical T-stage, which is due partly to the fact that surgically-treated NSCLC are typically lower-stage cancer cases, which results in a bounded higher probability for pathological upstaging. Individual T-stage categories Tis, T1a, T1b, T2a, T2b, T3, T4 invariant percent-concordances are 85.2 ± 9.7 + 10.3%; 72.7 ± 1.6 + 11.3%; 46.6 ± 1.8 + 10.9%; 54.6 ± 1.6 - 20.5%; 41.6 ± 3.3 - 0.1%; 54.7 ± 2.8 - 24.1%; 55.2 ± 4.7 + 2.6%, respectively. Each percent-concordance is referenced to an averaged number of pathological and clinical cases. The first error number quantifies statistical fluctuations; the second quantifies clinical and pathological staging biases. Lastly, comparison of over and under staging versus clinical characteristics provides further insights. CONCLUSIONS: Clinical NSCLC staging accuracy and concordance with pathological values can improve. As a first step, the framework enables standardizing comparing staging results and detecting possible problem areas. Cancer hospitals and registries can implement the efficient framework to monitor staging accuracy.