Sammanfattning
Idiopathic pulmonary fibrosis (IPF) is a chronic interstitial lung disease of a dismal prognosis. While IPF is a rare disease, it is still the most common of idiopathic interstitial pneumonias. The radiological and histopathological manifestation of the disease is the usual interstitial pneumonia (UIP) pattern. The etiology of IPF is unknown, but inorganic dust is one of the known risk factors for IPF. The diagnosis of IPF is usually based on clinical and radiological data, but a surgical lung biopsy is required for a minority of patients. Confirming the diagnosis can be challenging as many interstitial lung diseases share similar features, and interobserver variation between radiologists and pathologists is significant. Separating IPF from other interstitial lung diseases is crucial due to differences in treatment and prognosis. In this doctoral thesis, we hypothesized that histopathological features in IPF lung tissue would be associated with survival and lung function. In addition, we aimed at investigating interobserver agreement among pathologists, inorganic particulate matter (PM) in the lung tissue of patients with IPF, and the use of artificial intelligence (AI) in analyzing lung tissue samples of IPF patients. Our study cohort originated from a prospective, multicenter registry study, namely the FinnishIPF registry. We searched for patients with available histological lung tissue samples and compared the histopathological features to the registry data. In Study I, four pathologists experienced in pulmonary pathology re-evaluated 60 lung tissue samples using the 2011 diagnostic criteria of IPF. They also recorded atypical histopathological features for IPF. Most of the samples were re-evaluated as definite UIP (38/60, 63%). The most common atypical feature for IPF was abundant inflammation (15/60, 25%). Using Cohen’s κ coefficient, the interobserver agreement varied from slight to substantial (κ=0.04-0.78); the variation might be partly causative of differences in the interpretation of the presence of giant cells. Radiologically definite UIP associated with a poor survival. However, the histopathological UIP pattern or atypical features for IPF were not associated with survival. In Study II, we focused on inorganic PM in 73 IPF lung tissue samples. We developed a semiquantitative scoring method (0-5) for coal dust pigment and inorganic PM using polarizing light microscopy. PM scores were compared to clinical, population density, and air quality data. An energy dispersive spectrometry with a field emission scanning electron microscope was used to analyze the elemental compositions of six IPF lung tissue samples. There were high scores of inorganic PM in the samples from southern Finnish university hospital districts compared to the samples from northern districts (31/50, 62% vs. 7/23, 30%, p=0.02). The highest scores of 4 and 5 were connected to an exposure to inorganic dust (n=15, p=0.004). Aluminum, silicon, and potassium were found in all six samples. In Study III, we tested AI in the analysis of histopathological features in IPF samples. With 20 different IPF samples, we developed an AI model using a convolutional neural network in Aiforia® platform. The AI model was taught to recognize alveolar parenchyma from the lung tissue, fibroblast foci (FF), interstitial mononuclear inflammation, and intra-alveolar macrophages. The samples of 71 IPF patients were analyzed with the model. The high area of FF was associated with a poor survival (p=0.01), and we found that high amounts of interstitial mononuclear inflammation and intra-alveolar macrophages were associated with a prolonged survival (p=0.01 and p=0.01, respectively). FF and intra-alveolar macrophages also had a link with lung function. High numbers of FF were associated with a low diffusing capacity for carbon monoxide (p=0.03), whereas a high intra-alveolar macrophage density was associated with a high forced vital capacity of predicted (p=0.03). In conclusion, FF seem to be the most potent single histological prognostic markers of survival in IPF. Of the other markers, inflammatory cells appeared to predict a prolonged survival. The interobserver agreement on the histopathological features of IPF varied, and especially the interpretation of giant cells seemed to cause a discrepancy. Inorganic PM in the lung tissue of IPF patients was not associated with the survival. Instead, the histological PM could reflect the level of exposure to air pollution. In the prognostic evaluation of the histopathological features in IPF lung tissue samples, AI could function as a future tool.
Originalspråk | engelska |
---|---|
Handledare |
|
Utgivningsort | Helsinki |
Förlag | |
Tryckta ISBN | 978-951-51-7270-9 |
Elektroniska ISBN | 978-951-51-7271-6 |
Status | Publicerad - 2021 |
MoE-publikationstyp | G5 Doktorsavhandling (artikel) |
Bibliografisk information
M1 - 97 s. + liitteetVetenskapsgrenar
- 3121 Allmänmedicin, inre medicin och annan klinisk medicin