The effect of spatial resolution on deep learning classification of lung cancer histopathology.

Objective: The microscopic analysis of biopsied lung nodules represents the gold-standard for definitive diagnosis of lung cancer. Deep learning has achieved pathologist-level classification of non-small cell lung cancer histopathology images at high resolutions (0.5-2 µm/px), and recent studies have revealed tomography-histology relationships at lower spatial resolutions. Thus, we tested whether patterns for histological classification of lung cancer could be detected at spatial resolutions such as those offered by ultra-high-resolution CT. Methods: We investigated the performance of a deep convolutional neural network (inception-v3) to classify lung histopathology images at lower spatial resolutions than that of typical pathology. Models were trained on 2167 histopathology slides from The Cancer Genome Atlas to differentiate between lung cancer tissues (adenocarcinoma (LUAD) and squamous-cell carcinoma (LUSC)), and normal dense tissue. Slides were accessed at 2.5 × magnification (4 µm/px) and reduced resolutions of 8, 16, 32, 64, and 128 µm/px were simulated by applying digital low-pass filters. Results: The classifier achieved area under the curve ≥0.95 for all classes at spatial resolutions of 4-16 µm/px, and area under the curve ≥0.95 for differentiating normal tissue from the two cancer types at 128 µm/px. Conclusions: Features for tissue classification by deep learning exist at spatial resolutions below what is typically viewed by pathologists. Advances in knowledge: We demonstrated that a deep convolutional network could differentiate normal and cancerous lung tissue at spatial resolutions as low as 128 µm/px and LUAD, LUSC, and normal tissue as low as 16 µm/px. Our data, and results of tomography-histology studies, indicate that these patterns should also be detectable within tomographic data at these resolutions.

Breast cancer risk predictions by birth cohort and ethnicity in a population-based screening mammography program.

OBJECTIVES: To examine whether birth cohorts affect the risk of breast cancer for East Asian, First Nations, African, South Asian and Caucasian ethnicities in British Columbia (BC). METHODS: We used Cox PH models adjusted for well-known risk factors, such as age, breast density, mammographic features on false positives, and family history, to examine risk of breast cancer among East Asian, First Nations, African and South Asian ethnicities, relative to Caucasian, across three birth cohorts. RESULTS: There were 813,280 participants and 11,166 in situ and invasive breast cancer diagnoses. East Asians screened in BC were found to have a lower risk of breast cancer in the birth cohort born pre-1946 compared to Caucasian, but there was no statistically significant decrease for East Asians born after 1946. First Nations had an increased risk of breast cancer compared with Caucasian for all birth cohorts ranging from 1.1 to 2.0x the risk, which was statistically significant for those born after 1965. South Asians showed a statistically significant decrease in risk ranging from 0.58 to 0.81x lower compared with Caucasians for all birth cohorts. CONCLUSION: Risk of breast cancer for South Asians living in BC was found to be lower than Caucasians for each birth cohort examined, while East Asians had a comparable risk of breast cancer, First Nations had a consistently higher risk than Caucasians. ADVANCES IN KNOWLEDGE: When accounting for birth cohort, compared to Caucasians, South Asians have a decreased risk, First Nations have an increased risk, and East Asians have a similar risk of breast cancer.

A Semiautomated Chart Review for Assessing the Development of Radiation Pneumonitis Using Natural Language Processing: Diagnostic Accuracy and Feasibility Study.

BACKGROUND: Health research frequently requires manual chart reviews to identify patients in a study-specific cohort and examine their clinical outcomes. Manual chart review is a labor-intensive process that requires significant time investment for clinical researchers. OBJECTIVE: This study aims to evaluate the feasibility and accuracy of an assisted chart review program, using an in-house rule-based text-extraction program written in Python, to identify patients who developed radiation pneumonitis (RP) after receiving curative radiotherapy. METHODS: A retrospective manual chart review was completed for patients who received curative radiotherapy for stage 2-3 lung cancer from January 1, 2013 to December 31, 2015, at British Columbia Cancer, Kelowna Centre. In the manual chart review, RP diagnosis and grading were recorded using the Common Terminology Criteria for Adverse Events version 5.0. From the charts of 50 sample patients, a total of 1413 clinical documents were obtained for review from the electronic medical record system. The text-extraction program was built using the Natural Language Toolkit Python platform (and regular expressions, also known as RegEx). Python version 3.7.2 was used to run the text-extraction program. The output of the text-extraction program was a list of the full sentences containing the key terms, document IDs, and dates from which these sentences were extracted. The results from the manual review were used as the gold standard in this study, with which the results of the text-extraction program were compared. RESULTS: Fifty percent (25/50) of the sample patients developed grade ≥1 RP; the natural language processing program was able to ascertain 92% (23/25) of these patients (sensitivity 0.92, 95% CI 0.74-0.99; specificity 0.36, 95% CI 0.18-0.57). Furthermore, the text-extraction program was able to correctly identify all 9 patients with grade ≥2 RP, which are patients with clinically significant symptoms (sensitivity 1.0, 95% CI 0.66-1.0; specificity 0.27, 95% CI 0.14-0.43). The program was useful for distinguishing patients with RP from those without RP. The text-extraction program in this study avoided unnecessary manual review of 22% (11/50) of the sample patients, as these patients were identified as grade 0 RP and would not require further manual review in subsequent studies. CONCLUSIONS: This feasibility study showed that the text-extraction program was able to assist with the identification of patients who developed RP after curative radiotherapy. The program streamlines the manual chart review further by identifying the key sentences of interest. This work has the potential to improve future clinical research, as the text-extraction program shows promise in performing chart review in a more time-efficient manner, compared with the traditional labor-intensive manual chart review.

Target volume definition for post prostatectomy radiotherapy: Do the consensus guidelines correctly define the inferior border of the CTV?

AIM: We compare urethrogram delineation of the caudal aspect of the anastomosis to the recommended guidelines of post prostatectomy radiotherapy. BACKGROUND: Level one evidence has established the indications for, and importance of, adjuvant radiotherapy following radical prostatectomy. Several guidelines have recently addressed delineation of the prostate bed target volume including identification of the vesico-urethral anastomosis, taken as the first CT slice caudal to visible urine in the bladder neck. The inferior border of clinical target volume is then variably defined 5-12 mm below this anastomosis or 15 mm cranial to the penile bulb. METHODS AND MATERIALS: Thirty-three patients who received adjuvant radiotherapy following radical prostatectomy were reviewed. All underwent planning CT with urethrogram. The authors (MM, JC) independently identified the CT slice caudal to the last slice showing urine in the bladder neck (called the CT Reference Slice), and measured the distance between this and the tip of the urethrogram cone. Five patients also had a diagnostic MRI at the time of CT planning to better visualize the anatomy. RESULTS: Sixty-six readings were obtained. The mean distance between the Bladder CT Reference Slice and the most cranial urethrogram contrast slice was 16.1 mm (MM 16.4 mm, JC 15.8 mm), range: 6.8-34.2 mm. The mean distance between the urethrogram tip and the ischial tuberosities was 19.9 mm (range 12.5-29.8 mm). The mean distance between the CT Reference Slice and the ischial tuberosities was 36.9 mm (range 28.3-52.4 mm). CONCLUSIONS: Guidelines for prostate bed radiation post prostatectomy have been developed after publication of the trials proving benefit of such treatment, and are thus untested. The anastomosis is a frequent site of local relapse but is variably defined by the existing guidelines, none of which take into account anatomic patient variation and all of which are at variance with urethrogram data. We recommend the use of planning urethrogram to better delineate the vesico-urethral junction and minimize the potential for geographic misses.

Assessing the Effects of Participant Preference and Demographics in the Usage of Web-based Survey Questionnaires by Women Attending Screening Mammography in British Columbia.

BACKGROUND: Increased usage of Internet applications has allowed for the collection of patient reported outcomes (PROs) and other health data through Web-based communication and questionnaires. While these Web platforms allow for increased speed and scope of communication delivery, there are certain limitations associated with this technology, as survey mode preferences vary across demographic groups. OBJECTIVE: To investigate the impact of demographic factors and participant preferences on the use of a Web-based questionnaire in comparison with more traditional methods (mail and phone) for women participating in screening mammography in British Columbia, Canada. METHODS: A sample of women attending the Screening Mammography Program of British Columbia (SMPBC) participated in a breast cancer risk assessment project. The study questionnaire was administered through one of three modes (ie, telephone, mail, or website platform). Survey mode preferences and actual methods of response were analyzed for participants recruited from Victoria General Hospital. Both univariate and multivariate analyses were used to investigate the association of demographic factors (ie, age, education level, and ethnicity) with certain survey response types. RESULTS: A total of 1192 women successfully completed the study questionnaire at Victoria General Hospital. Mail was stated as the most preferred survey mode (509/1192, 42.70%), followed by website platform (422/1192, 35.40%), and telephone (147/1192, 12.33%). Over 80% (955/1192) of participants completed the questionnaire in the mode previously specified as their most preferred; mail was the most common method of response (688/1192, 57.72%). Mail was also the most preferred type of questionnaire response method when participants responded in a mode other than their original preference. The average age of participants who responded via the Web-based platform (age 52.9, 95% confidence interval [CI] 52.1-53.7) was significantly lower than those who used mail and telephone methods (age 55.9, 95% CI 55.2-56.5; P<.001); each decade of increased age was associated with a 0.97-fold decrease in the odds of using the website platform (P<.001). Web-based participation was more likely for those who completed higher levels of education; each interval increase leading to a 1.83 increase in the odds of website platform usage (P<.001). Ethnicity was not shown to play a role in participant preference for the website platform (P=.96). CONCLUSIONS: It is beneficial to consider participant survey mode preference when planning to collect PROs and other patient health data. Younger participants and those of higher education level were more likely to use the website platform questionnaire; Web-based participation failed to vary across ethnic group. Because mail questionnaires were still the most preferred survey mode, it will be important to employ strategies, such as user-friendly design and Web-based support, to ensure that the patient feedback being collected is representative of the population being served.

An Internet-Based Means of Monitoring Quality of Life in Post-Prostate Radiation Treatment: A Prospective Cohort Study.

BACKGROUND: Widespread integration of the Internet has resulted in an increase in the feasibility of using Web-based technologies as a means of communicating with patients. It may be possible to develop secure and standardized systems that facilitate Internet-based patient-reported outcomes which could be used to improve patient care. OBJECTIVE: This study investigates patient interest in participating in an online post-treatment disease outcomes and quality of life monitoring program developed specifically for patients who have received radiation treatment for prostate cancer at a regional oncology center. METHODS: Patients treated for prostate cancer between 2007 and 2011 (N=1113) at the British Columbia Cancer Agency, Centre for the Southern Interior were invited by mail to participate in a standardized questionnaire related to their post-treatment health. Overall participation rates were calculated. In addition, demographics, access to broadband Internet services, and treatment modalities were compared between participants and nonparticipants. RESULTS: Of the 1030 eligible invitees, 358 (358/1030, 34.7%) completed the online questionnaire. Participation rates were higher in individuals younger than age 60 when compared to those age 60 or older (42% vs 31%) and also for those living in urban areas compared with rural (37% vs 29%) and in those who received brachytherapy versus external beam radiotherapy (EBRT) (41% vs 31%). Better participation rates were seen in individuals who had access to Internet connectivity based on the different types of broadband services (DSL 35% for those with DSL connectivity vs 29% for those without DSL connectivity; cable 35% vs 32%; wireless 38% vs 26%). After adjusting for age, the model indicates that lack of access to wireless broadband connectivity, living in a rural area, and receiving EBRT were significant predictors of lower participation. CONCLUSIONS: This study demonstrates that participation rates vary in patient populations within the interior region of British Columbia, especially with older patients, those in rural areas, and those with limited access to quality Internet services.

Assessing the breast cancer risk distribution for women undergoing screening in British Columbia.

Breast cancer risk estimations are both informative and useful at the population level, with many screening programs relying on these assessments to allocate resources such as breast MRI. This cross-sectional multicenter study attempts to quantify the breast cancer risk distribution for women between the ages of 40 to 79 years undergoing screening mammography in British Columbia (BC), Canada. The proportion of women at high breast cancer risk was estimated by surveying women enrolled in the Screening Mammography Program of British Columbia (SMPBC) for known breast cancer risk factors. Each respondent's 10-year risk was computed with both the Tyrer-Cuzick and Gail risk assessment models. The resulting risk distributions were evaluated using the guidelines from the National Institute for Health and Care Excellence (United Kingdom). Of the 4,266 women surveyed, 3.5% of women between the ages of 40 to 79 years were found to have a high 10-year risk of developing breast cancer using the Tyrer-Cuzick model (1.1% using the Gail model). When extrapolated to the screening population, it was estimated that 19,414 women in the SMPBC are considered to be at high breast cancer risk. These women may benefit from additional MRI screening; preliminary analysis suggests that 4 to 5 additional MRI machines would be required to screen these high-risk women. However, the use of different models and guidelines will modify the number of women qualifying for additional screening interventions, thus impacting the MRI resources required. The results of this project can now be used to inform decision-making groups about resource allocation for breast cancer screening in BC.

A randomized trial comparing seed displacement of coated seeds to regular loose seeds at 30 days postimplant.

PURPOSE: To compare 30-day seed displacement and seed loss of standard loose seeds to specially engineered coated seeds. METHODS AND MATERIALS: Forty patients with prostate cancer were randomized and treated with either loose seeds or loose "coated" seeds. Implants were preplanned using transrectal ultrasound and performed using preloaded needles containing either standard or coated iodine-125 seeds according to randomization. Pelvic X-rays and CT were performed on Days 0 and 30 and a pelvic magnetic resonance scan on Day 30. Cranial-caudal displacement relative to the center of mass (COM) of the seed cloud of the six most peripheral basal and apical seeds was determined from Day 0 and 30 CT scans using custom software. Day 30 magnetic resonance-CT fusion was performed using a seed-to-seed match for soft tissue contouring on MRI. RESULTS: The mean displacement for the six basal seeds was 0.32 cm (standard deviation [SD], 0.25 cm) and 0.33 cm (SD, 0.27 cm) toward the COM for the regular and coated seeds, respectively (p = 0.35). For the apical seeds, mean displacement was 0.31 cm (SD, 0.35 cm) and 0.43 cm (SD, 0.26 cm) (p = 0.003) toward the COM. More regular seeds (n = 8) were lost from the apical region as compared with one coated seed (p = 0.015). There was a trend to reduction in total seeds lost: 1% for regular seeds as compared with 0.3% for coated seeds. CONCLUSIONS: Coated seeds were found to have a significant anchoring effect that was effective in reducing the number of apical seeds lost because of venous migration.

Defining a magnetic resonance scan sequence for permanent seed prostate brachytherapy postimplant assessment.

PURPOSE: We describe a magnetic resonance (MR) scan sequence for prostate brachytherapy postimplant assessment. METHODS AND MATERIALS: One brachytherapy team at the British Columbia Cancer Agency has incorporated MR-CT fusion into their permanent seed prostate brachytherapy quality assurance procedure. Several attempts were required to ensure that the diagnostic MR scanner at the adjoining general hospital performed the desired sequence, providing many examples of suboptimal scans and underlining the pitfalls for a center trying to incorporate the use of MR scanning into their brachytherapy program. RESULTS: The recommended sequence (Fast Spin Echo T2-weighted, repetition time [TR]/echo time [TE] 4500/90, echo train length [ETL] 10, 20×20 field of view [FOV], 80 bandwidth [BW]) is associated with superior edge detection when compared with those images in which a typical diagnostic sequence was used. The use of a low bandwidth sequence does not compromise edge detection or seed identification when compared with a higher bandwidth. CONCLUSIONS: We have defined a magnetic resonance imaging sequence, which appears to optimize both prostate delineation and identification of seeds, lending itself to straightforward fusion with CT images and allowing for less uncertainty in permanent seed prostate brachytherapy quality assurance.