RESUMO
Objetivo Evaluar la variabilidad interobservador en el uso de la versión 2.0 del PI- RADS (PI-RADS v2), en lectores experimentados y no experimentados. Materiales y Métodos Estudio retrospectivo de análisis de concordancia de lectores. Entre enero de 2015 y diciembre de 2016, 1.656 sujetos fueron estudiados mediante resonancia magnética multiparamétrica (RMmp) de próstata en nuestra institución. Se estimó la distribución porcentual del reporte en categoría PI-RADS v2, y con esa información, se realizó una selección de 150 casos con esquema de aleatorización estratificada a las distribuciones porcentuales de cada categoría. Dichos casos fueron anonimizados, presentados a tres lectores con cinco, cuatro y dos años de experiencia en lectura de RMmp además de tener más de un año de experiencia en el uso del PI- RADS v2 siendo leídos en forma individual. Los datos resultantes fueron analizados en forma independiente por un cuarto investigador. Resultados El valor de kappa ponderado para los observadores fue de 0,69 (IC 95: 0,64 a 0,75). La mayor concordancia correspondió a los lectores de mayor experiencia, donde alcanza un valor de 0,72 (IC 95%: 0,69 a 0,76). La concordancia entre los valores PI-RADS que determinan seguimiento o bien una intervención de acuerdo a elementos clínicos (1-2-3) y conducta activa (4-5) correspondió a 0,70 (IC 95%: 0,59 a 0,78). Discusión Se logró demostrar un acuerdo sustancial entre radiólogos utilizando el PI- RADS v2 para la detección en RMmp de lesiones sospechosas, mayor entre los dos lectores más experimentados. Sin embargo, la comparación del lector de menor experiencia con los de mayor experiencia también presentó una importante concordancia. Los valores de concordancia entre observadores para PI-RADS 2:4 fueron similares a los reportados en la literatura. Conclusiones El PI-RADSv2 ha demostrado en nuestro centro, con radiólogos dedicados a imágenes de abdomen y estudios de próstata, un alto nivel de acuerdo en la interpretación de la RMmp de próstata, encontrándose a tono con lo reportado en la literatura.
Purpose To evaluate the interobserver variability in the use of the PI-RADS 2.0 version, in experienced and non-experienced readers. Material and Methods Retrospective studyof readers' concordance analysis. Between January 2015 and December 2016, 1656 subjects were studied through multiparametric MRI (RMmp) of prostate in our institution. The percentage distribution of the report was estimated in each PI-RADS category, and a selection of 150 cases with a stratified randomization scheme was made to the percentage distributions of every category. These cases were anonymized, presented to three readers with 5, 4 and 2 years' experience in reading RMmp, and more than one-year experience with PI-RADS v2, and were read individually. The resulting data were analyzed independently by a fourth investigator. Results The weighted kappa value for the observers was 0.69 (IC 95: 0.64 to 0.75). The highest agreement corresponded to the two most experienced readers, where it reached a value of 0.72 (95% CI: 0.69 to 0.76). The concordance between the PI-RADS values that determine follow-up (1-2-3) and active behavior (4-5) corresponded to 0.70 (95% CI: 0.59 to 0.78). Discussion It was possible to demonstrate a substantial agreement between radiologists using the PI-RADS v2 for the detection in RMmp of suspicious lesions, greater among the two most experienced readers. However, the comparison of the less experienced reader with those of greater experience also presented an important concordance. The inter-observer concordance values for PI-RADS 2:4 were like those reported in the literature. Conclusions The PI-RADS v2 has demonstrated, in our center, with radiologists dedicated to abdominal images and prostate studies, a high level of agreement in the interpretation of prostate MRmp.
RESUMO
In the last decades, surgical treatment of breast cancer has enormously changed. As a result, nipple-sparing mastectomy (NSM) has evolved as an oncologically safe and cosmetic approach. NSM includes a subareolar frozen section to evaluate malignancy. We determined the accuracy of subareolar frozen section diagnosis, analyzed the discrepancy factor, and estimated the interobserver agreement of frozen section in NSM. A retrospective review of all NSMs at our institution from 2009 to 2015 was performed. Frozen sections were compared to the final diagnoses to analyze the accuracy of subareolar frozen sections. Discordant results were rigorously evaluated to identify discrepancy factors. Some cases were randomly chosen to assess the interobserver agreement (kappa) among pathologists. The agreement results were evaluated with and without knowledge of the tumor morphology. Among 34 NSMs, the frozen section false-negative and false-positive rate was 5.9% and 8.8%, respectively. The sensitivity and specificity was 77.8% and 88.0%, respectively. Sampling errors and diathermy artifacts explained our false-negative diagnoses. Freezing artifacts and an intraductal papilloma explained our false-positive diagnoses. The interobserver agreement between breast and general pathologists was 0.87 (p<0.0001) and 0.31 (p=0.0001), respectively. The interobserver agreement increased to 0.35 (p<0.0001) in general pathologists with knowledge of the tumor morphology. Subareolar frozen section showed to be a specific test with moderate sensitivity. Papillary lesions can mimic atypical cells and influence the frozen section interpretation. Frozen section in NSM had a better performance in breast pathologists (almost perfect) versus general pathologists (fair). Interobserver agreement may improve with knowledge of tumor morphology.
Assuntos
Neoplasias da Mama/patologia , Secções Congeladas , Mamilos/patologia , Adulto , Idoso , Carcinoma Intraductal não Infiltrante , Feminino , Secções Congeladas/métodos , Humanos , Masculino , Mastectomia Subcutânea/métodos , Pessoa de Meia-Idade , Variações Dependentes do Observador , Estudos RetrospectivosRESUMO
PURPOSE: To identify the main difficulties in postoperative clinical target volume (CTV) delineation in gastric cancer (GC). METHODS: Before and after a training course, 20 radiation oncology residents were asked to delineate the CTV for the postoperative GC case on four computed tomography scans: dome of the diaphragm, anterior abdominal wall, duodenal stump and porta hepatis level, and to determine the lower CTV border. CTV volume was reconstructed from requested planar contours. Area of intersection (AI) for each requested scan and volume of intersection (VI), defined as the overlap of delineated area/volume with respective reference area (RA)/reference volume (RV) proposed by the senior radiation oncologist, were computed. The degree of agreement between the reference and participants' contours was quantified using the Concordance Index (CI): AI/RA × 100% or VI/RV × 100%. The lower CTV border was analyzed separately. Pre- and post-training CIs were compared. A questionnaire investigated the difficulties with contouring. RESULTS: Mean CI value was the lowest for the dome of the diaphragm (24% pre-training, 35 % post-training) and for the duodenal stump (49% pre-training, 61% post-training). Mean CI for the CTV volume was 49% pre-training and 59% post-training, p = 0.39. Mean distance from the reference to the participants' lower CTV borders was 2.73 cm pre-training and 2.0 cm post-training, p = 0.71. In a questionnaire, 75% of respondents indicated the elective nodal area as the main difficulty. CONCLUSIONS: Delineation of the dome of the diaphragm and the duodenal stump, as yet not recognized as the source of variation, should be addressed in the international consensus guidelines and clarified.