Malignant Mesothelioma of the Pleura. The Reproducibility of the Immunohistological Diagnosis

https://doi.org/10.1016/S0344-0338(97)80054-4Get rights and content

Summary

The reproducibility of the histopathological diagnosis of pleural malignant mesothelioma (MM), after supplementing routine H&E stain by immunohistochemistry (IH) in 77 cases of original diagnoses of MM, was assessed by examining interobserver variation between five pathologists. A battery of commercial antibodies (cytokeratins, vimentin, HMFG-2, anti Leu-M 1 [CD15], BerEP4, B72.3 [TAG-72], carcinoembyonic antigen), considered to be useful in enhancing diagnostic accuracy, was used. The number of definitively classified tumors (accepted MM plus rejected MM) increased from 57 on H&E stain to 60 after IH, with 59 (76.6%) cases being accepted as true MM. Based on IH, the chance-adjusted interobserver agreement was poor (κw = 0.29) and lower than that observed on previous H&E alone. The intraobserver agreement for four of the five pathologists was rather good (κw = 0.54–0.56). The inter- and intraobserver concordance was higher in accepting than excluding the cases as MM. A larger number of cases were classified by all reviewers as mixed or sarcomatous variants after IH. In the interpretation of each immunostain, x values ranged from 0.19 for B72.3 to 0.62 for HMFG-2, which were respectively the least and the most consistently interpreted immunostains. The information additionally contributed by IH did not seem to change the pathologists' diagnoses very much in comparison with those made by routine H&E stain. Until highly specific and sensitive probes for the positive identification of MM become available, a careful scrutiny of routinely stained preparations still remains the most rewarding component of the diagnostic pathway.

Referencess (26)

  • BattiforaH et al.

    Tumors of the serosal membranes

  • BedrossianCWM et al.

    Differential diagnosis between mesothelioma and adenocarcinoma: a multimodal approach based on ultrastructure and immunocytochemistry

    Semin Diagn Pathol

    (1992)
  • CicchettiDV

    Assessing inter-rater reliability for rating scales: resolving some basic issues

    Br J Psychiatry

    (1976)
  • Cited by (30)

    • Automated classification of cells into multiple classes in epithelial tissue of oral squamous cell carcinoma using transfer learning and convolutional neural network

      2020, Neural Networks
      Citation Excerpt :

      These may be the make/model and status of the microscope, the lighting conditions of observation or data acquisition, the quality of the stains/slides, experience of the pathologist, time devoted for every observation and so on. All such factors may lead to diagnosis error or delay, affecting follow up action (Betta et al., 1997). Whereas automated systems applying traditional machine learning entail analysis of all features, deep learning eliminates the need for domain expertise and hardcore feature engineering.

    • A shape context fully convolutional neural network for segmentation and classification of cervical nuclei in Pap smear images

      2020, Artificial Intelligence in Medicine
      Citation Excerpt :

      According to The Bethesda System (TBS) [2] multi-class nomenclature, the cervical cells are grouped into NILM (Negative for Intra-epithelial malignancy), LSIL (Low-grade Squamous Intraepithelial Lesion), HSIL (High-grade Squamous Intraepithelial Lesion) and SCC (Squamous cell carcinoma) based on the degree of nuclear enlargement, hyper chromaticity, and other cytological changes. Class identification and quantification of cervical cancer from Pap smear images, when done manually by a pathologist, is a tedious, error-prone and time-consuming process [3]. Considering the already highly skewed ratio of pathologists/patients in underdeveloped and developing countries, like India, where the study is located, this appears to be a challenging task.

    View all citing articles on Scopus
    View full text